AI security course glossary

Every acronym and key term used in the Offensive Security and AI Red Teaming for People Systems course, with its full name and a plain-language definition. 2055 terms.

#

0-RTT — Zero round-trip time
TLS 1.3 early data sent on a resumed connection before the handshake completes; faster, but it can be replayed. Appears in module 0.

A

A/B test —
A randomized experiment that assigns units to get a change or not, so the difference estimates its effect. Appears in module 45.
AA — Identity Management, Authentication, and Access Control
CSF 2.0 category PR.AA: who can access what, and how they prove who they are. Appears in module 12.
AAAA — IPv6 address record
A DNS record that maps a name to an IPv6 address. Appears in modules 0, 2.
AAAI — Association for the Advancement of Artificial Intelligence
A scientific society that runs one of the main annual AI research conferences. Appears in module 7.
AAAI/ACM — Association for the Advancement of Artificial Intelligence / Association for Computing Machinery
The two societies that jointly run the AI, Ethics, and Society (AIES) conference. Appears in modules 7, 8, 16.
AAIR — Advanced in AI Risk
An ISACA credential on AI risk for people who already hold a qualifying risk, audit, security or privacy designation. Appears in module 46.
AAL — Authentication Assurance Level
NIST SP 800-63-4's measure of how strongly a sign-in proves the user controls authenticators bound to their account, from AAL1 to AAL3 (phishing-resistant, non-exportable key). Appears in module 25.
AAL1 — Authenticator Assurance Level 1
The lowest of NIST SP 800-63's three login-strength levels. Appears in module 25.
AAL2 — Authenticator Assurance Level 2
The middle of NIST SP 800-63's three login-strength levels; it requires two authentication factors. Appears in module 25.
AAL3 — Authenticator Assurance Level 3
The highest of NIST SP 800-63's three login-strength levels; it requires a phishing-resistant hardware-based authenticator. Appears in module 25.
AAR — After-Action Report
A document recording what happened in an incident or exercise, what worked, and what will change. Appears in modules 20, 43.
ABAC — Attribute-Based Access Control
Access decisions based on attributes and conditions, such as tags or time. Appears in module 17.
Ablation
Removing one component to measure how much it contributes. Appears in modules 8, 19, 23, 36.
ABLE — Actor, Behavior, Location, Evidence
PEAK's table for scoping a hypothesis-driven hunt (expansion as described in the PEAK articles). Appears in module 32.
Abuse monitoring —
A provider's processes to detect misuse of a model service, which may store samples of prompts and involve human review. Appears in module 31.
AC — Attack Complexity
CVSS metric for whether built-in security measures must be defeated. Appears in modules 6, 27.
AC — Access Control
SP 800-53 control family on limiting who and what can access systems, such as AC-6 least privilege. Appears in modules 12, 14, 25.
Accelerator
A chip built for the large matrix calculations in neural networks, such as a GPU. Appears in modules 17, 37.
ACDC — Automatic Circuit DisCovery
An algorithm that automates finding the subset of a model's components (a circuit) that implements a chosen behavior. Appears in module 38.
ACE — Access control entry
One line in an access control list, granting, denying or auditing a specific right for a specific identity. Appears in module 0.
ACH — Analysis of Competing Hypotheses
A structured method that ranks explanations by how much evidence is inconsistent with each. Appears in module 33.
ACK — Acknowledgment
A TCP flag that confirms receipt of data; also used in some scan types. Appears in modules 0, 2.
ACL — Access Control List
A list attached to an object that says who may do what to it. Appears in modules 0, 3, 7, 8, 18, 23, 31.
ACM — Association for Computing Machinery
A large professional society for computing that publishes journals and runs conferences. Appears in modules 7, 8, 9, 15, 16, 22, 24, 27, 32, 34, 36, 37, 39, 42, 43, 44.
Activity group —
In the Diamond Model, a cluster of similar intrusion activity that may or may not belong to one actor. Appears in module 33.
Activity thread —
In the Diamond Model, the ordered chain of events that make up one intrusion. Appears in module 33.
AD — Active Directory
Microsoft's directory service that stores users, computers, groups and permissions and runs sign-in for a Windows domain. Appears in modules 0, 3, 17, 23, 99.
ADA — Americans with Disabilities Act
US federal law prohibiting discrimination against people with disabilities, including in employment. Appears in modules 13, 19, 25, 30, 34, 36, 99.
Adaptive attack —
An attack designed with full knowledge of a specific defense and aimed at its weak point. Appears in modules 8, 9, 11, 20, 38.
ADEA — Age Discrimination in Employment Act
US federal law protecting workers aged 40 and over from age discrimination. Appears in modules 12, 13, 14, 21, 22.
Adequacy decision
A European Commission decision under GDPR Article 45 that a third country protects personal data adequately, so transfers need no extra safeguards. Appears in module 36.
ADM — Architecture Development Method
TOGAF's iterative cycle of phases for developing and governing an enterprise architecture. Appears in module 40.
Admission control
A check that runs before deployment and blocks anything that does not meet policy, such as unsigned models. Appears in modules 17, 18, 20, 31.
ADMT — Automated Decision-Making Technology
Colorado SB26-189's term for technology that processes personal data to produce outputs used to make, guide or assist decisions about individuals. Appears in modules 12, 19, 36.
ADR — Architecture Decision Record
A short, append-only document recording one architecture decision with its status, context, decision and consequences. Appears in module 44.
Adversarial example —
An input changed on purpose, often only slightly, so that a model gives a wrong or attacker-chosen answer. Appears in modules 9, 10, 38.
Adversary emulation
Reproducing the documented behavior of a specific real threat actor, using threat intelligence, to test defenses against it. Appears in modules 10, 29, 39.
Adverse action —
An unfavorable decision, such as a credit denial or an employment decision based on a consumer report. Appears in modules 12, 16, 23, 25, 36.
Adverse impact
A substantially different selection rate that works to the disadvantage of a race, sex or ethnic group. Appears in modules 1, 7, 8, 11, 12, 13, 14, 16, 18, 19, 20, 21, 22, 23, 24, 25, 26, 28, 30, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 99.
advisory board —
A group that advises management but has no legal authority to govern the company. Appears in modules 40, 47.
AE — Adverse Event Analysis
CSF 2.0 category DE.AE: analyzing anomalies and possible attacks. Appears in modules 12, 20.
AEAD — Authenticated encryption with associated data
Encryption that protects confidentiality and integrity together, and can also authenticate extra unencrypted data such as a record ID. Appears in modules 0, 34.
AEDT — Automated Employment Decision Tool
Under LL144, a computer-based tool using machine learning, statistics, data analytics or AI that substantially assists or replaces discretionary employment decisions. Appears in modules 12, 13, 14, 21, 22, 24, 26, 46.
AERA — American Educational Research Association
One of the three organizations that publish the Standards for Educational and Psychological Testing. Appears in modules 13, 16.
AES — Advanced Encryption Standard
The modern block cipher used for Kerberos encryption types 0x11 and 0x12. Appears in modules 0, 3, 34.
AFL — American Fuzzy Lop
An influential coverage-guided fuzzer by Michał Zalewski; AFL++ is its community-maintained fork. Appears in module 27.
Agent — LLM agent
A system that repeatedly calls an LLM, lets it choose tools, runs them and feeds results back. Appears in modules 7, 8, 15, 16, 23.
AGPL — GNU Affero General Public License
A copyleft software license that also requires sharing source code of modified versions offered over a network. Appears in modules 27, 33.
AHP — Analytic Hierarchy Process
An MCDA method, developed by Thomas Saaty, that derives weights from pairwise comparisons and checks their consistency. Appears in module 40.
AI — Artificial Intelligence
Computer systems that perform tasks such as ranking, predicting or generating text that normally need human judgment. Appears in modules 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 99.
AI 600-1 — NIST AI 600-1, Generative Artificial Intelligence Profile
NIST's July 2024 profile applying the AI RMF to generative AI risks. Appears in modules 12, 13, 14, 20, 24, 29, 30, 33, 45.
AI BOM — AI Bill of Materials
An inventory of models, adapters, datasets and related components. Appears in modules 8, 18.
AI literacy —
The skills and understanding needed to use AI systems in an informed way and to be aware of their risks. Appears in modules 12, 44, 45, 99.
AI RMF — Artificial Intelligence Risk Management Framework
NIST's voluntary framework (NIST AI 100-1, 2023) for managing AI risks, organized into the GOVERN, MAP, MEASURE and MANAGE functions. Appears in modules 12, 13, 14, 18, 19, 20, 23, 24, 26, 29, 30, 33, 36, 37, 43, 45, 47, 99.
AICPA — American Institute of Certified Public Accountants
The US professional body for accountants that publishes the attestation standards and the Trust Services Criteria behind SOC reports. Appears in module 12.
AIES — AAAI/ACM Conference on AI, Ethics, and Society
A peer-reviewed conference on the social and ethical aspects of AI. Appears in module 22.
AIGP — Artificial Intelligence Governance Professional
An AI governance certification offered by the IAPP. Appears in modules 12, 22, 23, 24, 36, 39, 45, 99.
AIID — AI Incident Database
A public collection of reports of harms and failures involving AI systems. Appears in module 33.
AIMS — Artificial Intelligence Management System
The policies, processes and records an organization uses to manage AI, the subject of ISO/IEC 42001. Appears in modules 12, 45.
AIRC — Trustworthy and Responsible AI Resource Center
NIST's website for AI RMF resources and playbooks. Appears in module 12.
AIRT — Microsoft AI Red Team
Microsoft's internal team that red teams AI products, and a family of PyRIT scenarios named after it. Appears in module 11.
AIVIA — Artificial Intelligence Video Interview Act
Illinois law (820 ILCS 42) requiring notice, explanation, consent, sharing limits and deletion when AI analyzes video interviews. Appears in module 19.
AKA — Also known as
Introduces another name for the same thing. Appears in module 32.
ALE — Annualized Loss Exposure
The expected loss per year from a risk; percentiles of the loss distribution matter as much. Appears in modules 12, 43.
Alert fatigue —
The loss of attention and accuracy that follows when people handle many alerts that turn out not to matter. Appears in module 32.
Alert suppression —
Raising one alert per entity per window instead of one per matching event. Appears in modules 29, 32.
Algorithmic discrimination
Unlawful differential treatment or impact caused by the use of an AI system. Appears in module 12.
ALL —
In Kubernetes security contexts, the keyword that drops every Linux capability from a container. Appears in module 17.
Allocative harm
Harm from unequal distribution of opportunities or resources, such as lower scores for equally qualified candidates. Appears in module 29.
Allow-list
A rule that permits only known-good values and rejects everything else. Appears in modules 0, 4, 5, 8, 10, 15, 17, 23, 27, 28, 35.
AM — Asset Management
CSF 2.0 category ID.AM: knowing what systems, data and software you have. Appears in module 12.
AMF — Annual Maintenance Fee
ISC2's yearly fee to keep its certifications active. Appears in module 39.
AML — Adversarial Machine Learning
The study of attacks on machine learning systems and the defenses against them. Appears in modules 8, 9, 10, 11, 28, 29, 30, 32, 33, 99.
AN — Analytic (ATT&CK ID prefix)
Platform-specific detection logic within an ATT&CK detection strategy. Appears in module 10.
AN — Incident Analysis
CSF 2.0 category RS.AN: investigating what happened during an incident. Appears in modules 12, 20, 30.
Analytic
In ATT&CK v18 and later, a specific piece of detection logic that implements part of a detection strategy. Appears in modules 10, 29.
Anchoring
The pull that a first number, such as a first offer, has on later judgments. Appears in modules 1, 26, 41, 46, 47.
Annex III — Annex III of the EU AI Act
The list of high-risk AI use areas; point 4 covers employment and workers' management. Appears in modules 12, 14, 19, 21, 22, 23, 24, 29, 31, 32, 35, 40, 44, 45, 46, 47, 99.
Anonymization
Processing data so that people are no longer identifiable by any means reasonably likely to be used. Appears in modules 7, 10, 19, 24, 36.
AP — Application (AP) exchange
The Kerberos step in which a client presents a service ticket to a service (KRB_AP_REQ and KRB_AP_REP). Appears in module 3.
APA — American Psychological Association
One of the three organizations that publish the Standards for Educational and Psychological Testing. Appears in modules 13, 16, 22, 47.
APAC — Asia-Pacific
The region covering East Asia, South Asia, Southeast Asia and Oceania. Appears in module 31.
API — Application programming interface
A defined way for programs to talk to each other. Appears in modules 0, 1, 2, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 20, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 44, 99.
API inventory —
A maintained list of every API host, version, environment and data flow, including third-party integrations. Appears in modules 4, 6.
API4 — API4:2023 Unrestricted Resource Consumption
The OWASP API Security Top 10 item on missing limits for CPU, memory, records or paid calls. Appears in modules 4, 18, 25.
APIs — Application Programming Interfaces
Plural of API: defined ways for programs to talk to each other. Appears in modules 0, 1, 4, 5, 6, 7, 8, 9, 11, 14, 15, 16, 17, 23, 24, 25, 31, 33, 34, 35, 38, 39, 44.
APK — Android Package
The file format used to distribute and install Android apps. Appears in module 35.
Applicant reference —
In NIST SP 800-63A-4, a representative who can vouch for an applicant's identity or circumstances without acting on the applicant's behalf. Appears in module 25.
Approval fatigue
People asked to approve many prompts start approving without reading, weakening human-in-the-loop controls. Appears in module 15.
AR — Availability Requirement
CVSS v4.0 environmental metric for how much the organization needs the system to be available. Appears in module 6.
Argon2id —
The primary variant of Argon2, a memory-hard password hashing function (RFC 9106) recommended by OWASP for new systems. Appears in module 0.
ARN — Amazon Resource Name
AWS's unique identifier for a resource or principal. Appears in module 17.
ARP — Address Resolution Protocol
The protocol that maps IP addresses to hardware (MAC) addresses on a local network; its cache is volatile evidence. Appears in module 30.
AS — Authentication Service
The KDC function that issues TGTs after checking a client's identity. Appears in module 3.
AS-REP Roasting — AS-REP Roasting (ATT&CK T1558.004)
Obtaining Kerberos replies for accounts without pre-authentication to guess their passwords offline. Appears in module 3.
ASA — American Statistical Association
The main professional society for statisticians in the United States. Appears in module 13.
ASan — AddressSanitizer
A compiler instrumentation and run-time library that reports out-of-bounds accesses, use-after-free and similar memory errors when they happen. Appears in module 27.
ASCII — American Standard Code for Information Interchange
A basic character encoding for English letters, digits and symbols. Appears in module 17.
ASI — Agentic Security Initiative
The OWASP GenAI Security Project initiative that publishes guidance on threats to AI agents. Appears in module 15.
ASL — AI Safety Level
The set of safeguard standards (for example ASL-3) in Anthropic's Responsible Scaling Policy. Appears in module 38.
ASLR — Address Space Layout Randomization
Placing code and data at random memory addresses each run so attackers cannot predict where things are. Appears in modules 5, 6, 27, 35.
ASQC — American Society for Quality Control
The former name of ASQ, the professional body for quality management. Appears in module 32.
ASR — Attack Success Rate
The share of attempts in which an attack achieves its goal. Appears in modules 8, 9, 11, 14, 15, 16, 23, 29, 38.
ASRs — Attack Success Rates
Plural of ASR: the share of attack attempts that achieve their goal. Appears in modules 8, 11, 14, 16, 38.
Assessment center
A method in which candidates do several job-like exercises observed and rated by several trained assessors. Appears in modules 16, 41, 47.
Associated data —
Data that is authenticated together with a ciphertext but not encrypted, used to bind it to its context. Appears in modules 0, 34.
ASVS — Application Security Verification Standard
OWASP's list of verifiable security requirements for applications. Appears in modules 4, 8.
AT — Attack Requirements
CVSS v4.0 metric for deployment or execution conditions needed for the attack. Appears in modules 2, 6, 27.
AT — Awareness and Training
SP 800-53 control family on security training for staff. Appears in module 12.
AT-C — Attestation Standards codification (AT-C sections)
The section numbering of the AICPA attestation standards; SOC 2 examinations are performed under AT-C 105 and 205. Appears in module 12.
ATLAS — Adversarial Threat Landscape for Artificial-Intelligence Systems
MITRE's public knowledge base of tactics and techniques used against AI and machine-learning systems. Appears in modules 1, 5, 7, 8, 9, 10, 11, 12, 14, 15, 16, 20, 23, 24, 26, 28, 29, 30, 32, 33, 39, 46, 99.
Atomic Red Team
Red Canary's open library of small tests mapped to ATT&CK techniques, defined in YAML files. Appears in module 29.
Atomic test
A small, self-contained test that performs one ATT&CK technique in one way, as in Atomic Red Team. Appears in module 29.
ATR — Agent Threat Rules
An open rule format for AI-agent security threats, listed in the SigmaHQ README, whose tool can export rules to Sigma. Appears in module 32.
ATS — Applicant tracking system
Software that manages job applications and candidates. Appears in modules 0, 1, 2, 4, 5, 6, 7, 8, 12, 14, 15, 16, 17, 18, 20, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 36, 37, 38, 40, 41, 42, 43, 44, 45.
ATT — Authorization to Test
The signed letter stating who may test which systems, when and from where. Appears in module 14.
ATT&CK — Adversarial Tactics, Techniques, and Common Knowledge
MITRE's public catalog of the tactics and techniques attackers use, with IDs such as T1558. Appears in modules 1, 5, 6, 9, 10, 15, 17, 20, 28, 29, 30, 31, 32, 33, 35, 39, 99.
ATT&CK — MITRE ATT&CK
A public knowledge base of observed adversary tactics and techniques. Appears in modules 2, 3, 23.
ATT&CK Navigator
MITRE's web application for annotating ATT&CK matrices with layers, for example to show detection coverage. Appears in modules 10, 29, 33.
Attack advantage
In membership inference, the true-positive rate minus the false-positive rate; zero means no leakage detected. Appears in module 23.
Attack surface
All the points where an attacker could enter, cause an effect or extract data. Appears in modules 0, 1, 2, 4, 5, 6, 7, 10, 18, 27, 35, 37, 40, 99.
Attack tree —
A diagram that breaks an attacker's goal into the alternative and combined steps that could achieve it. Appears in module 1.
Attention
A mechanism that weighs earlier tokens by relevance when processing each token. Appears in modules 7, 37, 38.
attestation —
A signed report from a device, such as TPM PCR values, that lets a remote party decide whether to trust it. Appears in modules 34, 35.
Attestation engagement
An engagement in which an independent practitioner gives an opinion on a subject matter or someone's statement about it. Appears in module 12.
Attribute condition —
In Google Cloud workload identity federation, an expression that must be true for an external token to be accepted. Appears in module 31.
AU — Automatable
A CVSS v4.0 supplemental metric: can an attacker automate the attack across many targets? Appears in module 6.
AU — Audit and Accountability
SP 800-53 control family on logging events and protecting logs, such as AU-2. Appears in modules 12, 14, 25.
AUC — Area Under the ROC Curve
A single number summarizing how well a classifier separates two groups across all thresholds. Appears in modules 9, 13, 32, 38.
Audit log —
A record of who did what, to which object and when, inside an application. Appears in modules 1, 5, 8, 10, 17, 18, 20, 25, 28, 29, 30, 31, 32, 34, 35, 40, 99.
AUP — Acceptable Use Policy
A platform's rules on what users may and may not do. Appears in modules 23, 39.
Automation bias
The tendency to accept automated suggestions without enough scrutiny. Appears in modules 7, 8, 12, 19, 45.
AV — Attack Vector
CVSS metric for how remote the attacker can be: Network, Adjacent, Local or Physical. Appears in modules 2, 6.
AVID — AI Vulnerability Database
An open database of AI failure modes and vulnerabilities, named in NIST AI 600-1 as a reporting resource. Appears in module 20.
AWQ — Activation-aware Weight Quantization
A 4-bit weight quantization method that protects the small share of weight channels that matter most, chosen by looking at activations. Appears in module 37.
AWS — Amazon Web Services
Amazon's cloud computing platform. Appears in modules 10, 17, 30, 31, 34, 39, 99.
Azure Policy —
The Azure service that evaluates resource requests and resources against rules whose effects include deny, audit, modify and deployIfNotExists. Appears in modules 17, 31.

B

BAC — Budget at Completion
The total approved budget for a project. Appears in module 40.
Backdoor
Hidden model behavior activated only by a trigger input. Appears in modules 8, 9, 18, 38.
Backend — Sigma backend
A pySigma plugin that writes a parsed Sigma rule as a query in one target language, such as SPL or SQL. Appears in module 32.
Backtest —
Running a detection over past data to estimate how often it would have fired before deploying it. Appears in module 32.
BACS — Building Automation and Control Systems
A broader term for building automation, used in the draft NIST SP 800-82 Revision 4. Appears in module 35.
BARS — Behaviorally anchored rating scale
A rating scale where each point is described by a concrete example of behavior at that level, introduced by Smith and Kendall in 1963. Appears in modules 26, 47.
BAS — Breach and Attack Simulation
Products that automatically and repeatedly run attacker techniques to check that defenses still work. Appears in module 29.
Base rate
How common a condition really is in the population being tested. Appears in modules 2, 3, 6, 8, 10, 13, 16, 20, 21, 22, 23, 24, 25, 26, 28, 32, 33, 38.
Baseline — Baseline configuration
The recorded model, settings, prompt and flags that all test results refer to. Appears in module 14.
BATNA — Best Alternative To a Negotiated Agreement
What you will do if a negotiation fails; your strength in a negotiation comes largely from it. Appears in modules 26, 41, 47.
BBQ — Bias Benchmark for QA
A hand-built question-answering benchmark that measures social bias across nine social dimensions relevant to US English-speaking contexts. Appears in module 37.
BCBS — Basel Committee on Banking Supervision
The international committee of bank supervisors that sets global standards for banking regulation. Appears in module 46.
BCMS — Business Continuity Management System
The organization-wide management system that ISO 22301 specifies. Appears in module 12.
BCP — Business Continuity Plan
A plan to sustain business processes during and after a disruption. Appears in module 12.
BCP — Best Current Practice
An IETF document series that records recommended operational practice. Appears in modules 0, 17, 25, 30.
bcrypt —
An older, deliberately slow password hashing function with a tunable work factor and a 72-byte input limit. Appears in module 0.
BE — Base and Environmental (CVSS v4.0 nomenclature)
A CVSS v4.0 score label meaning the Base metrics plus the Environmental metrics were used. Appears in modules 2, 6.
Bearer token —
A credential that grants access to whoever holds it, without further proof of identity; SCIM integrations commonly use them. Appears in modules 0, 4, 17, 25.
Beaver triple —
Precomputed secret-shared values a, b and a times b, used to multiply secret-shared numbers in MPC. Appears in module 34.
BEC — Business Email Compromise
Fraud in which an attacker uses a compromised or impersonated email account to trick staff into sending money or information. Appears in module 33.
Behavioral baseline —
A description of an entity's or group's normal behavior, usually as a distribution of a measure over time. Appears in module 32.
Benchmark contamination
Test items or close variants appearing in a model's training data, which distorts benchmark scores. Appears in modules 16, 37.
Benjamini–Hochberg procedure —
A method for flagging results from many tests while controlling the false discovery rate. Appears in module 32.
BERT — Bidirectional Encoder Representations from Transformers
An early transformer model that reads text in both directions; widely used to produce embeddings. Appears in module 7.
Best-of-N — Best-of-N jailbreaking
An attack that tries many random variants of a prompt and succeeds if any one works. Appears in modules 16, 38.
BF16 — bfloat16 (brain floating point, 16-bit)
A 2-byte number format with the same exponent range as FP32 but fewer digits of precision; common for training and serving LLMs. Appears in module 37.
BFLA — Broken Function Level Authorization
An API flaw where a caller can use a function, such as an admin export, that its role should never reach (API5:2023). Appears in modules 4, 18.
BFV — Brakerski/Fan-Vercauteren
A homomorphic encryption scheme for exact arithmetic on encrypted integers. Appears in module 34.
BGV — Brakerski-Gentry-Vaikuntanathan
A homomorphic encryption scheme for exact arithmetic on encrypted integers. Appears in module 34.
BH — Benjamini–Hochberg procedure
A step-up method for many tests that controls the false discovery rate and is more powerful than family-wise corrections. Appears in modules 21, 24.
BIA — Business Impact Analysis
The analysis that identifies and ranks systems and processes by the impact of their disruption. Appears in module 12.
Bias audit — Bias audit (LL144)
An impartial evaluation by an independent auditor including selection or scoring rates and impact ratios by sex, race/ethnicity and intersectional categories. Appears in modules 4, 7, 12, 13, 14, 16, 18, 21, 22, 23, 24, 25, 26, 29, 34, 37, 38, 40, 42, 44, 46, 47.
biometric —
A measurable physical or behavioral characteristic, such as a fingerprint or face, used to recognize a person. Appears in modules 13, 16, 19, 25, 35, 36, 37, 39, 40.
BIPA — Biometric Information Privacy Act
Illinois law (740 ILCS 14) regulating collection and use of biometric identifiers such as face geometry and voiceprints. Appears in modules 19, 25, 35, 36.
BIPIA — Benchmark for Indirect Prompt Injection Attacks
A research benchmark of indirect prompt-injection test cases used to compare attacks and defenses. Appears in module 8.
Birthday bound —
The fact that collisions among n-bit digests become likely after about 2^(n/2) attempts, which halves a hash's collision resistance. Appears in modules 0, 34.
BISO — Business Information Security Officer
A security leader embedded with a business unit who links that unit's needs with central security standards. Appears in modules 42, 43.
Blast radius
How much an attacker can reach from a single compromised component. Appears in modules 0, 1, 5, 15, 17, 20, 27, 31, 34, 37.
Blind baseline —
An attack that never looks at the model; if it does as well as a real attack, the benchmark is measuring something other than model leakage. Appears in module 38.
Blind index —
A keyed hash stored beside an encrypted field so equal values can be matched without decryption. Appears in module 34.
BLS — Bureau of Labor Statistics
The U.S. Department of Labor agency that publishes official data on jobs, pay and prices. Appears in module 26.
BLUF — Bottom Line Up Front
A writing rule: state the main judgment and the decision you need first. Appears in module 33.
board skills matrix —
A table of the skills and experience a board needs, rated for each director, used to find gaps; a job analysis of the board as a team. Appears in module 47.
BOD — Binding Operational Directive
A directive from CISA that U.S. federal civilian executive branch agencies must follow, such as BOD 22-01 on known exploited vulnerabilities. Appears in module 6.
BoK — Body of Knowledge
The published list of topics an exam covers. Appears in modules 24, 99.
BOLA — Broken Object Level Authorization
An API flaw where the server does not check that the caller may act on the specific object named in the request (API1:2023). Appears in modules 4, 17, 18, 25.
BOM — Bill of Materials
A list of every component in a product. In software and AI, it lists libraries, models and datasets with versions. Appears in modules 8, 18.
BOMs — Bills of Materials
Plural of BOM. Appears in module 18.
Bonferroni correction —
Dividing the significance level by the number of tests; strict, and very conservative with many tests. Appears in modules 11, 21, 32.
bootloader —
The small program that runs first after power-on and loads the next stage of software. Appears in module 35.
Bootstrapping —
The FHE operation that refreshes a noisy ciphertext so computation can continue; it is expensive. Appears in module 34.
BOPLA — Broken Object Property Level Authorization
An API flaw where the server does not control which fields of an object a caller may read or change (API3:2023). Appears in module 4.
Bottom line — Bottom line (Uniform Guidelines)
The result of the total selection process for a job, as opposed to its components. Appears in modules 13, 21.
BPDA — Backward Pass Differentiable Approximation
An adaptive-attack technique: run a defense's non-differentiable step normally, but use a smooth stand-in (often the identity) when computing gradients. Appears in module 38.
BPE — Byte Pair Encoding
A method for building a subword vocabulary by repeatedly merging frequent symbol pairs. Appears in modules 7, 37.
BPF — Berkeley Packet Filter
A small in-kernel filter language that Linux seccomp uses to express system-call rules. Appears in module 27.
Break-glass account —
An emergency administrator account kept offline and monitored, used only when normal SSO access fails. Appears in module 25.
Bridge letter — Bridge (gap) letter
An unaudited letter from a vendor's management covering the time between the end of a SOC 2 period and a recent date. Appears in module 12.
Brier score —
The average squared gap between stated probabilities and outcomes (1 or 0); lower means better calibrated, more accurate judgments. Appears in modules 0, 33, 39.
BSD — Berkeley Software Distribution
A family of Unix-like operating systems that began at the University of California, Berkeley. Appears in module 35.
BSides — Security BSides
Independently organized, community security conferences held in many cities. Appears in modules 24, 39.
BT — Base and Threat (CVSS v4.0 nomenclature)
A CVSS v4.0 score label meaning the Base metrics plus the Threat metrics were used. Appears in modules 2, 6, 8.
BTE — CVSS Base, Threat and Environmental score
A CVSS v4.0 score using all three scoring groups. Appears in modules 2, 6.
BU — Business unit
A part of a company that runs a distinct business with its own leader and, usually, its own P&L. Appears in module 42.
Buffer overflow
Writing past the end of a fixed-size block of memory (CWE-787 and its variants CWE-121 and CWE-122). Appears in module 27.
Bug bounty program —
A vulnerability disclosure policy that also pays rewards for qualifying reports. Appears in modules 27, 39.
Building automation system — Building Automation System
OT that controls a building's heating, ventilation, lighting, fire, access control and other utilities. Appears in module 35.
Bundle — Model bundle
The versioned set of everything that decides an output: model, adapter, prompt template, decoding settings and threshold. Appears in module 37.
Business glossary —
The agreed, owned definitions and formulas for business terms such as headcount. Appears in module 45.
business judgment rule —
A presumption that directors acted on an informed basis, in good faith and in the company's interest, so courts do not second-guess decisions on outcome alone. Appears in module 47.
BYOD — Bring Your Own Device
A model in which employees use their own devices for work. Appears in module 35.

C

C-SCRM — Cybersecurity Supply Chain Risk Management
NIST's term for a systematic process for managing exposure to cybersecurity risks throughout the supply chain (SP 800-161r1). Appears in module 40.
C2 — Command and Control
The channel an attacker uses to send instructions to code on a compromised machine and receive results. Appears in modules 0, 1, 28, 29, 33.
CA — Assessment, Authorization, and Monitoring
SP 800-53 control family on assessing controls, including penetration testing (CA-8). Appears in modules 2, 12, 14.
CAB — Change Advisory Board
A group that reviews and advises on higher-risk changes before they are authorized. Appears in module 40.
CAIO — Chief AI Officer
An executive accountable for an organization's use and governance of artificial intelligence. Appears in modules 43, 44, 45, 47.
CAISP — Certified AI Security Professional
An AI security certification offered by Practical DevSecOps. Appears in modules 39, 99.
CAISSI — Center for Advancing Innovation and Standards for Super Intelligence
The NIST center that, as of November 2026, NIST describes as industry's main U.S. government contact for testing and research on advanced AI systems. Appears in module 38.
Caldera
MITRE's open-source platform for automating adversary emulation with agents, abilities and adversary profiles; for isolated labs only. Appears in module 29.
Calibration — Calibration (sufficiency)
Among people given the same score, outcomes are the same across groups. Appears in modules 13, 41.
Calibration —
Agreement between a model's stated confidence and how often it is actually right. Appears in modules 0, 16, 33.
CaMeL — CaMeL (Debenedetti et al., 2025)
A defense design that separates control flow from untrusted data and checks capabilities at tool calls. Appears in modules 7, 8, 15.
Campaign —
A planned red-team engagement with scope, methods, success criteria and reporting. Appears in modules 11, 29.
Campbell's law
The more a quantitative indicator is used for decisions, the more it is corrupted and the more it distorts what it monitors. Appears in modules 41, 44.
Canary sequence
A random made-up secret inserted into training data on purpose so that its memorization can be measured afterwards. Appears in module 37.
Canary test
A small fixed set of inputs re-run regularly to detect unexpected changes in a system. Appears in module 14.
Canary token
A unique secret string placed somewhere so that its appearance elsewhere signals a leak. Appears in modules 20, 32.
Capability —
On Linux, one separately grantable piece of root's power, such as binding to ports below 1024. Appears in module 0.
Capability map —
A one-page structured list of what an organization must be able to do, annotated with importance, performance and risk. Appears in modules 40, 44.
CapEx — Capital Expenditure
Spending on assets that give value over several years, which accounting rules may allow to be recorded as an asset and spread over time. Appears in modules 40, 42.
Capture–recapture —
Estimating a total count from the overlap between two independent detectors; it fails when the detectors share blind spots. Appears in modules 4, 16, 32.
Cardinality —
The number of distinct values in a field. Appears in module 32.
Caremark duty — Duty of oversight
The duty to make a good-faith effort to ensure reporting systems bring key risk and compliance information to the board, and to act on red flags. Appears in module 47.
Carve-out method
Reporting approach that excludes a subservice organization's controls from the SOC 2, so assurance about them must come from elsewhere. Appears in module 12.
Case study (ATLAS) —
A documented real incident or demonstration, mapped step by step to ATLAS tactics and techniques. Appears in module 10.
CBC — Cipher Block Chaining
An older block-cipher mode that XORs each plaintext block with the previous ciphertext block; it needs padding and has no built-in integrity check. Appears in module 34.
CBOM — Cryptography Bill of Materials
A machine-readable inventory of algorithms, keys, certificates and protocols and where they are used; CycloneDX defines one. Appears in module 34.
CBRN — Chemical, Biological, Radiological and Nuclear
A category of severe-harm information risks named in NIST AI 600-1. Appears in modules 12, 38.
CC — Common Criteria (Trust Services)
Prefix for the SOC 2 common criteria series, such as CC6 for logical and physical access controls. Appears in module 12.
CC — Creative Commons
A non-profit that publishes standard open licenses such as CC BY 4.0. Appears in modules 22, 24.
CC — Certified in Cybersecurity
ISC2's entry-level cybersecurity certification. Appears in module 39.
CC BY — Creative Commons Attribution license
A Creative Commons license that lets anyone share and adapt a work, even commercially, if they give credit. Appears in modules 24, 35, 39.
CC BY 4.0 — Creative Commons Attribution 4.0 International
An open license that allows reuse and adaptation, including commercially, with attribution. Appears in module 24.
CC0 — Creative Commons Zero
A Creative Commons tool that places a work as close to the public domain as the law allows, so anyone can reuse it without conditions. Appears in module 8.
CCISO — Certified Chief Information Security Officer
An EC-Council certification program aimed at security executives. Appears in module 43.
CCL — Critical Capability Level
In Google DeepMind's Frontier Safety Framework, a capability level at which a model without extra mitigations could pose severe risk. Appears in module 38.
CCPA — California Consumer Privacy Act
California's main privacy law, which gives residents, including job applicants, rights over their personal information. Appears in modules 19, 33, 36.
CCS — ACM Conference on Computer and Communications Security
A major academic security conference run by the ACM. Appears in modules 9, 27, 39.
CCSK — Certificate of Cloud Security Knowledge
The Cloud Security Alliance's cloud security knowledge credential. Appears in module 31.
CCSP — Certified Cloud Security Professional
ISC2's cloud security certification, which requires an exam and several years of relevant work experience. Appears in modules 31, 39.
CD — Continuous Delivery or Deployment
Automatically releasing tested changes; paired with CI as CI/CD. Appears in modules 11, 30.
CDAO — Chief Data and AI Officer
One executive who holds both the data and AI officer roles. Appears in module 45.
CDE — Critical Data Element
A field important enough to govern tightly because key decisions or reports depend on it. Appears in module 45.
CDMP — Certified Data Management Professional
DAMA's data management credential, based on the DMBOK. Appears in module 45.
CDN — Content delivery network
A network of servers spread around the world that serves or receives content close to users on a site's behalf. Appears in modules 0, 2.
CDO — Chief Data Officer
The executive accountable for an organization's data as an asset: governance, quality, definitions and permitted use. Some organizations use the same initials for a Chief Digital Officer. Appears in modules 44, 45.
CE — Continuing Education
CompTIA's program for keeping a certification active by earning credits or retaking an exam. Appears in modules 39, 99.
CEL — Common Expression Language
A small expression language used for Google Cloud IAM conditions and workload identity attribute conditions. Appears in module 31.
Center for Creative Leadership
A leadership research and education organization whose studies founded research on executive derailment. Appears in module 47.
Central DP — Central differential privacy
DP where a trusted curator holds raw data and adds noise to the outputs. Appears in module 36.
CEO — Chief Executive Officer
The most senior executive, accountable to the board for the whole company. Appears in modules 41, 42, 43, 45, 46, 47.
CERT — CERT Coordination Center (CERT/CC)
The vulnerability coordination center at Carnegie Mellon University's Software Engineering Institute. Appears in modules 22, 25, 27.
CERT/CC — CERT Coordination Center
A vulnerability coordination center at Carnegie Mellon University's Software Engineering Institute. Appears in modules 22, 39.
Certificate authority —
An organization that signs certificates binding names to public keys; browsers trust a built-in list of root certificate authorities. Appears in modules 0, 2.
Certified radius —
The distance within which a model's prediction is proven not to change. Appears in module 9.
CET — Control-flow Enforcement Technology
Intel processor features, a shadow stack and indirect branch tracking, that block control-flow hijacking. Appears in module 27.
CEU — Continuing Education Unit
CompTIA's name for the credits that renew its certifications. Appears in modules 39, 99.
CFA — Confirmatory Factor Analysis
A statistical model that tests whether observed measures load on hypothesized underlying factors; used to test measurement invariance. Appears in module 16.
CFAA — Computer Fraud and Abuse Act
The main U.S. federal law against accessing computers without authorization (18 U.S.C. § 1030). Appears in modules 1, 2, 3, 4, 5, 6, 8, 9, 10, 11, 14, 15, 16, 17, 22, 23, 25, 27, 28, 29, 30, 31, 33, 34, 35, 37, 38, 39, 40, 42.
CFI — Control-Flow Integrity
A defense that checks a program only jumps to legitimate destinations. Appears in modules 5, 27.
CFO — Chief Financial Officer
The executive responsible for an organization's finances. Appears in modules 12, 41, 42, 43, 44, 45, 46, 47.
CFP — Call for Papers
An invitation from a conference to submit talk or paper proposals by a deadline. Appears in module 24.
CFPB — Consumer Financial Protection Bureau
The U.S. agency that administers Regulation B and issues guidance on consumer credit. Appears in module 16.
CFR — Code of Federal Regulations
The official compilation of U.S. federal agency rules, such as 29 CFR Part 1607. Appears in modules 1, 2, 12, 13, 14, 16, 19, 21, 22, 23, 24, 25, 26, 30, 34, 35, 36, 42, 43, 44, 46, 47, 99.
CFReDS — Computer Forensic Reference Data Sets
NIST's documented sets of simulated digital evidence for tool validation, training and proficiency testing. Appears in module 30.
CGEIT — Certified in the Governance of Enterprise IT
ISACA's certification for IT governance professionals. Appears in module 40.
CGRC — Certified in Governance, Risk and Compliance
An ISC2 certification for authorizing and maintaining information systems under risk frameworks. Appears in module 39.
Chain of custody
A record of who held evidence and when, showing it was not altered. Appears in modules 6, 22, 28, 30.
chain of trust —
A sequence in which each stage verifies or measures the next, starting from a root of trust. Appears in module 35.
Change enablement —
The ITIL 4 practice that assesses, authorizes and schedules changes to maximize successful changes. Appears in module 40.
Chargeback —
Billing a business unit's budget for the technology it uses. Appears in modules 40, 42, 44.
Chat template
The fixed format, with special tokens, that marks system, user and assistant turns for an instruction-tuned model. Appears in modules 11, 37, 38.
Checkov —
An open-source static analysis tool for infrastructure as code with built-in checks for the major clouds. Appears in module 31.
Choke point —
A step in an attack chain where one control would stop everything after it. Appears in modules 10, 28.
CHRO — Chief Human Resources Officer
The executive accountable for an organization's people function. Appears in modules 1, 17, 44, 45, 47.
CI — Container Inherit
A Windows access control flag meaning a permission is inherited by subfolders. Appears in module 0.
CI — Continuous integration
Automatically building and testing code on each change. Appears in modules 4, 11, 14, 17, 18, 20, 22, 23, 24, 27, 28, 31, 32, 34, 37.
CI — Confidence Interval
A range of plausible values for an unknown quantity, given the data. Appears in modules 13, 16, 21.
CI/CD — Continuous Integration and Continuous Delivery
Automated pipelines that test and release software changes. Appears in modules 11, 17, 31.
CIA — Confidentiality, Integrity, Availability
The three core security goals: keep data secret, unaltered and available. Appears in modules 0, 1.
CIDR — Classless Inter-Domain Routing
A way of writing a block of IP addresses as an address plus the number of fixed leading bits, such as 10.20.0.0/22. Appears in modules 0, 32.
CIFAR — Canadian Institute For Advanced Research (dataset name)
CIFAR-10 and CIFAR-100 are small labeled image datasets named after the institute, widely used in robustness research. Appears in modules 9, 38.
CIM — Common Information Model
Splunk's normalization schema, which maps many data sources onto shared data models and field names. Appears in module 32.
CIMA — Chartered Institute of Management Accountants
A UK-based professional body for management accountants. Appears in module 12.
CIO — Chief Information Officer
The executive accountable for an organization's information technology. Appears in modules 1, 6, 40, 41, 42, 43, 44, 45, 46, 47.
Ciphertext —
Encrypted data, unreadable without the right key. Appears in modules 0, 34.
CIPM — Certified Information Privacy Manager
IAPP certification on building and running a privacy program. Appears in modules 36, 39.
CIPP — Certified Information Privacy Professional
An IAPP certification in privacy law, with regional variants such as CIPP/E and CIPP/US. Appears in modules 36, 39.
CIPP/E — Certified Information Privacy Professional/Europe
IAPP certification on European data protection law, centered on the GDPR. Appears in modules 36, 39.
CIPP/US — Certified Information Privacy Professional/United States
IAPP certification on US privacy law, including a workplace privacy domain. Appears in modules 36, 39.
CIPT — Certified Information Privacy Technologist
IAPP certification on privacy in technology, including privacy by design and privacy engineering. Appears in modules 36, 39.
Circuit —
A set of model components that together implement a specific behavior. Appears in module 38.
Circular reporting —
When many reports repeat one original source, making a claim look independently confirmed when it is not. Appears in module 33.
CIS — Center for Internet Security
A nonprofit that publishes configuration benchmarks, including for Kubernetes. Appears in module 17.
CISA — Cybersecurity and Infrastructure Security Agency
The U.S. federal agency that publishes the Known Exploited Vulnerabilities catalog and binding directives. Appears in modules 6, 17, 22, 25, 27, 35, 39.
CISM — Certified Information Security Manager
ISACA's certification for information security managers. Appears in modules 39, 40, 43.
CISO — Chief Information Security Officer
The executive accountable for an organization's information security program. Appears in modules 1, 6, 12, 14, 26, 29, 39, 40, 42, 43, 44, 45, 46, 47.
CISSP — Certified Information Systems Security Professional
ISC2's broad security certification for experienced practitioners. Appears in modules 31, 39, 43.
CJEU — Court of Justice of the European Union
The EU's highest court on questions of EU law, which interprets the GDPR. Appears in module 19.
CK — Common Knowledge
The last part of MITRE ATT&CK (Adversarial Tactics, Techniques, and Common Knowledge). Appears in modules 19, 20.
CKKS — Cheon-Kim-Kim-Song
A homomorphic encryption scheme for approximate arithmetic on encrypted real or complex numbers. Appears in module 34.
Clean-label poisoning —
Poisoning in which every poisoned example carries a label that looks correct. Appears in module 9.
CLI — Command-Line Interface
A text-based way to run a program by typing commands. Appears in module 11.
Clustered standard error
A standard error that allows for correlation among test items that come in related groups, such as several questions about one résumé. Appears in modules 11, 16, 37.
CM — Continuous Monitoring / Configuration Management
In CSF 2.0, DE.CM means continuous monitoring; in SP 800-53, CM is the configuration management control family. Appears in modules 12, 20.
CMF — Certification Maintenance Fee
The IAPP fee that keeps certifications current for non-members. Appears in modules 36, 39, 99.
CMU — Carnegie Mellon University
The university that hosts the Software Engineering Institute and CERT/CC. Appears in module 22.
CMVP — Cryptographic Module Validation Program
The joint NIST and Canadian Centre for Cyber Security program that validates cryptographic modules against FIPS 140-3. Appears in module 34.
CNA — CVE Numbering Authority
An organization authorized to assign CVE identifiers. Appears in modules 22, 27.
CNAME — Canonical Name record
A DNS record that says one name is an alias for another. Appears in modules 0, 2.
CNCF — Cloud Native Computing Foundation
The foundation that hosts open-source cloud-native projects such as Kubernetes and OPA. Appears in module 31.
CNSSI — Committee on National Security Systems Instruction
A numbered policy document from the U.S. committee that sets security rules for national security systems. Appears in module 0.
CO — Incident Response Reporting and Communication / Incident Recovery Communication
CSF 2.0 categories RS.CO and RC.CO: telling the right people about an incident and the recovery. Appears in modules 12, 20, 39.
COAE — HTB Certified Offensive AI Expert
Hack The Box's hands-on AI red-teaming certification tied to its AI Red Teamer job-role path. Appears in modules 39, 99.
COBIT — Control Objectives for Information and Related Technologies
ISACA's framework for governance and management of enterprise IT; COBIT 2019 has 40 governance and management objectives. Appears in module 40.
Codespaces — GitHub Codespaces
A cloud development machine from GitHub that opens in the browser; the course uses it so learners install nothing locally. Appears in modules 0, 2, 5, 7, 8, 10, 11, 15, 16, 17, 18, 21, 23, 24, 29, 30, 32, 33, 35, 36, 38, 39.
COFF — Common Object File Format
An object file format whose header is part of the PE format. Appears in module 28.
cognitive load — Cognitive load
The amount of knowledge and attention a team or person must hold to do the work; too much lowers quality and speed. Appears in module 44.
COGS — Cost of goods sold
The direct cost of delivering what a company sells; for software companies, often hosting and support. Also called cost of revenue. Appears in module 42.
Cohen's kappa —
A measure of agreement between two raters that corrects for agreement expected by chance. Appears in modules 6, 8, 10, 11, 13, 16, 38, 40, 46.
Coherent risk measure
A risk measure with four properties set out by Artzner et al. (1999), including subadditivity; Expected Shortfall is coherent, Value at Risk is not. Appears in module 46.
Colab — Google Colaboratory
Google's hosted notebook service that runs Python in the cloud from a browser. Appears in modules 0, 3, 6, 7, 8, 9, 10, 11, 12, 13, 14, 16, 18, 19, 20, 21, 23, 24, 25, 27, 28, 29, 30, 32, 33, 34, 35, 36, 37, 38, 39, 46.
Collider
A variable caused by two other variables; conditioning on it can create a false association between them. Appears in module 21.
Commitment to change
A mindset binding a person to support a change, out of desire (affective), obligation (normative) or cost of not doing so (continuance). Appears in modules 40, 42.
Common criteria — Common criteria (CC1–CC9)
The security criteria required in every SOC 2, numbered CC1 to CC9. Appears in module 12.
common method bias — Common method bias
Inflated correlations that arise when the same people answer all measures in the same way at the same time. Appears in module 44.
compa-ratio — Comparative ratio
Pay divided by the midpoint of its pay range; 1.00 means exactly at the midpoint. Appears in module 26.
COMPAS — Correctional Offender Management Profiling for Alternative Sanctions
A commercial criminal-justice risk tool whose public data set is widely used in fairness research. Appears in module 16.
COMPASS — Conference on Computing and Sustainable Societies
An ACM SIGCAS conference on computing for social and sustainable development. Appears in module 34.
compensatory scoring —
A scoring rule, such as an average or weighted sum, where a high score in one area can make up for a low score in another. Appears in module 26.
competency model —
A set of competencies, each a cluster of related KSAOs shown in observable behavior, usually with descriptions for each level. Appears in modules 10, 26, 42, 44, 47.
Complementary suppression
Hiding extra cells so that a hidden small group cannot be worked out by subtraction. Appears in module 13.
Complete mediation
Checking every access against policy, every time, in code. Appears in modules 7, 8.
Compliance budget
The limited effort a person will spend on security tasks, which must be managed like a financial budget. Appears in module 43.
Composition — Composition (differential privacy)
The property that privacy losses from several DP releases add up to a bounded total, so a budget can be tracked. Appears in module 36.
CompTIA — Computing Technology Industry Association
A trade association that offers vendor-neutral IT and security certifications. Appears in modules 39, 99.
concentration risk — Concentration risk
The risk from heavy dependence on one supplier or a few closely connected suppliers, so one failure affects much of the business. Appears in modules 44, 46.
Concept drift
A change in the relationship between inputs and outcomes, so a model's learned rule stops fitting. Appears in modules 18, 37.
conduit —
In ISA/IEC 62443, a controlled communication path between zones. Appears in module 35.
Confabulation
Confidently stated but false output from a generative model, often called hallucination. Appears in modules 8, 12.
Confidence calibration —
How well a person's confidence in their answers matches how often those answers are right. Not the same as score calibration in fairness work (Module 13). Appears in module 0.
Confidence interval —
A range produced by a method that contains the true value a stated share of the time over repeated samples. Appears in modules 8, 9, 11, 12, 13, 14, 16, 21, 22, 23, 24, 25, 34, 37, 38, 39, 45.
Configuration drift —
A difference between the infrastructure declared in code and what is actually deployed. Appears in module 31.
Configuration flag
A setting that turns a planted weakness on or off so fixes can be tested. Appears in modules 8, 14.
conflict check —
A routine check of a new client and engagement against past and current clients, work and interests before accepting the work. Appears in modules 22, 24, 26.
Conftest —
An open-source tool that tests structured configuration files, including Terraform, against Rego policies. Appears in module 31.
Confused deputy
A privileged component tricked into misusing its rights on behalf of a less privileged party. Appears in modules 5, 7, 8, 15, 23.
conjunctive scoring —
A multiple-hurdle rule: you meet a level only when every requirement for it is met, so a strength cannot make up for a gap. Appears in modules 26, 47.
Consequential decision
In Colorado's AI legislation, a decision with a material effect on matters such as employment. Appears in modules 12, 23, 29, 44.
Constant-time code —
Code whose running time and memory access do not depend on secret values. Appears in modules 34, 35.
constant-time comparison —
Checking a secret by always examining every byte, so the time taken does not reveal how much was correct. Appears in modules 25, 34, 35.
Construct
The unobservable property a measurement is meant to reflect, such as susceptibility to prompt injection. Appears in modules 13, 16.
Construct validity —
How well a measure reflects the construct (the underlying property) it is meant to measure, judged from several kinds of evidence. Appears in modules 1, 3, 13, 16, 24, 25, 32, 37, 38, 44.
Consumer report
Under the FCRA, a report from a consumer reporting agency bearing on a person's characteristics, such as many employment background checks. Appears in modules 6, 16, 36.
Context window
The maximum number of tokens a model can consider at once. Appears in modules 7, 8.
Continuous batching
Adding new requests to a running batch, and removing finished ones, at every generation step to keep the accelerator busy. Appears in module 37.
Control — Security or privacy control
A specific safeguard that reduces a risk, such as enforcing least privilege. Appears in modules 12, 24.
Control plane —
The part of a cloud service used to create and configure resources. Appears in modules 30, 31.
Control task —
A task with random labels per word type, used to measure how much a probe succeeds by memorization; selectivity is real-task accuracy minus control accuracy. Appears in module 38.
Converter (PyRIT) —
A PyRIT component that transforms a prompt or response, such as Base64 encoding. Appears in module 11.
Conway's law
The observation that organizations design systems whose structure copies their own communication structure. Appears in modules 41, 42, 44.
COP — Call-Oriented Programming
A code-reuse technique like ROP that chains existing code fragments ending in indirect calls. Appears in module 27.
COPE — Corporate-Owned, Personally Enabled
A device model in which the organization owns the phone but allows some personal use. Appears in module 35.
Copula
A function that joins separate distributions for each risk into one joint distribution with a chosen dependence structure. Appears in module 46.
Corpus — Fuzzing corpus
The set of inputs a fuzzer keeps because each reached something new. Appears in module 27.
Correlation rule —
A Sigma meta-rule that links events matched by other rules, counting events or distinct values, or requiring rules to match together in a time window. Appears in module 32.
CORS — Cross-Origin Resource Sharing
Browser rules, set by the server, about which other sites may read its responses. Appears in module 4.
Cosine similarity
A measure of how closely two vectors point in the same direction, from -1 to 1. Appears in module 7.
COSO — Committee of Sponsoring Organizations of the Treadway Commission
The body that publishes the 2013 Internal Control—Integrated Framework, whose 17 principles underlie SOC 2 criteria CC1–CC5. Appears in modules 12, 43, 46.
Cost of delay
The value lost for each unit of time a piece of work is not done. Appears in modules 0, 10, 41.
Cost of risk
The expected loss from a risk you keep, plus the cost of transferring it and any capital held against it. Appears in module 42.
cost taxonomy — Cost taxonomy
An agreed set of categories into which every technology cost is sorted so costs can be compared across time, units and organizations. Appears in module 44.
Counter-metric —
A second measure paired with a target metric to reveal gaming or side effects. Appears in modules 40, 46.
Counterfactual —
What would have happened without the intervention; value is the difference from it. Appears in module 45.
Counterfactual fairness
A causal fairness definition requiring the same decision in a world where the person's protected attribute, and its consequences, were different. Appears in modules 13, 24, 37.
Counterfactual flip rate
The share of minimal-pair test items whose output changes when only a signal of a protected attribute, such as a name, is changed. Appears in module 37.
Courses of action matrix —
A table listing, for each kill chain phase, the defensive actions available, such as detect, deny or disrupt. Appears in module 33.
Covariate shift
A change in the inputs a model sees, with no change in the relationship between inputs and outcomes. Appears in module 37.
Coverage —
The share of exploited vulnerabilities that a prioritization rule flags; recall in statistical terms. Appears in module 6.
CPA — Certified Public Accountant
A licensed accountant; only a CPA firm can issue a SOC 2 report. Appears in module 12.
CPE — Common Platform Enumeration
A standard name for a product and version. Appears in module 2.
CPE — Continuing Professional Education
Ongoing learning credits that certification holders must earn to stay certified. Appears in modules 24, 33, 36, 39, 99.
CPEs — Continuing Professional Education credits
Hours of learning logged to keep a certification active. Appears in modules 28, 33, 36, 39, 99.
CPI — Cost Performance Index
Earned value divided by actual cost; below 1 means over cost. Appears in module 40.
CPO — Chief Product Officer
The executive accountable for a company's products; in this module's worked example, the owner of product-related risks. Appears in module 43.
CPPA — California Privacy Protection Agency
The California agency that makes rules under and enforces the CCPA. Appears in modules 19, 36.
CPRA — California Privacy Rights Act
A 2020 ballot measure that amended the CCPA and created the CPPA. Appears in modules 19, 36.
CPU — Central Processing Unit
The main processor of a computer; one of the resources an API must limit per request. Appears in modules 4, 7, 8, 11, 15, 27, 28, 37, 38.
CR — Confidentiality Requirement
CVSS v4.0 Environmental metric for how much confidentiality matters to the organization. Appears in modules 6, 39.
CRA — Consumer Reporting Agency
Under the FCRA, a firm that prepares consumer reports, such as many background-check providers. Appears in module 36.
Crackme
A small program written for reverse-engineering practice, usually asking for a password or key. Appears in modules 28, 35.
Crash triage
Reproducing, deduplicating, minimizing and ranking the crashes a fuzzer finds. Appears in module 27.
Credible interval
A Bayesian interval that contains the unknown value with a stated probability, given the model and prior. Appears in module 21.
CredSSP — Credential Security Support Provider
A Windows protocol that can pass user credentials to a remote server. Appears in module 3.
Crescendo —
A multi-turn jailbreak that starts with harmless questions and escalates gradually. Appears in module 11.
CRISC — Certified in Risk and Information Systems Control
ISACA's certification for IT risk professionals. Appears in modules 40, 46.
Criterion
The measure of job performance or success used to evaluate a predictor. Appears in modules 13, 41, 47.
Criterion contamination
Irrelevant influences, such as rater bias, in the performance measure. Appears in modules 8, 13, 14, 22, 37, 38, 44.
Criterion-related validity
How well scores on a selection method predict a later outcome such as job performance, usually reported as a correlation. Appears in modules 12, 13, 26, 41, 47.
CRL — Certificate Revocation List
A signed list, published by a certificate authority, of certificates it has canceled before their expiry date. Appears in module 0.
CRMP — Certified Risk Management Professional
The enterprise risk management credential awarded by RIMS. Appears in module 46.
CRO — Chief Risk Officer
The executive who leads enterprise risk management. Appears in modules 43, 46.
Cross-Region inference —
Routing a model request to another cloud Region, within a geography or globally, to gain capacity. Appears in module 31.
Cross-site scripting —
Injection of script into a web page so it runs in other users' browsers. Appears in modules 4, 8, 23.
Crosswalk
A table linking requirements or outcomes across different frameworks, laws or standards. Appears in modules 9, 12, 19, 24, 26, 36, 45, 99.
CRQ — Cyber Risk Quantification
Expressing cyber risk as probability distributions of loss, usually in money. Appears in module 12.
CRQC — Cryptographically relevant quantum computer
A quantum computer large and reliable enough to break deployed public-key cryptography; none is publicly known to exist. Appears in module 34.
CRUD — Create, Read, Update, Delete
The four basic operations on stored data; OCSF's API Activity class covers them. Appears in module 32.
Crypto agility —
The capability to replace and adapt cryptographic algorithms without breaking security or operations. Appears in module 34.
Crypto-shredding —
Making encrypted data permanently unreadable by destroying its key. Appears in modules 0, 31, 34, 36.
Cryptoperiod —
The time span during which a specific key is authorized for use (NIST SP 800-57). Appears in modules 31, 34.
CRYSTALS — Cryptographic Suite for Algebraic Lattices
The research suite behind Kyber and Dilithium, which NIST standardized as ML-KEM and ML-DSA. Appears in module 34.
CS — Case study (ID prefix)
The prefix of ATLAS case-study identifiers, as in AML.CS0035. Appears in module 10.
CS — Computer Science
The academic field of computing. Appears in module 22.
CSA — Cloud Security Alliance
A non-profit organization that publishes cloud security guidance and the CCSK credential. Appears in module 31.
CSF — Cybersecurity Framework
NIST's high-level framework for managing cybersecurity risk. Appears in modules 5, 8, 9, 12, 20, 26, 30, 33, 36, 39, 40, 42, 43, 44, 47, 99.
CSOC — Complementary Subservice Organization Control
A control a vendor expects one of its own providers (a subservice organization) to perform. Appears in module 12.
CSP — Content Security Policy
A browser policy that limits which scripts and resources a page may load. Appears in modules 4, 8.
CSPM — Cloud Security Posture Management
Continuous checking of deployed cloud resources against security recommendations and standards. Appears in module 31.
CSPRNG — Cryptographically secure pseudorandom number generator
A random number generator safe for creating secrets such as session IDs. Appears in modules 0, 4, 34.
CSRC — Computer Security Resource Center
NIST's website for its security publications, at csrc.nist.gov. Appears in modules 12, 20, 24, 34, 99.
CSRF — Cross-site request forgery
Making a victim's browser send an unintended request carrying the victim's credentials. Appears in modules 4, 8, 99.
CSSLP — Certified Secure Software Lifecycle Professional
An ISC2 certification for building security into each stage of software development. Appears in module 39.
CSV — Comma-Separated Values
A plain-text table format: one row per line, with commas between the fields. Appears in modules 0, 7, 8, 11, 13, 14, 16, 18, 29, 30, 34.
CSVs — Comma-Separated Values files
Plural of CSV. Appears in module 14.
CSWP — Cybersecurity White Paper
A NIST publication series; CSF 2.0 is NIST CSWP 29. Appears in modules 12, 26, 34, 36, 42, 43, 44, 47.
CT — Certificate Transparency
Public, append-only logs of issued TLS certificates. Appears in modules 0, 2, 36.
CTF — Capture The Flag
A security exercise or competition in which players find hidden 'flags' by solving challenges. Appears in modules 23, 27, 30, 39.
CTI — Cyber Threat Intelligence
Information about threats, such as indicators, behaviors and actors, used to defend systems. Appears in modules 28, 29, 33.
CTID — Center for Threat-Informed Defense
MITRE's research center that publishes open threat-informed defense projects, including the Adversary Emulation Library. Appears in module 29.
CTO — Chief Technology Officer
The executive accountable for an organization's technology and engineering strategy, architecture and the engineering organization that builds its products. Appears in modules 43, 44, 45, 47.
CTPH — Context Triggered Piecewise Hashing
A fuzzy hashing method that splits a file at content-chosen points and hashes the pieces; used by ssdeep. Appears in module 28.
CTT — Classical Test Theory
A measurement model in which an observed score equals a true score plus random error; it underlies most reliability coefficients. Appears in module 16.
CUEC — Complementary User Entity Control
A control a vendor assumes its customers will run, such as removing their own users' access. Appears in module 12.
CV — Curriculum Vitae
An academic or professional record of publications, roles and qualifications. Appears in modules 4, 12, 22, 24, 31, 36.
CVD — Coordinated Vulnerability Disclosure
A process in which the finder of a flaw, the vendor and sometimes a coordinator work together so a fix is ready before details are published. Appears in modules 22, 23, 27, 39.
CVE — Common Vulnerabilities and Exposures
A public identifier for a specific known vulnerability. Appears in modules 2, 3, 4, 6, 8, 10, 17, 20, 22, 24, 27, 33, 34, 35, 39.
CVSS — Common Vulnerability Scoring System
FIRST's standard for describing a vulnerability's characteristics and turning them into a 0-10 severity score. Appears in modules 1, 2, 3, 4, 5, 6, 7, 8, 14, 16, 18, 22, 23, 26, 27, 29, 30, 33, 39, 42.
CVSS-B — CVSS Base score
A CVSS v4.0 score computed from Base metrics only. Appears in modules 2, 4, 6, 8.
CVSS-BE — CVSS Base and Environmental score
A CVSS v4.0 score using Base and Environmental metrics. Appears in modules 2, 6.
CVSS-BT — CVSS Base and Threat score
A CVSS v4.0 score using Base and Threat metrics. Appears in modules 2, 6, 8, 23.
CVSS-BTE — CVSS Base, Threat and Environmental score
A CVSS v4.0 score using all three scoring groups. Appears in modules 2, 6.
CWB — Counterproductive Work Behavior
Voluntary behavior by employees that harms the organization or its members, from minor production deviance to theft and aggression. Appears in module 25.
CWE — Common Weakness Enumeration
MITRE's catalog of software weakness types, such as CWE-89 for SQL injection. Appears in modules 4, 5, 18, 27, 34, 39.
Cyber Kill Chain —
Lockheed Martin's seven-phase model of an intrusion, from reconnaissance to actions on objectives. Appears in module 33.
Cyber range —
A hosted environment that simulates networks, hosts and users for security training and exercises. Appears in module 39.
Cyber Trust Mark —
A U.S. cybersecurity labeling program for consumer smart devices, announced by the White House in July 2023. Appears in module 35.
CyberLive
GIAC's hands-on exam format using virtual machines. Appears in modules 28, 33, 99.
CYOD — Choose Your Own Device
A model in which employees buy a device for personal use from an approved list and use it for work. Appears in module 35.
CySA+ — CompTIA Cybersecurity Analyst
A CompTIA certification for security operations and threat detection work. Appears in module 39.

D

D&O — Directors and Officers
Short for directors and officers insurance, which covers claims against them for wrongful acts in their roles. Appears in modules 43, 47.
D3FEND —
MITRE's knowledge graph of defensive cybersecurity techniques; sigma-cli's validators download its data when checking tagged rules. Appears in module 32.
DACL — Discretionary access control list
The Windows list on an object that decides which identities are allowed or denied which kinds of access. Appears in module 0.
DAG — Directed Acyclic Graph
A diagram of nodes joined by one-way arrows that never loop back on themselves. Appears in module 1.
DAMA — Data Management Association International
A professional association for data management that publishes the DMBOK and runs the CDMP credential. Appears in module 45.
DAN — Do Anything Now
A family of role-play jailbreak prompts that ask a chatbot to act as an unrestricted persona. Appears in module 11.
Dangerous-capability evaluation —
A test of whether a model can do something that could cause severe harm. Appears in module 38.
DAST — Dynamic Application Security Testing
Testing a running application from the outside for vulnerabilities. Appears in modules 8, 18.
Data action
NIST's term for any operation a system performs on data, such as collection, retention, analysis, disclosure or deletion. Appears in module 36.
Data contract —
A machine-checkable agreement between a data producer and its consumers on schema, meaning, quality and change notice. Appears in module 45.
Data event —
In AWS CloudTrail, a record of a data-plane operation on or in a resource, such as reading an S3 object; not logged by default. Appears in modules 30, 31.
Data leakage
Information from the test set reaching training, which makes test scores look better than real performance. Appears in module 37.
Data lineage
A record of where each piece of data came from and every transformation applied to it. Appears in modules 18, 37.
Data map
A structured record of which personal data a system holds, about whom, for what purpose, where it flows, who can access it and how long it is kept. Appears in module 36.
Data mesh —
An organizational model for data with domain ownership, data as a product, a self-serve platform and federated governance. Appears in module 45.
Data minimization
The principle that personal data must be adequate, relevant and limited to what is necessary for the purpose. Appears in modules 0, 8, 11, 18, 19, 20, 25, 30, 31, 32, 35, 36, 41, 42.
Data perimeter —
Always-on controls that allow access only when the identity is trusted, the resource is trusted and the network is expected. Appears in module 31.
Data plane —
The part of a cloud service used to read and write the data inside resources. Appears in modules 30, 31.
Data poisoning —
Inserting or changing training data so the trained model behaves as the attacker wants. Appears in modules 6, 8, 9, 10, 11, 12, 15, 18, 38.
Data product —
A dataset or interface built and run for named consumers, with an owner, documentation and quality promises. Appears in module 45.
Data protection by design and by default
The GDPR Article 25 duty to build data protection principles into systems and to process only the data needed for each purpose by default. Appears in module 36.
Data residency —
A requirement that data is stored or processed only in specified locations. Appears in modules 31, 37, 42.
Data steward —
The person who looks after a data domain day to day: definitions, quality issues and questions. Appears in module 45.
Data-only attack
Changing a program's behavior by corrupting ordinary data, not code pointers, which bypasses most control-flow defenses. Appears in module 27.
DataFrame
pandas' table object: rows of records with named columns. Appears in module 7.
Dataset inference —
Testing whether a whole dataset was used to train a model by combining weak signals across many documents. Appears in module 38.
Datasheet for a dataset
A document describing a dataset's motivation, composition, collection, processing, uses, distribution and maintenance. Appears in module 37.
Daubert — Daubert v. Merrell Dow Pharmaceuticals (1993)
The Supreme Court case that made trial judges gatekeepers of the reliability of expert testimony. Appears in module 22.
DB — Database
An organized store of data, such as a table of candidate records. Appears in modules 1, 8.
DC — Domain controller
A server that holds the AD database and answers authentication requests. Appears in module 3.
DC — Description Criteria
The AICPA's criteria for what a SOC 2 system description must contain (DC section 200). Appears in module 12.
DC — District of Columbia
Washington, DC, the US capital. Appears in modules 13, 16.
DCS — Distributed Control System
A control system that runs a process within one site, using many coordinated controllers under a supervisory layer. Appears in module 35.
DCSync — DCSync (ATT&CK T1003.006)
Misusing directory replication rights to obtain password data from a domain controller. Appears in module 3.
DCWP — Department of Consumer and Worker Protection
The New York City agency that enforces LL144. Appears in modules 12, 13, 14, 21, 22, 24.
DE — Detect
The CSF 2.0 Detect function: finding and analyzing possible attacks. Appears in modules 12, 20.
Decision log
A per-decision record of the request, the output and the exact bundle version that produced it. Appears in module 37.
Decision record —
The set of inputs needed to reconstruct an AI action: model version, instructions, context, user input, sampling settings, output and actor. Appears in modules 26, 30, 37, 43, 44, 47.
Decision rights
A clear statement of who decides, who must be asked first and who must be told after. Appears in modules 26, 41, 42, 43, 44, 45, 47.
Decode
The phase of generation in which output tokens are produced one at a time. Appears in module 37.
Decompilation
Reconstructing approximate higher-level code, such as C, from machine code; the result is an interpretation, not the original source. Appears in module 28.
Deconfliction
Quickly establishing whether suspicious activity is the authorized test or a real attacker, by matching it to the test's activity log. Appears in module 29.
Deduplication
Removing repeated or near-identical records before training, which reduces memorization. Appears in module 37.
Deep uncertainty
A situation in which decision makers cannot agree on the models, probabilities or valuations that describe the outcomes. Appears in module 46.
Deepfake —
Synthetic or altered media, produced with machine learning, showing a person saying or doing something they did not say or do. Appears in modules 25, 33.
Defense in depth
Using several independent controls so one failure does not mean compromise. Appears in modules 1, 4, 8, 20, 27, 31, 39.
DEFF — Design Effect
The factor by which clustering inflates the variance of an estimate compared with independent observations. Appears in module 16.
DEK — Data encryption key
In envelope encryption, the key that encrypts the actual data; it is stored only in wrapped (encrypted) form. Appears in modules 0, 31, 34.
deliberate practice —
Effortful activity designed specifically to improve a skill, usually with feedback, as defined by Ericsson and colleagues in 1993. Appears in modules 26, 39.
demographic differential —
A difference in error rates between demographic groups for the same system and threshold. Appears in modules 25, 35.
Demographic parity — Demographic parity (independence)
Equal selection rates across groups. Appears in module 13.
Deny-list
A rule that blocks known-bad values and permits everything else; fragile against new variants. Appears in module 4.
DEP — Data Execution Prevention
Windows' name for marking data memory as non-executable. Appears in modules 5, 27.
Deployer — Deployer (EU AI Act)
The organization that uses an AI system under its authority, such as an employer using a screener. Appears in modules 12, 44, 45.
Deployment type —
A provider setting for a model deployment that decides where prompts and responses may be processed. Appears in module 31.
Deprecated —
NIST status: an algorithm may still be used, but with some risk that the data owner must assess. Appears in module 34.
derailment —
When a manager expected to rise further is fired, demoted or plateaus below the expected level. Appears in module 47.
Derivative
The slope of a function at a point: how fast the output changes when the input is nudged. Appears in module 9.
derivative suit —
A lawsuit a shareholder brings on the corporation's behalf against its own directors or officers. Appears in module 47.
DES — Data Encryption Standard
An obsolete cipher disabled by default in modern Windows Kerberos. Appears in module 3.
Deserialization —
Rebuilding program objects from stored or received data. Deserializing untrusted data without checking it is a known weakness (CWE-502). Appears in modules 5, 18, 27, 28.
Design effect —
How much clustering of attempts inflates the variance of an estimate. Appears in modules 11, 16, 45.
Desirable difficulty —
A practice condition that feels harder now but produces better learning later, such as spacing or retrieval. Appears in module 39.
DET — Detection strategy (ATT&CK ID prefix)
An ATT&CK object describing how to detect a technique, introduced in v18. Appears in module 10.
Detecting — Detecting (LINDDUN)
A LINDDUN threat: someone can tell that an item of interest about a person exists, without seeing it. Appears in module 36.
Detection coverage
The share of a defined, justified set of tested behaviors for which the defense reached a defined outcome level. Appears in modules 3, 5, 10, 20, 28, 29, 31, 39.
Detection engineering —
Designing, building, testing and maintaining the logic that spots attacks in logs. Appears in modules 0, 3, 5, 10, 17, 20, 28, 29, 32, 33, 36, 39.
Detection strategy
In ATT&CK v18 and later, the documented approach for detecting a technique, linked to specific analytics. Appears in modules 2, 10, 29, 32.
Detection-as-code
Managing detection rules like software: version control, review, tests and automated deployment. Appears in modules 10, 20, 32, 39.
Detective control —
A control that finds non-compliance after it exists and raises an alert. Appears in modules 8, 31.
Detector (garak) —
A garak plugin that decides whether an output shows the failure being tested. Appears in module 11.
Detonation
Informal term for deliberately running a suspected malicious file to observe its behavior. Appears in module 28.
DeTT&CT — Detect Tactics, Techniques & Combat Threats
An open-source framework for scoring data source quality, visibility and detection coverage against ATT&CK. Appears in module 29.
DevEx — Developer Experience
How developers experience their daily work; one model frames it as feedback loops, cognitive load and flow state. Appears in module 44.
DFD — Data-flow diagram
A picture of where data goes, who handles it and where trust changes. Appears in modules 1, 7, 15, 36.
DFIR — Digital Forensics and Incident Response
The defenders who investigate incidents, preserve evidence and lead response. Appears in modules 29, 30, 39.
DGCL — Delaware General Corporation Law
Delaware's corporate statute, which governs most large U.S. public companies' boards, duties, exculpation and indemnification. Appears in module 47.
DH — Diffie-Hellman
A key-exchange method that lets two parties derive a shared secret over an open network from public values. Appears in modules 0, 34.
DHCP — Dynamic Host Configuration Protocol
The protocol that hands out IP addresses to devices on a network automatically. Appears in module 2.
Diamond Model —
A model that describes each intrusion event by four linked features: adversary, capability, infrastructure and victim. Appears in module 33.
DIF — Differential Item Functioning
When an item behaves differently across groups that are equal on the construct being measured. Appears in module 16.
Difference-in-differences —
An estimate of an effect from the change in a treated group minus the change in a comparison group; assumes parallel trends. Appears in modules 26, 44, 45.
Differential privacy —
A mathematical guarantee that an algorithm's output changes very little whether or not any one person's data is included. Appears in modules 3, 8, 9, 19, 21, 23, 30, 34, 36, 37, 38.
Digital Omnibus on AI — Regulation (EU) 2026/1744
The 2026 EU regulation that amended the AI Act, moving Annex III high-risk obligations to 2 December 2027. Appears in modules 12, 44, 45, 99.
Digital signature —
A value made with a private key that anyone holding the matching public key can check, proving origin and integrity. Appears in modules 0, 18, 34, 35.
DIRM — Digital Identity Risk Management
NIST SP 800-63-4's process for assessing risk and selecting IAL, AAL and FAL for a service. Appears in module 25.
Disallowed —
NIST status: an algorithm is no longer allowed for the stated purpose. Appears in module 34.
Disassembly
Translating machine code back into human-readable assembly instructions. Appears in modules 27, 28.
Disassociability
A NIST privacy engineering objective: data can be processed without being tied to individuals beyond what the system needs. Appears in module 36.
Discount rate —
The rate used to convert future money into today's value, often set from an organization's cost of capital. Appears in modules 40, 42.
Disparate impact
A disproportionate adverse effect of a neutral practice on a protected group. Appears in modules 12, 13, 19, 21, 22, 23, 25, 32, 34, 38.
Disparate treatment
Treating people differently because of a protected characteristic. Appears in modules 13, 14, 19, 34.
Distinct-count estimation —
Approximating the number of distinct values to save time and memory, at the cost of small errors that matter near thresholds. Appears in module 32.
Divestiture
Selling or spinning off part of a company, which requires separating its systems, data and people. Appears in module 42.
DLL — Dynamic-Link Library
A shared code file loaded by Windows programs at run time. Appears in module 5.
DMBOK — Data Management Body of Knowledge
DAMA's reference book that describes data management as knowledge areas around data governance. Appears in module 45.
DMCA — Digital Millennium Copyright Act
A US copyright law whose anti-circumvention rules can affect security research. Appears in module 22.
DMZ — Demilitarised Zone
A buffer network between two zones where shared services sit, so the zones never talk to each other directly. Appears in module 35.
DNP3 — Distributed Network Protocol 3
An industrial protocol used mainly in electric and water utilities for communication between control centers and remote sites; published as IEEE 1815. Appears in module 35.
DNS — Domain Name System
The internet system that maps names to IP addresses. Appears in modules 0, 2, 3, 8, 15, 28, 33.
DNSSEC — Domain Name System Security Extensions
Additions to DNS that let resolvers check digital signatures on records, so forged answers can be detected. Appears in module 0.
DOCX — Word document format
Microsoft Word's file format; a ZIP archive of XML files. Appears in modules 5, 8, 27, 28.
DoH — DNS queries over HTTPS
Sending DNS lookups inside HTTPS so that the local network cannot read or alter them. Appears in module 0.
DOI — Digital Object Identifier
A permanent ID for a published work; adding it to https://doi.org/ links to the work. Appears in modules 15, 16, 21, 24.
DOJ — Department of Justice
The U.S. federal department responsible for enforcing federal law, including prosecutions of hiring-fraud schemes. Appears in modules 25, 33, 39.
DORA — DevOps Research and Assessment
A research program known for four software delivery metrics: deployment frequency, lead time for changes, change failure rate and time to restore service. Not the EU's Digital Operational Resilience Act. Appears in modules 40, 44.
DoS — Denial of Service
Making a service unavailable to legitimate users. Appears in modules 2, 8.
Dot product
Multiply two vectors position by position and add the results; how a linear model combines features and weights. Appears in modules 7, 9.
Double extortion
A ransomware tactic that both encrypts data and threatens to publish stolen data. Appears in module 43.
DoW — Denial of Wallet
Driving up a pay-per-use service's costs through excessive usage. Appears in module 8.
DP — Differential Privacy
A guarantee that a model's behavior changes very little whether or not one person's data is included. Appears in modules 8, 9, 19, 36, 37, 38.
DP-SGD — Differentially Private Stochastic Gradient Descent
Training with per-example gradient clipping and added noise to achieve differential privacy. Appears in modules 8, 9, 37.
DPA — Data Processing Agreement
The contract between a controller and a processor that GDPR Article 28(3) requires. Not to be confused with a data protection authority, which this course calls a supervisory authority. Appears in modules 36, 40.
DPF — EU-US Data Privacy Framework
The transfer mechanism recognized by the Commission's adequacy decision 2023/1795 of 10 July 2023 for US organizations that certify under it. Appears in module 36.
DPIA — Data Protection Impact Assessment
A documented assessment of the privacy risks of processing personal data and how they will be reduced, required by the GDPR for high-risk processing. Appears in modules 1, 3, 9, 12, 19, 20, 32, 35, 36.
DPIAs — Data Protection Impact Assessments
Plural of DPIA. Appears in modules 19, 36.
DPO — Direct Preference Optimization
A method that trains on preference data with a simple classification-style loss instead of reinforcement learning. Appears in modules 7, 37.
DPoP — Demonstrating Proof of Possession
An OAuth mechanism that binds tokens to a key the client holds, so stolen tokens are harder to reuse (RFC 9449). Appears in module 17.
DPRC — Data Protection Review Court
A US body created under US regulations to review complaints about US intelligence access to data; the DPF adequacy decision relies on it. Appears in module 36.
DPRK — Democratic People's Republic of Korea
The official name of North Korea. Appears in module 33.
DRBG — Deterministic random bit generator
NIST's term for a cryptographically secure pseudorandom generator that expands a secret seed into random-looking bits (SP 800-90A). Appears in module 0.
Drift
A change over time in input data or in the relationship between inputs and correct outputs. Appears in modules 18, 37.
DRP — Disaster Recovery Plan
A plan for major, usually physical disruptions that deny access to the primary facility for an extended period. Appears in module 12.
DS — Data Security
CSF 2.0 category PR.DS: protecting data's confidentiality, integrity and availability. Appears in module 12.
DSAR — Data Subject Access Request
A request by a person to see the personal data an organization holds about them, under GDPR Article 15; often used loosely for all privacy rights requests. Appears in module 36.
DSRM — Directory Services Restore Mode
A recovery mode for domain controllers with its own administrator password. Appears in module 3.
DSS — Digital Signature Standard
NIST's signature standard, FIPS 186. Appears in modules 0, 34.
DSSE — Dead Simple Signing Envelope
A simple format for wrapping a signed payload together with its signature. Appears in module 18.
DTLS — Datagram Transport Layer Security
A version of TLS for datagram (UDP) traffic. Appears in module 34.
DTO — Data Transfer Object
A small type that lists only the fields a given caller may send or receive, used to prevent mass assignment. Appears in module 4.
dual ladder —
A job architecture with parallel individual contributor and management tracks at comparable levels. Appears in module 26.
Dual LLM pattern
A privileged model that plans and calls tools never sees untrusted text; a separate quarantined model handles that text. Appears in module 15.
Due diligence
An investigation done before a commitment, such as buying a company or signing a vendor, to confirm what you are getting and find risks. Appears in modules 8, 12, 25, 27, 34, 39, 40, 42, 46, 47.
duty of care —
A director's duty to make decisions on an informed basis after reasonable deliberation. Appears in modules 1, 8, 24, 27, 47.
duty of loyalty —
A director's duty to put the corporation's interests ahead of personal interests and to act in good faith. Appears in module 47.
DVC — Data Version Control
An open-source tool that tracks versions of datasets and models alongside code. Appears in module 8.
DVWA — Damn Vulnerable Web Application
A deliberately insecure web app used to teach web security safely. Appears in modules 14, 39.
Dynamic analysis
Running a file in a controlled environment and observing what it does. Appears in modules 28, 35.
Dynamic secret —
A credential created on demand for one task that expires automatically. Appears in modules 17, 34.

E

E — Exploit Maturity
CVSS v4.0 Threat metric: Attacked, Proof-of-Concept, Unreported or Not Defined. Appears in module 6.
E-Verify
A U.S. government system that electronically compares information from an employee's Form I-9 with government records. It does not prove the presenter owns the identity. Appears in module 25.
E&O — Errors and Omissions insurance
Professional liability insurance for claims that services were performed negligently. Appears in modules 6, 26.
E01 — Expert Witness Format
A forensic disk image format that stores the copied bytes together with case metadata and embedded hashes. Appears in module 30.
EA — Enterprise Architecture
The discipline of describing how business processes, information, applications and technology fit together now and in a target future, and planning the move between them. Appears in modules 40, 44.
EAC — Estimate at Completion
A forecast of a project's final cost, often BAC divided by CPI. Appears in module 40.
EBITDA — Earnings before interest, taxes, depreciation and amortization
A common non-GAAP measure of operating profit before financing, tax and non-cash charges; companies define it themselves. Appears in module 42.
EC-Council — International Council of E-Commerce Consultants
A certification body that runs security certifications such as CEH and CCISO. Appears in module 43.
EC2 — Elastic Compute Cloud
Amazon Web Services' service for renting virtual machines. Appears in module 31.
ECB — European Central Bank
The central bank of the euro area; it publishes the TIBER-EU red-teaming framework. Appears in module 29.
ECDH — Elliptic-curve Diffie-Hellman
Diffie-Hellman key exchange performed with points on an elliptic curve, such as X25519. Appears in modules 0, 34.
ECDHE — Elliptic-curve Diffie-Hellman, ephemeral
ECDH with fresh, throw-away key pairs for each session, which gives forward secrecy. Appears in modules 0, 34.
ECDSA — Elliptic Curve Digital Signature Algorithm
A digital signature algorithm based on elliptic curves, widely used in certificates. Appears in modules 0, 34.
ECE — Expected Calibration Error
A weighted average of the gaps between stated confidence and actual accuracy across confidence bins. Appears in module 16.
eCFR — Electronic Code of Federal Regulations
The official, continuously updated online edition of the U.S. Code of Federal Regulations. Appears in modules 13, 16, 19, 34, 43, 44, 46, 47.
ECMA — Ecma International
A standards body (originally the European Computer Manufacturers Association) that publishes CycloneDX as ECMA-424. Appears in modules 18, 34.
ECOA — Equal Credit Opportunity Act
The U.S. law, implemented by Regulation B, that bans credit discrimination and requires adverse-action reasons. Appears in module 16.
ECPA — Electronic Communications Privacy Act
U.S. federal law that includes the wiretap rules on intercepting communications, with exceptions such as prior consent of a party. Appears in module 25.
ECS — Elastic Common Schema
An open schema of field names for logs, metrics and security events, now converging into the OpenTelemetry semantic conventions. Appears in module 32.
EDA — Exploratory Data Analysis
Summarizing and plotting data to see its patterns before testing any hypothesis; PEAK's name for baseline hunting. Appears in module 32.
EdDSA — Edwards-curve Digital Signature Algorithm
A modern digital signature scheme (RFC 8032); Ed25519 is its most widely used instance. Appears in modules 0, 34.
EDPB — European Data Protection Board
The EU body of national data protection authorities that issues guidance on the GDPR. Appears in module 19.
EDR — Endpoint Detection and Response
Software on a host that records activity and detects or blocks attacks. Appears in modules 5, 28, 29.
EEA — European Economic Area
The EU member states plus Iceland, Liechtenstein and Norway. Appears in modules 31, 36.
EEI — Essential Element of Information
An observable fact that answers an intelligence requirement. Appears in module 33.
EEO — Equal Employment Opportunity
The principle and body of law against employment discrimination; the EEO-1 report collects workforce demographics. Appears in module 13.
EEOC — Equal Employment Opportunity Commission
The US federal agency that enforces federal employment anti-discrimination laws. Appears in modules 12, 13, 19, 21, 24, 25, 46, 47, 99.
Effect size
A standard measure of how big a difference or relationship is, such as d, g or δ in standard-deviation units. Appears in modules 13, 14, 21, 22, 26, 39, 41, 47.
Effective challenge —
Critical review of a model by people who are competent, independent and able to make changes happen. Appears in modules 45, 46.
Efficiency —
The share of flagged vulnerabilities that were actually exploited; precision in statistical terms. Appears in module 6.
Effort —
The share of all vulnerabilities a prioritization rule asks you to act on. Appears in module 6.
Egress blocking
Preventing a server from opening outbound network connections, so a compromised server cannot fetch or send data. Appears in module 37.
EICAR — European Institute for Computer Antivirus Research
The organization that publishes the EICAR test file, a harmless file antivirus products agree to detect so people can test scanning safely. Appears in module 28.
ELF — Executable and Linkable Format
The file format of Linux programs and libraries; it starts with the bytes 7F 45 4C 46. Appears in module 28.
Elicitation —
The effort used to draw out a model's capability, such as prompting, tools, scaffolding or fine-tuning. Appears in module 38.
Eligible assignment —
A privileged role a person may activate when needed, rather than holding permanently. Appears in module 31.
Embargo
An agreed period during which vulnerability details are kept private while a fix is prepared. Appears in modules 22, 24.
Embedding
A list of numbers (vector) representing a token or text so similar meanings are close together. Appears in modules 7, 8.
Embedding inversion
Reconstructing original text from its embedding vector. Appears in modules 7, 8.
Emergent strategy
A pattern of action that forms over time from many decisions without being planned in advance (Mintzberg and Waters). Appears in module 42.
EMM — Enterprise Mobility Management
The family of tools an organization uses to manage mobile devices, apps and data. Appears in module 35.
EMNLP — Empirical Methods in Natural Language Processing
A leading annual research conference on language technology. Appears in modules 7, 37.
Employee monitoring —
Any collection or analysis of data about employees' activity by their employer, including security monitoring. Appears in modules 3, 19, 32, 35, 43.
Emulation plan
An ordered, intelligence-backed list of adversary steps with technique IDs, safe substitutions, expected telemetry and detections, and cleanup. Appears in module 29.
ENBIS — Expected Net Benefit of Information Security investment
In the Gordon-Loeb model, the expected loss a security investment avoids minus what the investment costs. Appears in module 43.
Encapsulation —
Wrapping data from a higher network layer inside a lower layer's header, like putting a letter in an envelope. Appears in module 0.
Encrypt-then-MAC —
Encrypting first, then computing a MAC over the ciphertext and verifying it before any decryption. Appears in module 34.
Encryption context —
Associated data bound to a ciphertext or wrapped key that must match exactly for decryption to succeed. Appears in module 34.
Endpoint —
One address and method an API accepts, such as GET /candidates/{id}. Appears in modules 4, 11, 37.
Enrichment —
Adding context to an event in the pipeline, such as a role's sensitivity or whether a domain is internal. Appears in modules 20, 32.
enrollment —
Capturing a person's biometric and storing their template for the first time. Appears in module 35.
enterprise wipe —
Deleting only the work data, settings and apps from a device, leaving personal data untouched. Appears in module 35.
Entropy — Shannon entropy
A measure of unpredictability in bits. For the bytes of a file it runs from 0 to 8 bits per byte, and compressed or encrypted data scores close to 8. Appears in modules 0, 4, 27, 28, 35.
Envelope encryption —
Encrypting data with a data key, then encrypting that data key with a master key held separately. Appears in modules 0, 31, 34.
EO — Executive Order
A directive issued by the U.S. President, such as EO 14110 (revoked by EO 14148). Appears in modules 18, 33, 34.
EOT — Expectation Over Transformation
Attacking a randomized model by following the average gradient over its random transformations or noise. Appears in modules 9, 38.
EPA — Extended Protection for Authentication
A Windows feature that binds authentication to the secure channel or service, which helps block NTLM relay. Appears in module 3.
Epoch
One full pass of training over every example in the training set. Appears in modules 9, 37.
EPSS — Exploit Prediction Scoring System
FIRST's daily estimate of the probability that a published vulnerability will be exploited in the next 30 days. Appears in modules 1, 2, 3, 5, 6.
EQ — Equivalence set (CVSS v4.0)
One of six groupings of CVSS v4.0 metrics; each vector gets a level in each group, and the levels together form its MacroVector. Appears in module 6.
Equalized odds — Equalized odds (separation)
Equal true positive and false positive rates across groups. Appears in module 13.
ERM — Enterprise Risk Management
The process a company uses to identify, assess and manage all its major risks together, against its objectives. Appears in modules 42, 43, 46.
ERP — Enterprise Resource Planning
Software that runs a company's core business processes, such as finance, purchasing and HR, in one system. Appears in module 40.
ES — Expected Shortfall
The average loss in the outcomes worse than the Value at Risk; a coherent measure of how bad the tail is. Appears in module 46.
Escalation of commitment —
Continuing to invest in a failing course of action because of what has already been spent. Appears in module 40.
Escalation primitive —
A single permission, such as passing a role or minting a token, that creates one hop on a privilege escalation path. Appears in module 31.
ESI — Electronically Stored Information
The US civil procedure term for digital information that may be subject to discovery, such as databases, logs, email and chat. Appears in module 30.
Estimative language —
Standard words, such as likely or unlikely, used to express probability in intelligence judgments. Appears in module 33.
ESXi — VMware ESXi
A widely used hypervisor that runs virtual machines directly on server hardware; an ATT&CK Enterprise platform. Appears in module 10.
etcd —
The key-value database that stores Kubernetes cluster state, including Secrets. Appears in module 17.
EU — European Union
The union of 27 European member states whose regulations, such as the AI Act, apply directly in each of them. Appears in modules 12, 19, 20, 21, 22, 23, 24, 25, 26, 29, 30, 31, 32, 35, 36, 37, 40, 42, 43, 44, 45, 46, 99.
EU AI Act — Regulation (EU) 2024/1689 laying down harmonized rules on artificial intelligence
The EU's binding AI law, which classifies hiring AI as high-risk under Annex III. Appears in modules 12, 14, 15, 19, 20, 21, 22, 23, 24, 29, 31, 35, 36, 37, 39, 40, 42, 44, 45, 46, 47, 99.
EUR — Euro
The EU currency (in fines) and, in EUR-Lex, part of the name of the EU's official law database. Appears in modules 12, 36, 43, 45, 46, 99.
EV — Earned Value
The budgeted cost of the work actually completed to date. Appears in module 40.
Eval harness — Evaluation harness
Software that runs evaluations in a standard, logged way: it loads tasks, formats prompts, calls the model and scores outputs. Appears in module 37.
Event class —
In OCSF, a named kind of event (such as Account Change) with its own identifier and attributes. Appears in module 32.
Evidence index
A table linking each test to its stored evidence files and their hashes. Appears in modules 6, 14.
Evidence pack
A set of dated, version-specific artifacts mapped to the requirements they show were met. Appears in module 37.
EVM — Earned Value Management
A method that compares planned value, earned value and actual cost to measure project cost and schedule performance. Appears in module 40.
EVTX — Windows XML Event Log file
The file format Windows uses to store event logs. Appears in module 32.
Exception — Exception (deviation)
A sampled instance in which a control did not operate as described; a report can list exceptions and still have a clean opinion. Appears in module 12.
Excessive agency
OWASP LLM06: harmful actions caused by giving an LLM system too much functionality, permission or autonomy. Appears in modules 5, 7, 8, 12, 14, 15, 20, 23, 24, 32, 39, 99.
exculpation —
A charter provision removing directors' personal money liability for certain breaches of duty, such as care but not loyalty. Appears in module 47.
executive presence —
The impression that a person is ready for senior leadership; research describes it as an audience perception, so it can carry the audience's biases. Appears in modules 39, 47.
Exfiltration —
Stealing data out of a network or system; in MITRE ATT&CK it is the tactic TA0010. Appears in modules 8, 9, 10, 11, 15, 17, 20, 29, 30, 31, 32, 33, 43, 99.
EXP-301 — OffSec Windows User Mode Exploit Development
OffSec's exploit development course, which leads to the OSED certification. Appears in modules 27, 39.
Expected loss
The probability of a loss event times the size of the loss. Appears in modules 1, 2, 9, 12, 26, 34, 40, 42, 43, 44, 46.
Expertise reversal effect —
Instructional support that helps beginners, such as worked examples, can stop helping or even hurt more experienced learners. Appears in module 39.
Explicit deny
A policy statement that forbids an action and overrides any allow. Appears in module 17.
Exploit
A method that triggers a vulnerability to make a system do something unintended. Appears in modules 5, 27.
Exposure — Exposure (memorization metric)
log2 of the number of possible canary fillings minus log2 of the canary's rank; how many bits of a secret a model reveals. Appears in module 37.
Exposure index —
This module's rough measure of credential risk: the sum, over credentials, of the number of sensitive operations each allows times its lifetime in days. Appears in module 25.

F

F3EAD — Find, Fix, Finish, Exploit, Analyze, Disseminate
An operations-focused intelligence cycle for answering urgent, narrow questions. Appears in module 33.
FAIR — Factor Analysis of Information Risk
A quantitative risk model that splits risk into loss event frequency and loss magnitude. Appears in modules 12, 14, 42, 43, 46.
FAL — Federation Assurance Level
NIST SP 800-63-4's measure of how strongly a federated identity assertion is protected, from FAL1 (forged assertions) to FAL3 (identity provider compromise). Appears in module 25.
FAL1 — Federation Assurance Level 1
The lowest of NIST SP 800-63's three levels for how securely a sign-in is passed from an identity provider to an application. Appears in module 25.
FAL2 — Federation Assurance Level 2
The middle of NIST SP 800-63's three federation levels. Appears in module 25.
FAL3 — Federation Assurance Level 3
The highest of NIST SP 800-63's three federation levels. Appears in module 25.
False discovery rate —
The expected share of false alarms among the results that are flagged as significant. Appears in modules 21, 32.
False positive rate —
The share of truly negative cases that a test or model wrongly flags as positive. Appears in modules 2, 9, 10, 12, 13, 14, 20, 25, 28, 32, 33, 35, 38.
FAQ — Frequently Asked Questions
A document of common questions and official answers, such as DCWP's AEDT FAQ. Appears in modules 10, 12, 13, 14, 24, 30, 36, 37, 39, 47, 99.
FAR — Federal Acquisition Regulation
The rules for U.S. federal purchasing, which can impose security requirements on contractors. Appears in module 34.
FAT — File Allocation Table
An older, simple file system family; it stores times in local time and with coarse resolution. Appears in modules 30, 35, 37.
FBI — Federal Bureau of Investigation
The principal U.S. federal investigative agency; its Internet Crime Complaint Center (IC3) publishes public service announcements on fraud. Appears in modules 25, 27, 33.
FCA — Financial Conduct Authority
The UK regulator for financial services firms and markets. Appears in module 46.
FCC — Federal Communications Commission
The U.S. regulator for communications, which participates in the U.S. Cyber Trust Mark labeling program. Appears in module 35.
FCRA — Fair Credit Reporting Act
U.S. law that regulates consumer reports, including many background checks used in hiring. Appears in modules 6, 16, 36.
FDIC — Federal Deposit Insurance Corporation
The US federal agency that insures bank deposits and supervises many banks. Appears in modules 45, 46.
FDR — False Discovery Rate
The expected share of false alarms among the results that are flagged as significant. Appears in modules 21, 24, 32.
Federated model
An operating model where a central team sets standards and platforms and business units run their own teams within them. Appears in module 42.
Feedback intervention theory
Kluger and DeNisi's account of why feedback helps when it focuses on the task and can harm when it focuses on the self. Appears in modules 26, 41, 47.
FEHA — Fair Employment and Housing Act
California's main anti-discrimination statute for employment and housing. Appears in modules 12, 19.
FGSM — Fast Gradient Sign Method
A one-step attack that nudges every input feature by a fixed amount in the direction that most increases the model's error. Appears in module 9.
FHE — Fully homomorphic encryption
Encryption that allows any computation to be performed on ciphertexts, producing an encrypted result. Appears in module 34.
FI — Feedback Intervention
Feedback given to someone about their performance, as studied in Kluger and DeNisi's research on when feedback helps or harms. Appears in module 41.
fiduciary duty —
A legal duty to act in another party's interest; directors owe fiduciary duties to the corporation and its shareholders. Appears in module 47.
FIN — Finish
A TCP flag that closes a connection; also used in some scan types. Appears in module 2.
Fine-tuning —
Further training of an already-trained model on a smaller, specific dataset to change its behavior. Appears in modules 7, 8, 9, 12, 16, 18, 23, 24, 30, 31, 37, 38.
FinOps — Cloud Financial Operations
A practice for managing variable cloud and technology spending through collaboration between engineering, finance and business teams. Appears in modules 40, 44, 45.
FIPPs — Fair Information Practice Principles
A long-standing set of rules for handling personal data: notice, choice, purpose limitation, minimization, accuracy, security, access and accountability. Appears in module 36.
FIPS — Federal Information Processing Standard
A standard issued by NIST for U.S. federal computer systems, such as FIPS 197 for AES. Appears in modules 0, 34, 39.
Firewall —
A device or program that allows or blocks network traffic according to rules; default deny is the safest policy. Appears in modules 0, 1, 2, 3, 4, 5, 6, 17, 29, 31, 35, 40, 41, 43, 46.
firmware —
Software stored in a device's non-volatile memory that starts and runs the device. Appears in modules 2, 34, 35.
FIRST — Forum of Incident Response and Security Teams
The body that maintains CVSS and EPSS. Appears in modules 2, 4, 5, 6, 8, 14, 18, 22, 27, 28, 33, 39.
Fisher's exact test
An exact significance test for a 2x2 table, used when expected cell counts are small. Appears in modules 6, 8, 13, 14, 21, 23, 35.
Five forces — Porter's five forces
Rivalry, threat of entrants, threat of substitutes, buyer power and supplier power, which together shape an industry's profit potential. Appears in module 42.
Fixed-pie bias
Assuming the other side's interests are directly opposed to yours, which hides trades that help both. Appears in module 41.
Flip rate —
The share of pairwise judge verdicts that change when only the order of the two responses is swapped. Appears in modules 14, 16, 36, 37.
FLSA — Fair Labor Standards Act
The U.S. federal law that sets minimum wage, overtime pay and child labor rules. Appears in module 12.
FMR — False Match Rate
The share of comparisons between two different people that a biometric system wrongly declares a match. Appears in modules 25, 35.
FNIR — False Negative Identification Rate
In one-to-many search, the share of searches for enrolled people that fail to find them. Appears in module 35.
FNMR — False Non-Match Rate
The share of comparisons of the same person that a biometric system wrongly declares not a match. Appears in modules 25, 35.
FNR — False Negative Rate
The share of truly qualified people that a tool rejects; equal to one minus the TPR. Appears in module 13.
FOCUS — FinOps Open Cost and Usage Specification
An open specification that defines a common schema for technology cost and usage data across providers. Appears in module 40.
Forecast calibration —
How closely stated probabilities match how often events actually happen. Appears in module 33.
Forensic image —
A bit-for-bit copy of a storage device or memory, including unused space, made for examination. Appears in module 30.
Form 10-K
The annual report a U.S. public company files with the SEC; Item 1C holds its cybersecurity disclosure. Appears in modules 43, 47.
Form 8-K
The SEC form a public company files to report specified major events between periodic reports. Appears in modules 42, 43, 47.
Format-string bug
Passing untrusted data as the format argument of a printf-style function (CWE-134). Appears in module 27.
Forward proxy —
A server in front of users that makes outbound requests on their behalf, often for filtering and logging. Appears in module 0.
Forward secrecy —
A property of key exchange in which stealing a long-term private key later does not expose sessions recorded earlier. Appears in module 0.
Four-fifths rule — Four-fifths (80%) rule
The Uniform Guidelines' rule of thumb that an impact ratio below 0.8 is generally regarded as evidence of adverse impact. Appears in modules 0, 1, 7, 12, 13, 14, 16, 19, 21, 22, 24, 25, 26, 33, 35, 36, 38, 40, 41, 43, 44, 45, 99.
FP16 — 16-bit floating point
A half-precision number format that uses 2 bytes per value, with a smaller range than FP32. Appears in module 37.
FP32 — 32-bit floating point
A number format that uses 4 bytes per value; the usual format for optimizer state in mixed-precision training. Appears in module 37.
FPE — Format-preserving encryption
Encryption whose output has the same format as its input, for example nine digits to nine digits; NIST SP 800-38G covers it. Appears in module 34.
FPIR — False Positive Identification Rate
In a one-to-many biometric search, the share of searches that wrongly return another person as a match. Appears in modules 25, 35.
FPR — False Positive Rate
The share of real negatives a test wrongly flags. Appears in modules 9, 13, 14, 20, 25, 28, 32, 38.
FPRs — False Positive Rates
Plural of FPR. Appears in module 13.
FR — Federal Register
The U.S. government's daily publication of rules, notices and presidential documents. Appears in modules 16, 18, 25, 42, 43, 46, 47.
Framing effect
Different choices depending on whether the same options are described as gains or losses. Appears in module 41.
FRCP — Federal Rules of Civil Procedure
The rules for civil cases in U.S. federal courts. Rule 26 sets what a retained expert's written report must contain. Appears in modules 22, 26, 30.
FRE — Federal Rules of Evidence
The rules for admitting evidence in U.S. federal courts; Rule 702 covers expert testimony. Appears in modules 21, 22, 30.
Free cash flow
Commonly, operating cash flow minus capital expenditure: cash available for debt, buybacks or acquisitions. Appears in module 42.
FRIA — Fundamental Rights Impact Assessment
An assessment that certain deployers of high-risk AI must perform under Article 27 of the AI Act. Appears in module 12.
FRM — Financial Risk Manager
A credential from GARP focused on financial risk measurement and management. Appears in module 46.
FRVT — Face Recognition Vendor Test
NIST's ongoing evaluation of face recognition algorithms; Part 3 (NISTIR 8280) studied demographic effects. Appears in modules 25, 35.
FSB — Financial Stability Board
An international body that coordinates financial regulators; it published Principles for an Effective Risk Appetite Framework in 2013. Appears in module 46.
FSF — Frontier Safety Framework
Google DeepMind's published framework of protocols for severe risks from frontier models; Version 3.0 was published on 22 September 2025. Appears in module 38.
FSTM — Firmware Security Testing Methodology
An OWASP methodology for analyzing firmware security step by step. Appears in module 35.
FTC — Federal Trade Commission
The US federal agency that enforces consumer protection law, including parts of privacy and credit reporting law. Appears in modules 36, 43.
FTE — Full-Time Equivalent
A unit of staff time equal to one person working full time. Appears in module 40.
FTP — File Transfer Protocol
An old protocol for moving files that sends data, including passwords, unencrypted. Appears in module 2.
FTSE — Financial Times Stock Exchange
The name behind the FTSE stock indexes, such as the FTSE 100 of large UK companies. Appears in module 47.
function code —
The number in a Modbus request that says what to do, such as 3 to read holding registers or 6 to write one register. Appears in module 35.
Fuzz target — Fuzz target (harness)
The small function that passes the fuzzer's bytes to the code under test. Appears in module 27.
Fuzzing — Fuzz testing
Automatically feeding a program many generated inputs and watching for crashes, hangs or sanitizer reports. Appears in modules 5, 27, 35.
Fuzzy hash
A hash designed so that similar files get similar values, giving a similarity score instead of an exact match. Appears in module 28.
FWER — Family-Wise Error Rate
The probability of at least one false alarm across a set of related tests. Appears in module 21.

G

GAAP — Generally accepted accounting principles
The accounting rules U.S. companies follow when preparing financial statements. Appears in module 42.
GAI — Generative Artificial Intelligence
AI that produces text, images, audio or code. NIST AI 600-1 is the Generative AI Profile of the AI RMF. Appears in modules 12, 20, 30.
GAI PWG — Generative AI Public Working Group
The NIST public working group whose input shaped NIST AI 600-1. Appears in module 12.
GAIPS — GIAC AI Platform Security
A GIAC certification in securing AI platforms. Appears in module 39.
garak — Generative AI Red-teaming & Assessment Kit
NVIDIA's open-source scanner that probes language models for known weaknesses. Appears in modules 7, 11, 12, 14, 23, 24, 29, 39.
Garbled circuit —
Yao's two-party technique in which one party encrypts a Boolean circuit and the other evaluates it without learning intermediate values. Appears in module 34.
GARP — Global Association of Risk Professionals
The association that awards the Financial Risk Manager credential. Appears in module 46.
Gate —
A pass-or-fail minimum requirement that a proposal must meet before it is scored. Appears in module 40.
GB — Gigabyte
A unit of data size, about a billion bytes. Appears in modules 5, 18, 20, 28, 32, 34, 37.
GC — General Counsel
The chief lawyer of an organization; at the U.S. NLRB, the official whose memos set enforcement priorities. Appears in modules 25, 32.
GCC — GNU Compiler Collection
A widely used open-source compiler suite for C, C++ and other languages. Appears in module 27.
GCDA — GIAC Certified Detection Analyst
A GIAC certification in building detections and security monitoring. Appears in module 39.
GCFA — GIAC Certified Forensic Analyst
A GIAC certification covering forensic collection and analysis and advanced incident handling. Appears in modules 30, 39.
GCFE — GIAC Certified Forensic Examiner
A GIAC certification covering forensic examination of computer systems. Appears in module 30.
GCFR — GIAC Cloud Forensics Responder
A GIAC certification in cloud forensics and incident response. Appears in module 30.
GCG — Greedy Coordinate Gradient
The search method Zou et al. (2023) used to find adversarial suffixes. Appears in modules 8, 38.
GCHQ — Government Communications Headquarters
The UK's signals intelligence and cyber security agency. Appears in module 0.
GCM — Galois/Counter Mode
An authenticated-encryption mode for block ciphers such as AES, specified in NIST SP 800-38D; its IVs must never repeat under one key. Appears in modules 0, 34.
GCSA — GIAC Cloud Security Automation
A GIAC certification on DevSecOps, infrastructure as code security and cloud automation. Appears in module 31.
GCTI — GIAC Cyber Threat Intelligence
A GIAC certification covering strategic, operational and tactical cyber threat intelligence. Appears in modules 33, 39.
GDPR — General Data Protection Regulation
The EU's data protection law, Regulation (EU) 2016/679. Appears in modules 12, 15, 19, 20, 25, 30, 31, 32, 33, 35, 36, 40, 46.
GenAI — Generative AI
AI systems that produce new content, such as text, images or code, rather than only sorting or scoring inputs. Appears in modules 9, 15, 16, 20, 22, 23, 24, 28, 30, 32, 39, 99.
Generator (garak) —
A garak plugin that sends prompts to the target and returns its outputs. Appears in module 11.
GGUF — GGUF (llama.cpp model file format)
The single-file model format used by llama.cpp, which stores weights (often quantized) together with the tokenizer and metadata. Appears in module 37.
GHASH —
The polynomial hash inside GCM that computes the authentication tag using a secret hash subkey H. Appears in module 34.
Ghidra
An open-source software reverse engineering framework from the NSA Research Directorate, with disassembly, decompilation and scripting. Appears in module 28.
GIAC — Global Information Assurance Certification
The certification body associated with SANS that offers GPEN and GWAPT. Appears in modules 27, 28, 30, 31, 33, 35, 39, 99.
GICSP — Global Industrial Cyber Security Professional
A GIAC certification covering industrial control system security. Appears in modules 35, 39.
GID — Group identifier
The number Linux uses to identify a group of users. Appears in module 0.
GINA — Genetic Information Nondiscrimination Act
The US federal law that prohibits genetic information discrimination in employment and health insurance. Appears in module 30.
glass cliff —
The finding that women are more likely than men to be appointed to leadership roles in organizations already in difficulty. Appears in module 47.
Global filter —
A Sigma filter file that applies the same exclusion to many rules, used for local tuning without editing shared rules. Appears in module 32.
GMAC — Galois Message Authentication Code
The authentication part of AES-GCM, used on its own to check that data was not altered. Appears in modules 0, 34.
GMLE — GIAC Machine Learning Engineer
A GIAC certification in applying machine learning to security problems. Appears in module 39.
GMOB — GIAC Mobile Device Security Analyst
A GIAC certification in mobile device and app security. Appears in module 39.
gMSA — Group Managed Service Account
An AD service account whose password the domain generates and rotates automatically. Appears in module 3.
GNFA — GIAC Network Forensic Analyst
A GIAC certification in network forensic analysis. Appears in module 30.
GNU — GNU's Not Unix
A free-software project started in 1983; its tools and licenses (such as the GNU GPL) are used throughout Linux systems. Appears in modules 0, 27, 33.
GOAA — GIAC Offensive AI Analyst
A GIAC certification on offensive uses and testing of AI. Appears in module 39.
Goal-setting theory
The theory, summarized by Locke and Latham, that specific, difficult goals generally raise performance more than vague goals. Appears in module 41.
GOAT — Generative Offensive Agent Tester
Meta's automated multi-turn red-teaming method, offered as a promptfoo strategy. Appears in module 11.
Golden set
A small, stable, hand-checked set of test items used to detect changes between model versions. Appears in module 37.
Goodhart's law —
The idea that when a measure becomes a target, it stops being a good measure. Appears in modules 26, 29, 38, 40, 41, 44, 45.
Gordon-Loeb model
An economic model of the optimal security investment for an information set, given its vulnerability and potential loss. Appears in modules 42, 43.
Govern Function —
The NIST CSF 2.0 Function covering organizational context, risk strategy, roles, policy, oversight and supply chain risk. Appears in modules 12, 26, 39, 42, 43, 44.
GPCS — GIAC Public Cloud Security
A vendor-neutral GIAC certification on securing AWS, Azure and Google Cloud. Appears in modules 31, 39.
GPEN — GIAC Penetration Tester
A GIAC certification covering penetration-test planning, exploitation, password and Azure attacks. Appears in modules 39, 99.
GPO — Group Policy Object
A bundle of Windows settings applied to users and computers through Active Directory. Appears in module 6.
GPS — Global Positioning System
The U.S. satellite system that lets a device work out its precise location. Appears in module 35.
GPT — Generative Pre-trained Transformer
A family of large language models; the name describes how they are built and trained. Appears in modules 1, 7, 8, 9, 10, 16, 20, 22, 23, 24, 37, 38, 45.
GPT-4 — Generative Pre-trained Transformer 4
An OpenAI large language model often used as an LLM judge in research. Appears in modules 1, 2, 7, 8, 15, 16.
GPTQ — GPTQ (post-training quantization method)
A method that quantizes a trained model's weights layer by layer, using approximate curvature information to compensate for rounding errors. Appears in module 37.
GPU — Graphics Processing Unit
A processor built for parallel math, widely used to train and run AI models. Appears in modules 5, 8, 11, 16, 17, 18, 24, 25, 27, 36, 37.
GQA — Grouped-Query Attention
An attention design in which several query heads share one key-value head, which shrinks the KV cache. Appears in module 37.
Graded outcome score
PTEF's 0 to 5 scale for each tested technique, from no visibility to full response. Appears in module 29.
Gradient
The list of slopes of a function, one per input; it points in the direction that increases the output fastest. Appears in modules 7, 9, 37, 38.
Gradient descent
Repeatedly taking a small step against the gradient to make a function smaller; how models are trained. Appears in modules 7, 8, 9, 37, 38.
Gradient masking —
When a defense makes gradients useless to an attacker without making the model robust; also called obfuscated gradients. Appears in module 38.
GRC — Governance, Risk and Compliance
The function that sets policies, manages the risk register and gathers evidence that controls work. Appears in module 41.
GREM — GIAC Reverse Engineering Malware
A GIAC certification in malware analysis and reverse engineering, aligned to SANS FOR610. Appears in modules 28, 39.
GRID — GIAC Response and Industrial Defense
A GIAC certification aligned with ICS detection and incident response. Appears in module 35.
Ground truth
The known, correct values used to score a measurement or an audit. Appears in modules 2, 8, 11, 13, 14, 16, 27, 38.
Grover's algorithm —
A quantum search algorithm that finds a marked item among N in about the square root of N steps. Appears in module 34.
GRTP — GIAC Red Team Professional
A GIAC certification in red team operations. Appears in module 39.
GSM1k — Grade School Math 1K
A newer set of about 1,000 problems built to mirror GSM8k, used to test for contamination. Appears in module 16.
GSM8k — Grade School Math 8K
A public benchmark of about 8,000 grade-school math word problems. Appears in module 16.
Guardrail —
A high-level cloud governance rule, preventive, detective or proactive, that workload teams cannot switch off. Appears in module 31.
GUID — Globally Unique Identifier
A long, effectively unique ID that is hard to guess; it slows but does not prevent BOLA. Appears in modules 4, 29.
GV — Govern
The CSF 2.0 Govern function, and the prefix for GOVERN suggested actions in NIST AI 600-1. Appears in modules 12, 20, 26, 30, 42, 43, 47.
GWAPT — GIAC Web Application Penetration Tester
A GIAC certification covering web application penetration testing. Appears in modules 39, 99.
GXPN — GIAC Exploit Researcher and Advanced Penetration Tester
A GIAC certification in exploit research; SANS SEC660 prepares learners for it. Appears in modules 27, 39.

H

Half-life (memory) —
The time after which the chance of recalling an item has dropped to one half. Appears in module 39.
Hallucination
Model output that is fabricated or unsupported by its source. Appears in modules 7, 8, 11, 12.
Handling policy — Malware handling policy
Written rules for how suspicious samples are collected, stored, analyzed, shared and destroyed. Appears in module 28.
Harm taxonomy —
The structured list of harms a campaign tests for. Appears in modules 11, 29.
Hash function —
A keyless function that turns any input into a fixed-length digest; a small change in the input changes the digest completely. Appears in modules 0, 6, 18, 20, 30, 34, 35.
HB — House Bill
A bill introduced in a state House of Representatives, such as Illinois HB 3773. Appears in modules 12, 19.
HCP — HashiCorp Cloud Platform
HashiCorp's hosted services, including HCP Terraform. Appears in module 31.
Heartbeat detection —
An alert that fires when a log source sends no events for longer than its normal quiet period. Appears in module 32.
HHI — Herfindahl-Hirschman Index
A concentration measure: the sum of the squared percentage shares of each supplier in a category; larger values mean more concentration. Appears in modules 44, 46.
HHS — Department of Health and Human Services
The US federal department that enforces HIPAA and receives HIPAA breach reports. Appears in module 36.
High-impact AI —
In OMB M-25-21, AI whose output is a principal basis for decisions with a legal, material, binding or significant effect on rights or safety. Appears in module 45.
HIPAA — Health Insurance Portability and Accountability Act
U.S. law that protects health information held by covered organizations. Appears in modules 6, 34, 36.
historian —
A database that records process values over time in an industrial or building system. Appears in module 35.
hit@k —
The share of prompts for which at least one of k attempts succeeds. Appears in module 11.
HITL — Human-in-the-Loop
A design where a person must approve certain actions before the system carries them out. Appears in module 15.
HKDF — HMAC-based Extract-and-Expand Key Derivation Function
A standard way (RFC 5869) to turn a shared secret into one or more independent keys for different purposes. Appears in module 34.
HMAC — Hash-based Message Authentication Code
A way to prove a message came from someone who holds a secret key and was not changed. Appears in modules 0, 3, 25, 34.
HMI — Human-Machine Interface
The screen and software an operator uses to see and command an industrial process. Appears in module 35.
HNDL — Harvest now, decrypt later
Recording encrypted data today in order to decrypt it once a capable quantum computer exists. Appears in module 34.
Homogeneity attack
Learning a person's sensitive value because everyone in their k-anonymous group shares it. Appears in modules 19, 36.
Homogenization
Many systems producing the same outputs, so one model's errors or blind spots spread across many users. Appears in module 12.
Honey SPN — Decoy service principal name
A fake service account that no legitimate process uses, so any ticket request for it is suspicious. Appears in module 3.
Horizon scanning
A systematic search for early signals of emerging risks and changes in the environment. Appears in module 46.
Hosted lab — Hosted simulated lab
A course-run, browser-based simulation of a hiring app, with synthetic data and planted flaws, that learners are authorized to test. Nothing is installed on the learner's computer. Appears in modules 1, 2, 3, 4, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 23, 24, 26, 39.
Hosted simulation
A deliberately vulnerable lab application the course runs on its own servers, reached through a browser. Appears in modules 8, 23.
HQC — Hamming Quasi-Cyclic
A post-quantum key-encapsulation mechanism that NIST selected for standardization in March 2025. Appears in modules 0, 34.
HR — Human Resources
The function that recruits, pays, develops and manages an organization's people. Appears in modules 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 99.
HRIS — Human resources information system
The system of record for employee data. Appears in modules 1, 2, 3, 4, 5, 12, 15, 17, 25, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 39, 40, 41, 42, 43, 44, 45, 46, 47.
HSM — Hardware Security Module
Tamper-resistant hardware that generates and stores cryptographic keys and performs operations with them. Appears in modules 31, 34, 35.
HSTS — HTTP Strict Transport Security
A response header that tells browsers to contact a site only over HTTPS for a stated period. Appears in module 0.
HTB — Hack The Box
A company that runs hosted, legal hacking and defense labs, including HTB Academy courses. Appears in modules 2, 3, 5, 6, 23, 39, 99.
HTML — HyperText Markup Language
The language web pages are written in. Appears in modules 4, 8, 10, 11, 12, 14, 23, 33.
HTTP — Hypertext Transfer Protocol
The protocol browsers and web servers use to exchange pages and data. Appears in modules 0, 2, 4, 5, 7, 11, 12, 13, 14, 15, 19, 22, 24, 25, 28, 35, 37, 44, 99.
HTTPS — HTTP Secure
HTTP carried over TLS encryption. Appears in modules 0, 4, 17, 28, 34, 35.
Hub-and-spoke model
An operating model with a central hub of experts and platforms and embedded spokes in business units. Appears in module 42.
Hunting hypothesis —
A testable statement about adversary behavior in your environment, specific enough that data can support or refute it. Appears in module 32.
HVAC — Heating, Ventilation and Air Conditioning
The systems that control a building's temperature, airflow and air quality. Appears in module 35.
Hypergeometric probability —
The probability of one particular split of passes between two groups when the total number of passes is fixed; the basis of Fisher's exact test. Appears in module 13.
Hyperparameter
A training setting chosen by the engineer rather than learned, such as the learning rate, number of epochs or LoRA rank. Appears in modules 18, 37, 38, 45.

I

I-9 — Form I-9, Employment Eligibility Verification
The U.S. form on which employers record that they examined documents showing a new hire's identity and authorization to work. Appears in modules 25, 34.
I-O — Industrial-Organizational (psychology)
The field of psychology that studies work, including personnel selection and its validity and fairness. Appears in modules 0, 1, 2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 15, 16, 17, 19, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 99.
I-O psychology — Industrial-organizational psychology
The study of people at work, including job analysis, selection, performance measurement and development. Appears in modules 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 22, 23, 24, 25, 26, 28, 29, 30, 32, 33, 35, 36, 38, 40, 41, 42, 43, 44, 45, 46, 47, 99.
IaaS — Infrastructure as a Service
Rented computing, storage and networking from a cloud provider. Appears in modules 10, 17.
IaC — Infrastructure as Code
Defining cloud infrastructure in versioned, reviewable files that automation applies. Appears in module 31.
IACS — Industrial Automation and Control Systems
The term the ISA/IEC 62443 standards use for industrial control systems and their supporting systems. Appears in module 35.
IAL — Identity Assurance Level
NIST SP 800-63-4's measure of how strongly a person's claimed identity has been proofed, from IAL1 (lowest) to IAL3 (highest). Appears in module 25.
IAL1 — Identity Assurance Level 1
The lowest of NIST SP 800-63's three identity-proofing levels. Appears in module 25.
IAL2 — Identity Assurance Level 2
The middle of NIST SP 800-63's three identity-proofing levels, the one most remote hiring checks aim for. Appears in module 25.
IAL3 — Identity Assurance Level 3
The highest of NIST SP 800-63's three identity-proofing levels, which calls for supervised, in-person-grade proofing. Appears in module 25.
IAM — Identity and Access Management
The rules and systems that decide who (or what) can do which actions on which resources. Appears in modules 15, 17, 30, 31, 34.
IANA — Internet Assigned Numbers Authority
The body that keeps the registries of Internet numbers, such as which port number is assigned to which service. Appears in modules 0, 34.
IAP — Identity-Aware Proxy
Google Cloud's service that checks identity and access for each request in front of an application. Appears in module 17.
IAPAR — Impostor Attack Presentation Accept Rate
The share of presentation attacks of a given type that a liveness system wrongly accepts as a live person. NIST SP 800-63A-4 requires less than 0.07. Appears in module 25.
IAPP — International Association of Privacy Professionals
A not-for-profit association that offers privacy and AI governance certifications. Appears in modules 12, 22, 24, 36, 39, 45, 99.
IAST — Interactive Application Security Testing
Testing with instrumentation inside the running application. Appears in module 8.
IBT — Indirect Branch Tracking
A CET feature that checks indirect calls and jumps land on compiler-marked targets. Appears in module 27.
IC — Individual contributor
A career track where seniority grows through the difficulty and reach of your own work rather than through managing people. Appears in module 26.
IC3 — Internet Crime Complaint Center
The FBI unit that receives internet crime complaints and publishes public service announcements, including on deepfakes in remote hiring. Appears in modules 25, 33.
ICC — Intraclass Correlation Coefficient
The share of total variance that lies between groups or clusters; also used as an agreement statistic for continuous ratings. Appears in modules 16, 40.
ICCP — Inter-Control Center Communications Protocol
A protocol used to exchange data between utility control centers. Appears in module 35.
ICD — Intelligence Community Directive
A directive issued by the US Director of National Intelligence; ICD 203 sets analytic standards. Appears in module 33.
ICFR — Internal Control over Financial Reporting
The controls that support reliable financial statements; the subject of SOC 1 reports. Appears in module 12.
ICLR — International Conference on Learning Representations
A major machine learning research conference. Appears in modules 9, 37, 38.
ICML — International Conference on Machine Learning
A major machine learning research conference. Appears in modules 9, 11, 34, 37, 38.
ICMP — Internet Control Message Protocol
The protocol behind ping and many network error messages. Appears in module 2.
ICS — Industrial Control Systems
Computer systems that run physical processes such as utilities and factories; ATT&CK has a separate matrix for them. Appears in modules 10, 35, 39.
ICT — Information and Communications Technology
Computing and communications systems and services; NIST uses the term in supply chain guidance. Appears in modules 40, 43, 44, 46.
ID — Identifier
A value that names one specific thing, such as a test case, a record or a technique. Appears in modules 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 14, 15, 16, 17, 19, 20, 23, 24, 25, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 39, 45, 99.
Idempotency key —
A unique message identifier the receiver records so that the same message is processed only once, defeating duplicates and replays. Appears in module 25.
identification (1:N) —
Searching a sample against all enrolled templates to find who the person is. Appears in module 35.
Identifying — Identifying (LINDDUN)
A LINDDUN threat: a person's identity can be learned from data. Appears in module 36.
Identity proofing —
Establishing that a person is who they claim to be: resolving the identity, validating evidence and attributes, and verifying that the person owns them. Appears in modules 10, 25, 33.
IDOR — Insecure direct object reference
Access to an object by changing its identifier, without an ownership check. Appears in module 4.
IdP — Identity Provider
The system that authenticates users and issues tokens or assertions to applications. Appears in modules 17, 25, 30, 31.
IDS — Intrusion Detection System
A tool that watches network or host activity and alerts on signs of attack. Appears in module 2.
IDs — Identifiers
Plural of ID. Appears in modules 0, 1, 3, 4, 5, 6, 8, 9, 10, 11, 12, 14, 15, 16, 17, 19, 20, 23, 24, 25, 27, 28, 29, 30, 31, 32, 33, 35, 36, 37, 39, 40, 99.
IEC — International Electrotechnical Commission
An international standards body that co-publishes information-technology standards with ISO. Appears in modules 12, 35, 43.
IEEE — Institute of Electrical and Electronics Engineers
A professional association that publishes journals and runs conferences, including the IEEE Symposium on Security and Privacy. Appears in modules 7, 8, 9, 27, 28, 34, 35, 36, 38, 39, 40, 44.
IETF — Internet Engineering Task Force
The body that develops internet standards and publishes them as RFCs. Appears in modules 0, 3, 4, 17, 22, 25, 34.
IG — Integrated Gradients
A feature-attribution method that adds up gradients along a straight path from a baseline input to the actual input. Appears in module 38.
IIA — The Institute of Internal Auditors
The professional body for internal auditors; it published the Three Lines Model in 2020. Appears in module 46.
IIoT — Industrial Internet of Things
Networked sensors and devices, often cloud-connected, used in industrial settings. Appears in module 35.
ILCS — Illinois Compiled Statutes
The official codification of Illinois statutes. Appears in modules 19, 25, 35, 36.
IM — Improvement
CSF 2.0 category ID.IM: improving processes from lessons learned. Appears in modules 12, 20, 30, 36.
Image digest —
A cryptographic hash that names exact container image content; unlike a tag, it cannot be moved. Appears in modules 8, 17, 18.
IMDS — Instance Metadata Service
A service inside cloud VMs, usually at 169.254.169.254, that provides instance information and temporary credentials. Appears in module 17.
Impact ratio —
A group's selection (or scoring) rate divided by the highest group's rate. Appears in modules 0, 7, 9, 12, 13, 14, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 29, 34, 36, 37, 38, 40, 44, 45, 46.
Impact tolerance
The maximum disruption to an important service an organization will tolerate, measured in time, volume or harm. Appears in module 46.
implementation intention —
An if-then plan that links a specific situation to an action, which helps people start and keep to goals. Appears in module 26.
Implicit trust zone —
An area where everything past a checkpoint is trusted; zero trust makes such zones as small as possible. Appears in module 17.
Improper linear model
A scoring rule with rough or equal weights; research shows it usually beats unaided intuition. Appears in modules 40, 41.
In-context learning
A model performing a task from instructions or examples placed in its prompt. Appears in module 7.
Incidence rate
The share of tested applications with at least one weakness in a category. Appears in module 4.
Incident command
A structure in which one person runs the response and others fill defined roles on a fixed briefing rhythm. Appears in modules 41, 43.
Inclusive method
Reporting approach that includes and tests a subservice organization's relevant controls in the SOC 2. Appears in module 12.
indemnification —
A company's promise to pay a director's or officer's legal costs and some liabilities arising from their service. Appears in modules 43, 47.
independent director —
A director with no material relationship with the company, such as employment or consulting fees, that could compromise judgment. Appears in modules 46, 47.
Indicator of compromise —
An observable such as a file hash, domain or file path that suggests a system has been compromised. Appears in modules 28, 33.
Indirect prompt injection
Instructions hidden in content the model reads, rather than typed by the user. Appears in modules 6, 7, 8, 9, 10, 12, 13, 14, 15, 16, 20, 23, 24, 25, 28, 33, 38, 39.
Influence tactic
A specific behavior used to change someone's attitude or action, such as rational persuasion or consultation. Appears in module 41.
Information credibility —
A rating (1 to 6) of how believable a particular piece of information is, judged separately from its source. Appears in module 33.
Information-processing view
Galbraith's theory that more uncertain work needs more information processing, which shapes how an organization should be designed. Appears in modules 41, 42.
Injection
A weakness in which user data is interpreted as code or syntax by an interpreter. Appears in modules 4, 5, 27.
Injection attack —
An attack that feeds forged or altered media into a verification or video stream after the camera, for example through a virtual camera, bypassing sensor-level checks. Appears in modules 4, 7, 9, 12, 25, 33, 34.
Input privacy —
Protection of each party's inputs from the other parties in a computation. Appears in module 34.
Inside view
Forecasting from the details of the current plan while neglecting statistics from similar past cases; linked to overly optimistic forecasts. Appears in modules 40, 42, 46.
Insider threat program —
A cross-discipline capability to deter, detect and respond to harm from people with authorized access; NIST SP 800-53 control PM-12. Appears in module 25.
Inspect — Inspect (inspect_ai)
The UK AI Security Institute's open-source (MIT) framework for writing and running LLM evaluations as tasks with datasets, solvers and scorers. Appears in module 37.
INT8 — 8-bit integer
A 1-byte whole-number format used to store quantized weights or activations, together with a scaling constant. Appears in module 37.
Integrative negotiation
Negotiation that creates value for both sides, often by trading on issues valued differently. Appears in modules 41, 47.
Intelligence cycle —
The loop of direction, collection, processing, analysis, dissemination and feedback that turns requirements into intelligence. Appears in module 33.
Inter-rater reliability —
How consistently different raters give the same rating to the same thing. Appears in modules 6, 7, 8, 16, 29.
Interleaving —
Mixing different types of problem in one practice session instead of practicing one type in a block. Appears in module 39.
Intersectional category
A combination of categories, such as sex and race/ethnicity together. Appears in modules 12, 13.
Intraclass correlation —
The share of total variance that lies between groups or clusters; also used as an agreement statistic for continuous ratings. Appears in modules 11, 16, 40.
Invariant — Invariant (disclosure avoidance)
A statistic published exactly, outside the privacy accounting, as a policy choice; it weakens the overall guarantee. Appears in module 36.
IOC — Indicator of Compromise
An observable such as a file hash, domain or file path that suggests a system has been compromised. Appears in modules 28, 33.
IoT — Internet of Things
Network-connected devices such as sensors, cameras and smart equipment. Appears in modules 6, 35.
IP — Internet Protocol
The protocol that addresses and routes packets between computers; an IP address names one machine on a network. Appears in modules 0, 1, 2, 3, 4, 5, 6, 14, 17, 25, 29, 30, 31, 32, 33, 47.
IP — Intellectual Property
Creations such as model weights, code or data that an organization owns legal rights over. Appears in modules 9, 10.
IPC — Inter-Process Communication
Ways for programs on the same machine to exchange data, such as pipes or Unix domain sockets. Appears in modules 15, 35.
IPD — Initial Public Draft
NIST's label for the first draft of a publication released for public comment, such as Privacy Framework 1.1. Appears in module 36.
IPI — Indirect Prompt Injection
Malicious instructions hidden in content an AI system retrieves, such as an email or web page, rather than typed by the user. Appears in module 15.
IPS — Intrusion Prevention System
An intrusion detection system that can also block the traffic it flags. Appears in module 2.
IPs — Internet Protocol addresses
Plural of IP address. Appears in modules 14, 30.
Ipse dixit
Latin for 'he himself said it': an assertion supported only by the speaker's authority. Joiner says courts need not admit such opinions. Appears in modules 22, 23.
IPsec — Internet Protocol Security
A suite of protocols that encrypts and authenticates traffic at the IP layer, used in many VPNs. Appears in module 34.
IPv4 — Internet Protocol version 4
The original Internet Protocol, with 32-bit addresses written as four numbers such as 192.0.2.10. Appears in modules 0, 2, 30.
IPv6 — Internet Protocol version 6
The newer Internet Protocol, with 128-bit addresses written in hexadecimal groups such as 2001:db8::10. Appears in modules 0, 2.
IQR — Interquartile Range
The distance between the 25th and 75th percentiles of a set of values; used in Tukey fences. Appears in module 32.
IR — Incident Response
The organized process of detecting, containing and recovering from a security incident. Appears in modules 6, 12, 14, 20, 30, 46.
IR — Impact Ratio
A group's selection (or scoring) rate divided by the highest group's rate. Appears in modules 13, 21, 22, 23, 24.
IRB — Institutional Review Board
A committee that reviews research involving human participants to protect their rights and welfare. Appears in module 22.
IRR — Internal Rate of Return
The discount rate at which a decision's NPV equals zero; useful as a supporting figure but blind to scale. Appears in modules 40, 42.
IRs — Impact Ratios
Plural of impact ratio (IR). Appears in modules 14, 21, 22, 24.
IRS — Internal Revenue Service
The U.S. federal tax agency; tax-exempt organizations report their governance to it on Form 990. Appears in module 47.
IRT — Item Response Theory
Models that link a hidden trait to the probability of each item response, with item difficulty and discrimination parameters. Appears in module 16.
IRTF — Internet Research Task Force
The research sibling of the IETF; its Crypto Forum Research Group publishes RFCs such as RFC 8452. Appears in module 34.
ISA — International Society of Automation
A professional body that develops automation standards, including the ISA/IEC 62443 industrial cybersecurity series with the IEC. Appears in module 35.
ISAC — Information Sharing and Analysis Center
A member organization, usually for one industry sector, through which companies share threat information. Appears in module 33.
ISACA — ISACA (formerly the Information Systems Audit and Control Association)
A professional association that offers audit, governance and security certifications such as CISA and CISM. Appears in modules 39, 40, 43, 46, 47.
ISAO — Information Sharing and Analysis Organization
A threat-sharing organization of any affinity (sector, region or topic), encouraged by US Executive Order 13691 of 2015. Appears in module 33.
ISC2 — ISC2 (formerly the International Information System Security Certification Consortium)
A certification body for security professionals, known for the CISSP. Appears in modules 22, 26, 31, 39, 43.
ISECOM — Institute for Security and Open Methodologies
The organization that publishes the OSSTMM. Appears in module 1.
ISI — Interstudy Interval
The gap between two practice sessions on the same material. Appears in module 39.
ISMS — Information Security Management System
The management system for information security that ISO/IEC 27001 certifies. Appears in modules 12, 43.
ISO — International Organization for Standardization
An international body that publishes standards, many of them sold rather than free. Appears in modules 0, 12, 20, 36, 39, 43, 45, 46.
ISO/IEC — International Organization for Standardization / International Electrotechnical Commission
The two bodies that jointly publish information technology standards such as ISO/IEC 42001. Appears in modules 12, 22, 24, 25, 26, 27, 34, 36, 39, 43, 45.
ISO/IEC 42001 — ISO/IEC 42001, AI management system
A certifiable international standard for an AI management system; its full text is paywalled. Appears in modules 12, 43, 45, 99.
Isolated lab — Isolated analysis lab
Machines, usually virtual, with no route to production networks, used only for analyzing suspicious files. Appears in modules 28, 39.
ISSAP — Information Systems Security Architecture Professional
An ISC2 concentration, added on top of CISSP, for security architects. Appears in module 39.
ISSEP — Information Systems Security Engineering Professional
An ISC2 concentration, added on top of CISSP, for security engineers. Appears in module 39.
ISSMP — Information Systems Security Management Professional
An ISC2 management concentration for CISSP holders. Appears in modules 39, 43.
Issue selling
Efforts to draw senior leaders' attention to an issue and shape how they understand it. Appears in module 41.
ISTG — IoT Security Testing Guide
OWASP's guide for testing the security of IoT devices. Appears in module 35.
IT — Information Technology
The computers, networks and software an organization runs, and the team that runs them. Appears in modules 0, 1, 2, 5, 6, 10, 12, 25, 26, 29, 31, 33, 35, 36, 40, 41, 42, 43, 44, 45, 46.
Item 1.05 — Form 8-K Item 1.05
The 8-K item for material cybersecurity incidents, due within four business days of determining materiality. Appears in modules 42, 43, 47.
Item 106 — Regulation S-K Item 106
The SEC rule requiring annual disclosure of cybersecurity risk management, strategy and governance. Appears in modules 26, 42, 43, 44, 46, 47.
Item response theory —
Models that link a hidden trait to the probability of each item response, with item difficulty and discrimination parameters. Appears in modules 0, 16.
ITIL — ITIL (formerly IT Infrastructure Library)
A widely used body of IT service management guidance; ITIL 4 is the current version. Appears in module 40.
ITSM — IT Service Management
The discipline of designing, delivering and running technology as services that users value. Appears in modules 40, 44.
ITU — International Telecommunication Union
The United Nations agency for telecommunications standards; with ISO, it publishes the OSI reference model. Appears in module 0.
IV — Initialization vector
A value fed into an encryption mode alongside the key; for AES-GCM it must be unique for every encryption under the same key. Also called a nonce. Appears in modules 0, 12, 22, 24, 34, 36, 37, 99.

J

Jailbreak
A prompt injection that makes a model disregard its safety rules. Appears in modules 7, 8, 9, 10, 11, 16, 20, 23, 29, 37, 38, 39.
JDK — Java Development Kit
The Java runtime and developer tools; Ghidra needs a 64-bit JDK. Appears in module 28.
JIT — Just-In-Time provisioning
Creating a user account in an application automatically at first sign-in through SSO, using attributes from the identity provider. Appears in modules 25, 31.
JML — Joiner-Mover-Leaver
The lifecycle of a person's access: granted on joining, changed on moving role, removed on leaving. Appears in module 25.
Job analysis —
A systematic study of a job's tasks and of the knowledge, skills, abilities and other characteristics needed to do it. Appears in modules 0, 1, 10, 11, 12, 13, 14, 19, 23, 26, 39, 40, 41, 42, 46, 47.
job architecture —
The structure of job families, tracks and levels an organization uses to set titles, pay ranges and promotion rules. Appears in modules 26, 42, 45.
Job embeddedness
The links, fit and sacrifices that keep a person in their job and community. Appears in module 41.
JOP — Jump-Oriented Programming
A code-reuse technique like ROP that chains existing code fragments ending in indirect jumps. Appears in module 27.
JSON — JavaScript Object Notation
A simple text format for structured data. Appears in modules 0, 4, 6, 7, 8, 10, 11, 14, 15, 17, 18, 20, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 40, 41.
JSON-RPC — JavaScript Object Notation Remote Procedure Call
A simple format for calling functions on another program by sending JSON messages. Appears in module 15.
JSONL — JSON Lines
A text format with one JSON object per line, used for garak reports. Appears in modules 11, 32.
JSTOR — Journal Storage
A digital library of academic journals. Appears in module 21.
Judge error —
Mistakes made by the human, rule or model that labels attempts as successes or failures. Appears in modules 11, 16.
Juice Shop — OWASP Juice Shop
A deliberately insecure web application for practicing security testing on your own computer. Appears in modules 4, 8, 14, 18, 39.
JupyterLite
A Jupyter notebook that runs entirely in the browser using Pyodide. Appears in modules 6, 9, 10, 12, 30, 46.
JWT — JSON Web Token
A signed token format often used to carry identity in web APIs. Appears in modules 4, 17, 31.

K

k-anonymity
A property of a dataset in which every combination of quasi-identifier values is shared by at least k records. Appears in modules 19, 36.
Kappa — Cohen's kappa
A measure of agreement between two raters that corrects for chance: 1 is perfect, 0 is no better than chance. Appears in modules 10, 13, 16, 40.
KB — Knowledge Base article
A numbered Microsoft support article, such as KB 243327. Appears in module 3.
KB — Kilobyte
About one thousand bytes of data. Appears in module 20.
KDC — Key Distribution Center
The Kerberos service, running on each domain controller, that issues tickets. Appears in module 3.
KDS — Key Distribution Service
The domain service whose root key is used to derive gMSA passwords. Appears in module 3.
KEK — Key encryption key
In envelope encryption, the master key, usually held in a key management service, that wraps data encryption keys. Appears in modules 0, 31, 34.
KEM — Key-encapsulation mechanism
A public-key method for two parties to establish a shared secret key, used where Diffie-Hellman is used today. Appears in modules 0, 34.
Kerberoasting — Kerberoasting (ATT&CK T1558.003)
Requesting service tickets in order to guess the service account's password offline. Appears in modules 3, 4, 23.
Kerckhoffs's principle —
The rule that a cryptosystem should stay secure even if everything about it except the key is public. Appears in modules 0, 8.
KEV — Known Exploited Vulnerabilities
CISA's catalog of vulnerabilities with evidence of active exploitation, each with a required action and due date. Appears in modules 6, 27.
Key hierarchy —
A tree of keys in which each key wraps the keys below it, from a root key down to data keys. Appears in module 34.
Key policy —
The resource policy on an AWS KMS key; without its permission, IAM policies cannot grant use of the key. Appears in module 31.
Keystream —
The random-looking bytes a counter-based cipher mode XORs with plaintext; reusing it leaks plaintext. Appears in modules 0, 34.
KL — Kullback–Leibler divergence
A measure of how different two probability distributions are. Appears in modules 7, 37.
KMS — Key Management Service
A cloud service that creates cryptographic keys and controls who may use them; AWS KMS is one example. Appears in modules 31, 34.
KPI — Key Performance Indicator
A metric chosen to track progress toward a goal. Appears in modules 6, 26, 41, 43, 44.
KQL — Kusto Query Language
Microsoft's pipe-based query language for Azure Data Explorer, Microsoft Sentinel, Azure Monitor and Microsoft Fabric. Appears in module 32.
KRI — Key risk indicator
A measure that signals rising risk, such as the number of high-risk AI systems not yet tested. Appears in modules 26, 43, 44, 46.
Krippendorff's alpha —
A chance-corrected agreement statistic for any number of raters, any scale type and missing ratings. Appears in module 16.
KSA — Knowledge, Skills and Abilities
The knowledge, skills and abilities a role needs; used here to plan what the module must teach and assess. Appears in module 35.
KSAO — Knowledge, Skills, Abilities and Other characteristics
The worker attributes a job analysis identifies as needed for the job. Appears in modules 13, 41, 47.
KSAOs — Knowledge, Skills, Abilities and Other characteristics
The worker attributes a job analysis identifies as needed for a job. Appears in modules 13, 26, 41, 47.
KV — Key-value
Short for the key and value vectors that attention computes for each token; see KV cache. Appears in module 37.
KV cache — Key-value cache
The stored key and value vectors for earlier tokens that a model reuses while generating, so it does not recompute them; it grows with every token of every active request. Appears in module 37.
KVM — Keyboard-Video-Mouse
A KVM switch or device lets one keyboard, screen and mouse control computers remotely or in turn; used in laptop farms to let workers abroad control company laptops. Appears in module 25.
KYC — Know Your Customer
Identity checks that financial services run when onboarding customers, often a document plus a selfie match. Appears in module 33.

L

l-diversity
A strengthening of k-anonymity requiring each group to contain at least l well-represented sensitive values. Appears in modules 19, 36.
L0 — Build Level 0
The SLSA build level with no guarantees. Appears in modules 17, 18.
L1 — Build Level 1
The SLSA build level at which provenance exists. Appears in modules 17, 18.
L2 — Build Level 2
The SLSA build level that requires a hosted build platform. Appears in modules 17, 18.
L3 — Build Level 3
The SLSA build level that requires hardened builds. Appears in modules 17, 18.
Label shift
A change in the mix of outcomes, with no change in what the inputs for each outcome look like. Appears in module 37.
Lagging indicator
A measure of the outcome itself, known only after the fact. Appears in modules 24, 41.
LAION — Large-scale Artificial Intelligence Open Network
A non-profit that published large web-scraped image-text datasets such as LAION-400M. Appears in modules 8, 38.
LAN — Local Area Network
A network that connects devices in one building or site. Appears in module 3.
Landing zone —
A pre-built multi-account, multi-subscription or multi-project cloud environment with central identity, logging, security tooling and guardrails. Appears in modules 31, 39.
LAPS — Local Administrator Password Solution
A Windows feature that gives each machine a unique, rotating local administrator password stored in the directory. Appears in modules 3, 4.
Laptop farm —
A location where company laptops issued to remote workers are hosted and connected to remote-access devices so that workers elsewhere appear to be local. Appears in modules 25, 33.
Lateral movement
Moving from one compromised system to others in the same environment. Appears in modules 5, 10, 14, 33, 35, 99.
Lattice —
A regular grid of points in many dimensions; hard lattice problems underpin ML-KEM, ML-DSA and most FHE. Appears in modules 0, 34, 39.
Lawful basis
One of the six grounds in GDPR Article 6(1) that makes processing lawful, such as contract steps or legitimate interests. Appears in modules 1, 3, 9, 10, 12, 15, 17, 20, 23, 25, 35, 36.
Layered notice
A privacy notice with a short top layer answering the main questions and a full layer with every required detail. Appears in module 36.
Layers
The number of management levels between a leader and the frontline. Appears in module 41.
LCA — Levels of Confidence in Assessment
Fixed terms (high, moderate, low) describing how good the evidence behind a judgment is. Appears in module 33.
LDA — Loss distribution approach
Modeling annual loss by combining how often loss events happen (frequency) with how large each one is (severity). Appears in module 46.
LDAP — Lightweight Directory Access Protocol
The protocol used to query and change directory data such as Active Directory users and groups. Appears in modules 3, 4.
Leading indicator
A measure that moves before the outcome you care about and helps predict it. Appears in modules 24, 33, 41, 44.
Learning rate
How large a step training takes when it updates the model's parameters. Appears in modules 9, 37.
least intrusive means —
The principle of choosing the control that meets a security need with the smallest intrusion into people's privacy. Appears in module 35.
Least privilege — Principle of least privilege
Giving every account or process only the rights it needs. Appears in modules 0, 1, 3, 4, 5, 7, 8, 12, 15, 17, 18, 20, 23, 25, 27, 31, 34, 35, 39.
Leaver latency —
The time from an HR termination date to the disabling of every account the person held; accounts never disabled must be counted separately. Appears in module 25.
LEF — Loss Event Frequency
How often a loss event occurs; threat event frequency times vulnerability. Appears in module 12.
Legal hold —
An instruction from counsel to preserve information relevant to anticipated or current litigation, overriding normal deletion schedules. Appears in modules 19, 30, 36, 43, 45.
LeSS — Large-Scale Scrum
A framework for running Scrum with many teams working on one product. Appears in module 40.
LIA — Legitimate Interests Assessment
The documented three-part test (interest, necessity, balancing) behind relying on GDPR Article 6(1)(f). Appears in module 36.
libFuzzer
LLVM's in-process, coverage-guided, evolutionary fuzzing engine. Appears in module 27.
LII — Legal Information Institute
Cornell Law School's free online source of US statutes and regulations, used for verification in this course. Appears in modules 36, 43.
Likelihood ratio —
How much more probable a piece of evidence is under one hypothesis than under another. Appears in modules 2, 8, 33, 38.
LIME — Local Interpretable Model-agnostic Explanations
An explanation method that fits a simple model around one prediction by perturbing the input. Appears in modules 16, 38.
LINDDUN — Linking, Identifying, Non-repudiation, Detecting, Data disclosure, Unawareness and Unintervenability, Non-compliance
A threat-modeling method focused on privacy harms. Appears in modules 1, 36, 39.
Lineage —
The record of where data came from and how it was changed on its way to a report or model. Appears in modules 18, 37, 45.
Linking — Linking (LINDDUN)
A LINDDUN threat: data items can be tied together as relating to the same person, even without knowing who. Appears in module 36.
LiRA — Likelihood Ratio Attack
A strong membership inference attack that compares how a model behaves with and without a record using shadow models. Appears in modules 8, 9, 38.
Liveness check —
A test that a live person, not a photo or replayed video, is in front of the camera. Appears in modules 10, 25, 33.
liveness detection —
Checks that a biometric sample comes from a live person rather than an artifact. Appears in modules 25, 35.
LL144 — New York City Local Law 144 of 2021
NYC law requiring a bias audit, publication of results and notice before employers use automated employment decision tools. Appears in modules 12, 13, 14, 20, 21, 22, 23, 24, 29, 37, 44, 46, 47.
llama.cpp
An open-source (MIT) C/C++ project for running LLMs efficiently on CPUs and GPUs, using the GGUF format and its llama-server program. Appears in modules 11, 37.
LLM — Large Language Model
A neural network trained on very large amounts of text that predicts and generates text. Appears in modules 0, 1, 2, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 22, 23, 24, 25, 26, 28, 29, 30, 31, 32, 34, 37, 38, 39, 40, 44, 46, 99.
LLM-as-a-judge
Using a language model to grade other models' outputs; the judge must itself be validated. Appears in module 16.
LLM01 — LLM01:2025 Prompt Injection
User or external input that alters an LLM's behavior or output in unintended ways. Appears in modules 1, 6, 7, 8, 9, 14, 15, 18, 20, 22, 23, 24, 28, 99.
LLM02 — LLM02:2025 Sensitive Information Disclosure
LLM applications exposing personal, confidential or credential data through their output. Appears in modules 7, 8, 14, 99.
LLM03 — LLM03:2025 Supply Chain
Risks from third-party models, adapters, datasets, packages and platforms. Appears in modules 7, 8, 28, 99.
LLM04 — LLM04:2025 Data and Model Poisoning
Manipulating pre-training, fine-tuning or embedding data to plant bias or backdoors. Appears in modules 5, 7, 8, 23, 99.
LLM05 — LLM05:2025 Improper Output Handling
Passing model output to downstream systems without validation, sanitization or encoding. Appears in modules 7, 8, 14, 23, 24, 99.
LLM06 — LLM06:2025 Excessive Agency
Damaging actions enabled by giving an LLM too much functionality, permission or autonomy. Appears in modules 5, 7, 8, 12, 14, 15, 23, 24, 32, 39, 99.
LLM07 — LLM07:2025 System Prompt Leakage
Exposure of system prompts that contain secrets or rules, or are relied on as controls. Appears in modules 7, 8, 14, 23, 24, 99.
LLM08 — LLM08:2025 Vector and Embedding Weaknesses
Weaknesses in how RAG vectors are generated, stored or retrieved. Appears in modules 7, 8, 99.
LLM09 — LLM09:2025 Misinformation
False or misleading LLM output that appears credible, made worse by overreliance. Appears in modules 7, 8, 14, 23, 24, 99.
LLM10 — LLM10:2025 Unbounded Consumption
Uncontrolled inference leading to denial of service, cost or model theft. Appears in modules 7, 8, 14, 23, 24, 99.
LLMs — Large Language Models
Plural of LLM. Appears in modules 1, 2, 5, 6, 7, 8, 10, 11, 13, 16, 18, 19, 20, 23, 24, 25, 27, 28, 29, 30, 31, 37, 38, 39, 44, 99.
LLVM — LLVM compiler infrastructure
An open-source compiler project that includes the Clang C/C++ compiler, the sanitizers and libFuzzer. Appears in module 27.
LM — Language Model
A model that assigns probabilities to sequences of text and can generate text. Appears in modules 8, 33.
lm-evaluation-harness — Language Model Evaluation Harness
EleutherAI's open-source (MIT) software for running standard language-model benchmarks in a consistent way. Appears in module 37.
LMS — Leighton-Micali Signature
A stateful hash-based signature scheme approved in NIST SP 800-208. Appears in module 34.
LNCS — Lecture Notes in Computer Science
A Springer book series that publishes many computer science conference proceedings. Appears in module 34.
Local DP — Local differential privacy
DP where each person's device adds noise before sending data, so the collector never sees raw values; it needs much more noise. Appears in module 36.
Log pipeline —
The chain from where an event happens to where it is queried: collection, parsing, normalization, enrichment, filtering, routing and storage. Appears in modules 0, 32.
Log source —
The logsource section of a Sigma rule, which names the category, product and service of the telemetry the rule applies to. Appears in module 32.
Logistic regression —
A statistical model that predicts the probability of a yes-or-no outcome from one or more inputs. Appears in modules 9, 21, 38.
Logit
A model's raw score for a possible next token, before conversion to a probability. Appears in modules 7, 8, 38.
Logit —
The log-odds of a probability, ln(p/(1−p)). Appears in module 9.
Logprobs — Log-probabilities
API output that exposes the model's probabilities for tokens. Appears in modules 7, 8.
LOI — Letter of intent
A usually non-binding document setting out the main terms of a proposed deal before detailed due diligence. Appears in module 42.
LoRA — Low-Rank Adaptation
A PEFT method that adds small trainable matrices to a frozen model; produces small adapter files. Appears in modules 7, 8, 37.
LoRA rank — LoRA rank (r)
The width of the two small matrices in a LoRA adapter; a higher rank gives the adapter more capacity, including more capacity to memorize. Appears in module 37.
Loss exceedance curve —
A chart showing, for each loss amount, the probability that annual loss exceeds it. Appears in modules 12, 46.
Loss function
A number that measures how wrong a model's prediction is; training makes it small and evasion attacks make it large. Appears in module 9.
LPTA — Lowest Price, Technically Acceptable
A source selection method that picks the cheapest offer meeting minimum requirements; NIST SP 800-161r1 advises avoiding it for ICT purchases. Appears in module 40.
LSA — Local Security Authority
The Windows component that enforces local security policy and handles sign-ins. Appears in modules 3, 5.
LSASS — Local Security Authority Subsystem Service
The Windows process that handles sign-in and holds credentials for active sessions. Appears in modules 3, 5.
LSB — Least Significant Byte first
In file-type output, means multi-byte numbers are stored little-endian, lowest byte first. Appears in modules 27, 28.
LZMA — Lempel-Ziv-Markov chain Algorithm
A lossless compression method used by the 7z and xz formats. Appears in module 35.

M

M-ATH — Model-Assisted Threat Hunting
PEAK's hunt type in which a statistical or machine-learning model ranks or clusters data for a human to review. Appears in module 32.
M&A — Mergers and acquisitions
Transactions in which companies combine or one buys another. Appears in module 42.
MA — Incident Management / Maintenance
In CSF 2.0, RS.MA means incident management; in SP 800-53, MA is the maintenance control family. Appears in modules 12, 20.
MAC — Message authentication code
A short value computed from a message and a shared secret key, used to detect tampering and confirm the message came from a key-holder. Appears in modules 0, 34.
MAC — Modified Attack Complexity
A CVSS v4.0 environmental metric that overrides the base Attack Complexity for your own environment. Appears in module 6.
MACB — Modification, Access, Change, Birth
Shorthand for the four kinds of file timestamp that timeline tools report. Appears in module 30.
Machine unlearning
Methods to remove the influence of specific training records from a trained model without retraining from scratch. Appears in modules 8, 36, 45.
MacroVector
A CVSS v4.0 group of metric combinations judged to be of similar severity, used in scoring. Appears in module 6.
MAD — Median Absolute Deviation
The median of the distances between each value and the median; a measure of spread that outliers barely move. Appears in module 32.
Magic number
Fixed bytes at the start of a file that identify its true format regardless of its name. Appears in modules 28, 35.
Maintenance group —
A set of assets with similar characteristics that share the same maintenance plans (NIST SP 800-40 Rev. 4). Appears in module 6.
Malicious adversary —
A corrupted party that may deviate from the protocol in any way. Appears in module 34.
Malicious AI artifact
This course's umbrella term for a model file, package or document in the AI supply chain that carries an attack. Appears in module 28.
Malware — Malicious software
Software covertly placed on a system to damage data, run intrusive programs or otherwise compromise confidentiality, integrity or availability. Appears in modules 0, 2, 5, 10, 14, 20, 28, 29, 30, 33, 35, 39, 43.
MAM — Mobile Application Management
Management of only specific work apps and their data, not the whole device. Appears in module 35.
Manageability
A NIST privacy engineering objective: data can be administered at a fine grain, including altered, deleted and selectively disclosed. Appears in module 36.
Managed identity
An Azure identity for resources that lets them get tokens without stored secrets. Appears in modules 17, 31.
Managed model endpoint —
A cloud provider's hosted API for AI model inference, where the customer manages identity, network, configuration and data. Appears in module 31.
Management event —
In AWS CloudTrail, a record of a control-plane operation that configures resources; included in the default 90-day event history. Appears in modules 3, 30, 31.
Management group —
An Azure container above subscriptions to which policies and access are assigned and inherited. Appears in modules 17, 31, 43.
Management system
A repeatable cycle of policy, planning, doing, checking and improving, with records proving each step. Appears in modules 12, 35, 39, 40, 41, 42, 43, 45.
MAS — Mobile Application Security
The OWASP project that maintains MASVS, MASWE and MASTG. Appears in module 35.
Masking —
When outliers inflate the mean and standard deviation so that they, and other outliers, no longer look unusual. Appears in module 32.
Mass assignment —
Automatic binding of request fields onto internal objects, which lets callers set fields they should not, such as a role or score. Appears in module 4.
Master data management — Master Data Management
The processes and tools that keep one trusted version of core shared data, such as worker records, and share it. Appears in module 45.
MASTG — Mobile Application Security Testing Guide
OWASP's manual for testing mobile apps against MASVS. Appears in modules 35, 39.
MASVS — Mobile Application Security Verification Standard
OWASP's standard of security and privacy requirements for mobile apps, organized into control groups such as STORAGE and PRIVACY. Appears in module 35.
MASWE — Mobile Application Security Weakness Enumeration
OWASP's catalog of specific mobile weaknesses that break MASVS requirements. Appears in module 35.
MAT — Modified Attack Requirements
A CVSS v4.0 environmental metric that overrides the base Attack Requirements for your own environment. Appears in module 6.
Matched pair
Two synthetic profiles identical except for one attribute, used to test whether that attribute changes a system's output. Appears in module 29.
Materiality
In U.S. securities law, information is material if there is a substantial likelihood a reasonable shareholder would consider it important. Appears in modules 42, 43.
MATS — ML Alignment and Theory Scholars
A research training program in AI alignment (details not verified from the authoring environment). Appears in module 38.
Maturity (ATLAS) —
ATLAS's rating of the evidence for a technique: Feasible, Demonstrated or Realized. Appears in module 10.
MAV — Modified Attack Vector
A CVSS v4.0 environmental metric that overrides the base Attack Vector for your own environment. Appears in module 6.
MB — Megabyte
About one million bytes of data. Appears in module 20.
MBA — Master of Business Administration
A graduate degree in business management. Appears in module 47.
MBR — Master Boot Record
The first sector of a disk, holding a small partition table and boot code. Appears in module 35.
MCDA — Multi-Criteria Decision Analysis
Methods for choosing between options scored on several criteria, such as weighted scoring or AHP. Appears in module 40.
MCMC — Markov Chain Monte Carlo
A family of simulation methods for drawing samples from a posterior distribution when it has no simple formula. Appears in module 21.
McNemar's test —
A test for paired yes/no outcomes that uses only the pairs where the two conditions disagree. Appears in modules 11, 13, 14, 16, 23, 24, 37, 38.
MCP — Model Context Protocol
An open protocol, built on JSON-RPC, that connects LLM applications (hosts) to servers offering resources, prompts and tools. Appears in modules 15, 16, 17, 99.
MCQ — Multiple-Choice Question
A question answered by picking from listed options. Appears in module 99.
MCSB — Microsoft Cloud Security Benchmark
Microsoft's set of security recommendations that Defender for Cloud uses to assess posture. Appears in module 31.
MD4 — Message Digest 4
An old, fast hash function; the NT hash is the MD4 hash of the password. Appears in module 3.
MD5 — Message-Digest Algorithm 5
An older hash function still used as a lookup key in reports; unsafe against deliberate collisions. Appears in modules 28, 30.
MDM — Mobile Device Management
Management of a whole mobile device: settings, passcode, apps and wipe. Appears in module 35.
measured boot —
A startup process in which each stage records a hash of the next in a TPM, creating a record for later checking. Appears in module 35.
Measurement invariance
The property that an instrument measures the same construct in the same way across groups. Appears in module 16.
mechanical combination —
Combining assessment evidence with a rule set in advance, such as a weighted sum, instead of holistic judgment. Appears in modules 40, 41, 47.
Mediator
A variable on the causal path from group membership to the outcome; controlling for it removes part of the effect being measured. Appears in module 21.
Membership inference —
Working out whether a particular record was in a model's training data. Appears in modules 8, 9, 10, 12, 13, 23, 34, 36, 38, 99.
Memorization
A model reproducing, or ranking far above similar alternatives, a specific sequence from its training data rather than a general pattern. Appears in modules 7, 8, 9, 16, 36, 37, 38.
Memory bandwidth
How many bytes per second a chip can read from its memory; usually the limit on how fast a model generates tokens. Appears in module 37.
Memory safety
A guarantee that every memory access stays within the right object (spatial) and only while it exists (temporal). Appears in modules 27, 34.
Meta-analysis
A statistical method that combines results from many studies to estimate an average effect and its variation. Appears in modules 25, 26, 39, 40, 41, 42, 43, 44, 46, 47.
Metadata —
Data about a communication rather than its content, such as addresses, sizes and timing, which encryption usually does not hide. Appears in module 0.
MFA — Multi-Factor Authentication
Signing in with two or more different kinds of proof, such as a password plus a phone prompt. Appears in modules 1, 4, 12, 17, 25, 31, 32, 33, 35, 43.
MFT — Master File Table
NTFS's central index, with at least one entry for every file on the volume. Appears in module 30.
MG — Manage
Prefix for MANAGE suggested actions in NIST AI 600-1. Appears in modules 12, 30.
MI — Incident Mitigation
CSF 2.0 category RS.MI: containing and eliminating an incident. Appears in modules 12, 20.
MIA — Membership Inference Attack
An attack that tries to decide whether a particular record was in a model's training data. Appears in module 38.
Micro emulation plan
A CTID plan type that emulates a compound behavior seen across many adversaries using a small number of techniques. Appears in module 29.
Microsegmentation —
Dividing a network into many small zones, each with its own gate. Appears in module 17.
Mid-p — mid-p value
An exact p-value that counts only half the probability of the observed table, making exact tests less conservative. Appears in modules 13, 21.
Minimum cut
The smallest set of graph edges whose removal disconnects one group of nodes from another. Appears in module 3.
Minimum detectable effect
The smallest true effect a study of a given size can reliably detect at the chosen error rates. Appears in modules 13, 16, 21, 22, 24.
Minutes per true positive —
Triage minutes per alert divided by precision: the analyst time each real finding costs. Appears in module 32.
Mirroring hypothesis
The idea, also called Conway's law, that a system's design copies the communication structure of the organization that builds it. Appears in modules 41, 42, 44.
MIS — Management Information Systems
The study and practice of information systems in organizations, as in the journal MIS Quarterly. Appears in modules 41, 43, 44.
MISP — MISP threat intelligence sharing platform (originally Malware Information Sharing Platform)
An open-source platform for storing, correlating and sharing threat information as events, attributes and objects. Appears in module 33.
Misprision of felony
A U.S. federal crime (18 U.S.C. 4) of concealing a known felony and not reporting it. Appears in module 43.
mission critical —
In oversight cases, a risk central to a company's business, for which courts expect the board to have a reporting system. Appears in module 47.
Misuse resistance —
Design of an API or construction so that common mistakes, such as nonce reuse, fail safely or are impossible. Appears in module 34.
MIT — Massachusetts Institute of Technology
A U.S. research university; MIT Press is its academic publisher. Appears in module 9.
MIT — MIT License
A short permissive open-source software license (named after the Massachusetts Institute of Technology). Appears in modules 3, 24, 30, 32, 35, 36, 37, 38.
MITRE — MITRE Corporation
A U.S. not-for-profit that maintains ATT&CK, ATLAS, CVE and CWE. Appears in modules 1, 2, 3, 4, 5, 7, 8, 9, 10, 11, 12, 14, 15, 17, 18, 20, 23, 24, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 39, 46, 99.
ML — Machine Learning
Building systems that learn patterns from data rather than following hand-written rules. Appears in modules 0, 7, 8, 9, 10, 12, 14, 17, 18, 19, 20, 23, 26, 27, 28, 31, 37, 38, 39, 44, 45, 99.
ML-BOM — Machine Learning Bill of Materials
A bill of materials that lists models and datasets as components, with details such as a model card. Appears in modules 8, 18, 20, 28, 37.
ML-DSA — Module-Lattice-Based Digital Signature Algorithm
NIST's main post-quantum digital signature standard, FIPS 204 (2024). Appears in modules 0, 34.
ML-KEM — Module-Lattice-Based Key-Encapsulation Mechanism
NIST's post-quantum key-encapsulation standard, FIPS 203 (2024). Appears in modules 0, 34.
MLaaS — Machine Learning as a Service
Cloud services that train or host models and let customers query them through an API. Appears in module 9.
MLOps — Machine Learning Operations
The practices and tools for building, deploying, monitoring and updating ML models in production. Appears in modules 8, 18.
MLP — Multilayer Perceptron
A basic neural network made of layers of weighted sums followed by simple non-linear functions. Appears in module 9.
MNIST — Modified National Institute of Standards and Technology (dataset)
A classic dataset of handwritten digit images, built from NIST data, used as a simple benchmark. Appears in module 9.
Mobile threat defense — Mobile Threat Defense
Software that detects threats on mobile devices, such as malicious apps or risky networks. Appears in module 35.
Modbus —
A simple industrial request-response protocol for reading and writing numbered registers and coils; in common deployments it carries no user authentication. Appears in module 35.
Model card
A short document that travels with a model and describes its intended use, evaluation results across groups and conditions, and limits. Appears in modules 8, 18, 20, 21, 23, 29, 37, 38.
Model extraction —
Copying a model's parameters or behavior by querying it and studying the answers. Appears in modules 8, 9, 10, 12, 15, 18, 23, 36, 38.
Model organism —
A deliberately built model used to study a failure mode, such as a backdoored model. Appears in module 38.
Model registry
A store of immutable model versions with their metadata, where a movable label such as 'production' points at the version in use. Appears in modules 5, 17, 18, 28, 30, 37.
Model risk
The potential for harm from decisions based on models that are wrong or used outside their intended purpose. Appears in modules 8, 26, 45, 46.
Model scanner
A tool that inspects model files for dangerous constructs without loading them. Appears in modules 18, 27.
Modified z-score —
0.6745 times the distance from the median divided by the MAD; a robust version of the z-score. Appears in module 32.
Monte Carlo simulation —
Estimating an outcome's distribution by drawing many random samples of uncertain inputs and computing the result for each. Appears in modules 12, 40, 46.
Mosca's inequality —
If data's required secrecy time plus migration time exceeds the time until a quantum computer, the data is at risk. Appears in module 34.
MP — Media Protection / Map
In SP 800-53, the media protection control family; in NIST AI 600-1 action IDs, the prefix for MAP actions. Appears in modules 12, 34.
MPC — Secure multi-party computation
Protocols that let several parties compute a function of their private inputs while revealing only the output. Appears in module 34.
MPR — Modified Privileges Required
A CVSS v4.0 environmental metric that overrides the base Privileges Required for your own environment. Appears in module 6.
MQV — Menezes-Qu-Vanstone
A family of authenticated Diffie-Hellman key-agreement schemes in NIST SP 800-56A. Appears in module 34.
MRM — Model Risk Management
The practice of governing models through sound development, independent validation and controls, from banking supervision. Appears in module 45.
MRTR — Multi Round-Trip Requests
The MCP 2026-07-28 pattern in which a server asks for more input by returning an 'input required' result that the client answers on a retry. Appears in module 99.
MS — Measure
Prefix for MEASURE suggested actions in NIST AI 600-1, such as MS-2.11-xxx. Appears in modules 3, 12.
MS-DOS — Microsoft Disk Operating System
Microsoft's early command-line operating system; every PE file still begins with a small MS-DOS stub. Appears in module 28.
MSA — Master Services Agreement
An overarching contract setting general terms such as liability and confidentiality. Appears in modules 6, 26.
MSC — Modified Subsequent System Confidentiality
A CVSS v4.0 environmental metric that overrides the confidentiality impact on downstream systems for your own environment. Appears in module 6.
MSI — Modified Subsequent System Integrity
A CVSS v4.0 environmental metric that overrides the integrity impact on downstream systems for your own environment. Appears in module 6.
MSL — Memory-Safe Language
A programming language that enforces bounds and lifetime checks by default, such as Rust, Go, Java, C# or Python. Appears in module 27.
MSRPC — Microsoft Remote Procedure Call
Microsoft's version of remote procedure call, which lets a program run functions on another Windows machine. Appears in module 3.
MSSP — Managed security service provider
An outside company contracted to run security services such as 24-hour monitoring. Appears in module 42.
MSTG — Mobile Security Testing Guide
OWASP's manual for testing the security of mobile apps, now part of the OWASP MAS project. Appears in module 35.
MT — Multi-Turn
In MT-Bench, the benchmark's use of multi-turn conversations to test chat assistants. Appears in module 16.
MTD — Maximum Tolerable Downtime
The total time leaders will accept a business process being disrupted. Appears in module 12.
mTLS — Mutual TLS
A form of TLS in which both sides, not just the server, prove their identity with certificates. Appears in module 17.
MTTR — Mean Time to Restore
The average time from the start of an incident to service being restored. Appears in module 40.
MUI — Modified User Interaction
A CVSS v4.0 environmental metric that overrides the base User Interaction for your own environment. Appears in module 6.
Multiple comparisons
Running many tests, which increases the chance of false alarms. Appears in modules 3, 4, 11, 13, 14, 21, 32, 37.
Multiplicative depth —
The longest chain of multiplications in a computation, which drives the cost of homomorphic encryption. Appears in module 34.
Multisource feedback
Ratings of a person from their manager, peers, direct reports and themselves; also called 360-degree feedback. Appears in modules 41, 47.
MVA — Modified Vulnerable System Availability
A CVSS v4.0 environmental metric that overrides the availability impact on the vulnerable system for your own environment. Appears in module 6.
MVC — Modified Vulnerable System Confidentiality
A CVSS v4.0 environmental metric that overrides the confidentiality impact on the vulnerable system for your own environment. Appears in module 6.
MVI — Modified Vulnerable System Integrity
A CVSS v4.0 environmental metric that overrides the integrity impact on the vulnerable system for your own environment. Appears in module 6.
MX — Mail Exchanger record
A DNS record that names the servers that receive email for a domain. Appears in modules 0, 2.

N

NACD — National Association of Corporate Directors
A U.S. membership organization for corporate directors that offers director education and a certification program. Appears in module 47.
Naive practice —
Repeating an activity you can already do; it stops producing improvement once the activity becomes automatic. Appears in module 39.
Namespace —
In Kubernetes, a scope for names and policies within a cluster; on its own not a security boundary. Appears in module 17.
NAPT — Network Address Port Translation
The common form of NAT that maps many internal addresses and ports onto one public address and different ports. Appears in module 0.
NAT — Network Address Translation
Rewriting IP addresses as packets pass through a router, typically so many private addresses can share one public address. Appears in module 0.
Native extension
A Python module compiled from C, C++, Rust or GPU code, outside Python's own memory safety. Appears in module 27.
NATO — North Atlantic Treaty Organization
A military alliance whose intelligence terms for probability and confidence FIRST recommends for CTI reports. Appears in module 33.
NBER — National Bureau of Economic Research
A private U.S. research organization that publishes working papers in economics. Appears in modules 26, 41.
NCME — National Council on Measurement in Education
One of the three organizations that publish the Standards for Educational and Psychological Testing. Appears in modules 13, 16.
NDA — Non-Disclosure Agreement
A contract restricting who may see confidential material, such as a full SOC 2 report. Appears in module 12.
NDSS — Network and Distributed System Security Symposium
A leading academic computer security conference run by the Internet Society. Appears in module 39.
Necessity — Necessity (data protection)
The test that processing is targeted and proportionate to its purpose, with no less intrusive way to achieve it. Appears in modules 12, 13, 19, 21, 22, 25, 32, 36.
NetworkPolicy —
A Kubernetes object that limits which pods may connect to which; it needs a network plugin that enforces it. Appears in module 17.
NeurIPS — Conference on Neural Information Processing Systems
A major machine learning research conference; its Datasets and Benchmarks Track publishes benchmarks. Appears in modules 9, 11, 37, 38.
newcomer adjustment — Newcomer adjustment
How well a new member of an organization gains role clarity, self-efficacy and social acceptance during socialization. Appears in modules 44, 47.
NF4 — NormalFloat 4
A 4-bit number format whose 16 levels sit where a normal distribution's quantiles fall, matching how model weights are usually spread. Appears in module 37.
NICE — National Initiative for Cybersecurity Education
The NIST-led initiative whose framework describes cybersecurity work roles, competencies, tasks, knowledge and skills. Appears in modules 26, 35, 36, 40, 41, 42, 44.
NIS 2 — NIS 2 Directive
Directive (EU) 2022/2555, the EU law requiring cybersecurity risk management and incident reporting from essential and important entities. Appears in module 43.
NIS2 — NIS 2 Directive
Directive (EU) 2022/2555, the EU law requiring cybersecurity risk management and incident reporting from essential and important entities. Appears in module 43.
NIST — National Institute of Standards and Technology
U.S. agency that publishes widely used security guidance such as SP 800-115. Appears in modules 0, 1, 2, 3, 5, 6, 8, 9, 10, 12, 13, 14, 17, 18, 19, 20, 22, 23, 24, 25, 26, 28, 29, 30, 31, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 99.
NIST AI 100-2 E2025 — NIST Adversarial Machine Learning: A Taxonomy and Terminology of Attacks and Mitigations, 2025 edition
NIST's shared vocabulary for classifying attacks on AI systems by goal, capability, knowledge and life-cycle stage. Appears in modules 9, 10, 11.
NIST IR — NIST Interagency or Internal Report
A NIST report series (NISTIR) for research results, guidance and other technical work outside the Special Publication series. Appears in modules 30, 42, 43, 46.
NISTAML — NIST adversarial machine learning taxonomy label
The prefix NIST AI 100-2 uses for identifiers in its taxonomy, such as NISTAML.01 for availability breakdown. Appears in module 9.
NISTIR — NIST Interagency or Internal Report
A NIST report series; NISTIR 8280 is the 2019 study of demographic effects in face recognition. Appears in modules 25, 34, 35, 36, 40.
NLP — Natural Language Processing
The field that builds computer methods for working with human language. Appears in modules 7, 8, 16.
NLRA — National Labor Relations Act
U.S. law protecting employees' rights to organize and to engage in concerted activities for mutual aid or protection (Section 7). Appears in module 25.
NLRB — National Labor Relations Board
The U.S. agency that enforces the National Labor Relations Act. Appears in module 25.
No-regret move
An action that is worthwhile in every scenario considered. Appears in module 42.
NOA — Notice of Adoption
An agency notice that publishes a final rule; DCWP's NOA adopted the LL144 rules. Appears in module 22.
Non-inferiority test
A test of whether a new version is no worse than the old one by more than a margin chosen in advance. Appears in module 37.
Non-repudiation — Non-repudiation (LINDDUN)
A LINDDUN threat: a person cannot deny an action when deniability would protect them. Appears in modules 1, 36.
Nonce —
A number used once: a per-encryption value that must never repeat under the same key. Appears in modules 0, 34.
Norm
A measure of a vector's size, used to state how big an input change is allowed to be. Appears in modules 9, 38.
Normalization —
Mapping fields from many data sources into one agreed set of names and meanings. Appears in module 32.
NPV — Net Present Value
The sum of a decision's cash flows, each converted to today's value at a discount rate; positive means value is created on those assumptions. Appears in modules 40, 42.
NSA — National Security Agency
The U.S. agency that co-authored the Kubernetes Hardening Guidance with CISA. Appears in modules 17, 27.
NSE — Nmap Scripting Engine
Nmap's system for running scripts against discovered services. Appears in module 2.
NT — NT hash
The hash of a Windows password used by NTLM; the name comes from Windows NT. Appears in module 3.
NTFS — New Technology File System
The default file system on Windows. Appears in module 30.
NTLM — NT LAN Manager
Microsoft's older challenge-response authentication protocol, now deprecated. Appears in modules 3, 4, 5.
Null hypothesis — Null hypothesis (H0)
The assumption, in a significance test, that there is no true difference. Appears in modules 13, 16, 21.
NVD — National Vulnerability Database
A U.S. public database of known vulnerabilities. Appears in modules 2, 6, 20.
NX — No-eXecute (also DEP, Data Execution Prevention)
Marking memory that holds data as not runnable, so injected bytes cannot run as code. Appears in modules 5, 27.
NYC — New York City
Used here for NYC Local Law 144 on automated employment decision tools. Appears in modules 4, 12, 13, 14, 18, 21, 22, 23, 24, 25, 26, 34, 37, 40, 44, 45.
NYCRR — New York Codes, Rules and Regulations
The official compilation of New York State regulations; Title 23, Part 500 is the NYDFS cybersecurity regulation. Appears in module 43.
NYDFS — New York Department of Financial Services
The New York regulator of banks, insurers and other financial firms; it issues the cybersecurity regulation 23 NYCRR Part 500. Appears in module 43.

O

O-RA — Open Group Risk Analysis
The Open Group standard that describes how to perform a FAIR risk analysis. Appears in module 12.
O-RT — Open Group Risk Taxonomy
The Open Group standard that defines FAIR's risk factors. Appears in module 12.
O*NET — Occupational Information Network
The U.S. Department of Labor's database of occupations, their tasks, knowledge, skills and abilities. Appears in modules 26, 40, 41, 44, 45, 47.
OAEP — Optimal Asymmetric Encryption Padding
The padding scheme that makes RSA encryption safe against chosen-ciphertext attacks; RSA without it is CWE-780. Appears in module 34.
OASIS — Organization for the Advancement of Structured Information Standards
A standards consortium that publishes SAML and other open standards. Appears in modules 17, 25, 28.
OAuth — OAuth 2.0
A framework for granting applications limited access using tokens. Appears in modules 0, 4, 5, 8, 15, 16, 17, 25, 31, 32, 99.
Obfuscation
Transforming code or data so it is harder to read without changing what it does. Appears in modules 8, 9, 10, 11, 28, 35, 38.
Observation period
The span of time, often three to twelve months, over which a Type 2 report tests controls. Appears in module 12.
OC — Organizational Context
CSF 2.0 category GV.OC: understanding the mission, stakeholders and legal duties that shape risk decisions. Appears in modules 12, 26, 42, 43.
OCB — Organizational Citizenship Behavior
Discretionary behavior that helps colleagues or the organization beyond formal job requirements. Appears in module 25.
OCC — Office of the Comptroller of the Currency
The US federal regulator of national banks. Appears in modules 45, 46.
OCM — Organizational Change Management
The structured effort to help people move from current ways of working to new ones by building understanding, ability and commitment. Appears in module 40.
OCSF — Open Cybersecurity Schema Framework
An open, vendor-neutral schema for security events, organized into categories and event classes with numeric identifiers. Appears in module 32.
OCTAVE — Operationally Critical Threat, Asset, and Vulnerability Evaluation
A risk-based security assessment method from Carnegie Mellon's Software Engineering Institute. Appears in modules 1, 36.
OECD — Organisation for Economic Co-operation and Development
An international organization whose AI work includes an AI incidents monitor and AI principles. Appears in module 20.
OEWS — Occupational Employment and Wage Statistics
The U.S. Bureau of Labor Statistics program that publishes employment and wage estimates by occupation. Appears in module 26.
OFAC — Office of Foreign Assets Control
The U.S. Treasury office that administers and enforces economic sanctions. Appears in module 43.
OG-WRL-008 — Privacy Compliance work role
The NICE Framework work role responsible for developing and overseeing an organization's privacy compliance program. Appears in modules 26, 36.
OIDC — OpenID Connect
An identity layer on top of OAuth 2.0 that tells an application who signed in, using a signed ID Token. Appears in modules 0, 17, 18, 25, 31.
OJ — Official Journal of the European Union
The EU's official publication in which regulations such as the AI Act appear. Appears in modules 12, 19, 24.
OKR — Objectives and Key Results
A goal format: a qualitative objective plus a few measurable key results for a period, usually a quarter. Appears in modules 41, 42.
OLA — Operating Level Agreement
A service-level agreement between internal teams. Appears in module 40.
OMB — Office of Management and Budget
The U.S. executive office that issues binding guidance to federal agencies, such as memorandum M-26-15 on PQC migration. Appears in modules 34, 45, 46.
ONNX — Open Neural Network Exchange
An open file format for machine learning models that stores the model as a graph, not as executable Python objects. Appears in modules 18, 27.
OOH — Occupational Outlook Handbook
The BLS guide that summarizes pay, outlook and duties for occupations. Appears in module 26.
OOPSLA — Object-Oriented Programming, Systems, Languages and Applications
A research conference on programming languages and software engineering. Appears in module 40.
OPA — Open Policy Agent
An open-source, general-purpose policy engine whose rules are written in the Rego language. Appears in module 31.
Open Badges —
A 1EdTech standard for digital credentials; version 3.0 builds on W3C Verifiable Credentials and carries a cryptographic proof. Appears in module 0.
Open-weight model
A model whose trained weights are published so others can run it themselves. Appears in modules 18, 22, 24, 38, 44.
OpenAPI —
A machine-readable format that describes an API's endpoints, parameters and responses. Appears in module 4.
OpenDP
An open-source differential privacy library written in Rust with Python and R bindings; version 0.16.0 as checked on 2026-11-08. Appears in module 36.
OpenSSF — Open Source Security Foundation
A Linux Foundation project that funds and coordinates work to make open-source software more secure. Appears in module 18.
Operating model
How a function delivers: structure, processes, decision rights, locations, sourcing and funding. Appears in modules 36, 39, 42, 44, 45.
Operating rhythm
The recurring meetings, reviews and reports that keep a multi-team organization coordinated. Appears in modules 41, 42.
Operational definition —
An observable rule that two raters could apply to decide whether an outcome counts. Appears in modules 11, 44.
Operational resilience
The ability to keep delivering important services within a stated tolerance for disruption, recover them, and learn. Appears in modules 43, 44, 46.
OpEx — Operating Expenditure
Spending recorded as an expense in the period it occurs, such as a subscription. Appears in modules 40, 42.
Orchestrator
The application code that runs an agent loop, validates and executes tool calls. Appears in modules 7, 8, 15.
Order of volatility —
The principle of collecting the most short-lived evidence first, from registers and memory down to disks and archives (RFC 3227). Appears in module 30.
Ordinal scale
A scale whose numbers show order but not distance, so multiplying them is not meaningful. Appears in modules 1, 6, 46.
Organization design
Deliberate choices about structure, roles, processes and decision rights. Appears in modules 41, 42, 44.
Organization Policy —
The Google Cloud service that restricts how resources may be configured, regardless of which principal asks. Appears in module 31.
Organizational justice —
Employees' perceptions of fairness at work, with four facets: distributive (outcomes), procedural (processes), interpersonal (respect) and informational (explanations). Appears in modules 25, 33.
ORM — Object-relational mapper
A library that maps program objects to database tables and builds queries. Appears in module 4.
OS — Operating System
The core software that runs a computer, such as Linux or Windows. Appears in modules 0, 2, 4, 5, 15, 27, 28, 29, 30, 34, 35.
OSCAL — Open Security Controls Assessment Language
NIST's machine-readable formats for security control catalogs and assessments, used to publish SP 800-53 in JSON. Appears in module 25.
OSCP — OffSec Certified Professional
OffSec's penetration testing certification earned through a proctored practical exam after PEN-200. Appears in modules 23, 24, 99.
OSCP+ — OffSec Certified Professional Plus
The time-limited version of OSCP awarded on passing the same exam; OffSec states it expires three years from issuance. Appears in modules 23, 24, 39, 99.
OSED — OffSec Exploit Developer
An OffSec exploit development certification. Appears in modules 27, 39, 99.
OSEE — OffSec Exploitation Expert
OffSec's most advanced exploitation certification. Appears in module 99.
OSEP — OffSec Experienced Penetration Tester
An OffSec certification that can be used to maintain OSCP+. Appears in modules 39, 99.
OSF — Open Science Framework
An open platform used, among other things, to register study plans before data collection (not verified from the authoring environment). Appears in module 38.
OSI — Open Systems Interconnection
A seven-layer reference model for describing network communication, from the physical layer up to the application layer. Appears in module 0.
OSINT — Open-source intelligence
Information gathered from publicly available sources. Appears in modules 2, 24, 33, 99.
OSS — Open-Source Software
Software whose source code is published under a license that lets anyone read, change and share it. Appears in module 27.
OSS-Fuzz
Google's free continuous fuzzing service for open-source projects. Appears in module 27.
OSSTMM — Open Source Security Testing Methodology Manual
ISECOM's methodology for measuring operational security. Appears in module 1.
OSV — Open Source Vulnerabilities
An open vulnerability database for open-source packages, served at OSV.dev. Appears in module 18.
OSWA — OffSec Web Assessor
An OffSec web application certification. Appears in module 99.
OT — Operational Technology
Hardware and software that monitors or controls physical equipment and processes. Appears in modules 6, 35, 39, 40.
OTel — OpenTelemetry
An open, vendor-neutral standard and toolkit for traces, metrics and logs. Appears in modules 20, 32.
OTRF — Open Threat Research
A community project that publishes free, recorded security event datasets (Security Datasets) for practicing detection. Appears in modules 3, 32.
OU — Organizational Unit
A grouping of accounts in AWS Organizations (or of objects in a directory) used to apply policy. Appears in modules 3, 17, 31.
Outcome bias
Judging a decision by how it turned out rather than by what was known when it was made. Appears in modules 43, 47.
Output privacy —
Limiting what a computation's released result reveals about individuals. Appears in module 34.
Outside view —
Forecasting by comparing a plan with the actual outcomes of a reference class of similar past projects. Appears in modules 40, 42, 46.
OV — Oversight
CSF 2.0 category GV.OV: reviewing and adjusting the risk strategy. Appears in modules 12, 26, 43, 47.
Overall assessment rating
The single combined score from an assessment center. Appears in modules 41, 47.
Overdispersion —
Count data varying more than a Poisson model predicts, which makes Poisson tail probabilities too small. Appears in module 32.
Overfitting
Fitting the training data so closely that performance on new data gets worse; on personal data it often shows up as memorization. Appears in modules 9, 18, 37.
OWASP — Open Worldwide Application Security Project
A non-profit community that publishes free application security guidance. Appears in modules 0, 1, 4, 5, 6, 7, 8, 9, 12, 14, 15, 16, 17, 18, 19, 20, 22, 23, 24, 25, 26, 28, 29, 31, 32, 34, 35, 38, 39, 43, 99.

P

p-value
The probability of a result at least as extreme as the one observed, assuming there is no true difference. Appears in modules 8, 11, 13, 14, 21, 24, 32.
P&L — Profit and loss statement
The financial statement showing revenue, costs and profit over a period; also called the income statement. Appears in module 42.
PA — Policy Administrator
The SP 800-207 component that opens or shuts the communication path between subject and resource. Appears in modules 3, 17.
PaaS — Platform as a Service
A cloud model where the provider runs the platform and the customer deploys code and configures it. Appears in module 17.
PAC — Privilege Attribute Certificate
Microsoft's authorization data, such as group memberships, carried inside Kerberos tickets. Appears in module 3.
Packing
Compressing or encrypting a program and wrapping it in a small stub that restores it in memory when it runs. Appears in module 28.
PACS — Physical Access Control System
A system that decides who may pass which door or gate, and when, and logs every attempt. Appears in module 35.
PAD — Presentation Attack Detection
Technology that tries to tell a live person at the sensor from an artifact such as a photo, mask or screen; often called liveness detection. Appears in module 25.
Padding oracle —
Any system behavior, such as an error or a timing difference, that reveals whether decrypted padding was valid. Appears in module 34.
PagedAttention
A way of managing KV cache memory in fixed-size blocks allocated on demand, like virtual memory, introduced with vLLM. Appears in module 37.
Paid placement
Content that appears editorial but was paid for; this program does not use it. Appears in module 24.
PAIR — Prompt Automatic Iterative Refinement
An automated jailbreak method in which an attacker model repeatedly rewrites a prompt based on the target's replies. Appears in modules 8, 14, 38.
Paired testing
Comparing a tool's outputs on inputs that differ only in a group marker. Appears in modules 11, 13, 14, 24, 29.
pandas
The standard Python library for working with tables of data. Appears in modules 3, 7.
Parameterized query — Parameterized query (prepared statement)
A query whose structure is fixed before user data is supplied, so the data cannot change it. Appears in modules 4, 8.
PASTA — PASTA (expansion not verified)
A risk-centered threat-modeling method that links business objectives to application threats. Appears in modules 1, 36.
Payload
The action or code that runs after an exploit succeeds. Appears in modules 5, 14, 27, 28.
Payroll diversion —
Redirecting an employee's salary by changing their direct-deposit details, usually after account takeover or impersonation. Appears in module 33.
PBKDF2 — Password-Based Key Derivation Function 2
A standard way to turn a password into a cryptographic key by hashing it many times. Appears in modules 0, 3.
PCR — Platform Configuration Register
A TPM register that can only be extended, holding a running hash of boot measurements. Appears in module 35.
PDF — Portable Document Format
A common file format for documents; often uploaded as résumés. Appears in modules 0, 1, 5, 6, 8, 9, 12, 14, 17, 18, 19, 22, 24, 25, 26, 27, 28, 30, 34, 40, 43, 44, 47.
PDFs — Portable Document Format files
Plural of PDF. Appears in modules 12, 14, 15, 21, 22, 24, 36.
PDP — Policy Decision Point
The logic that decides access; in SP 800-207, the policy engine plus the policy administrator. Appears in module 17.
PE — Policy Engine
The SP 800-207 component that makes and logs the decision to grant, deny or revoke access. Appears in modules 17, 28.
PEAK — Prepare, Execute, and Act with Knowledge
A threat-hunting framework from Splunk's SURGe team with three hunt types: hypothesis-driven, baseline and model-assisted. Appears in module 32.
PEC — Privacy-Enhancing Cryptography
NIST's program tracking cryptographic techniques such as MPC, FHE and zero-knowledge proofs. Appears in module 34.
Peer group —
A set of entities expected to behave alike, used as a comparison for each member. Appears in module 32.
PEFT — Parameter-Efficient Fine-Tuning
Fine-tuning methods that adjust only a small number of parameters. Appears in modules 7, 8, 37.
PEM — Privacy-Enhanced Mail (format)
A text encoding for certificates and keys, with lines such as BEGIN CERTIFICATE around base64 data. Appears in module 0.
PEN-200 — Penetration Testing with Kali Linux
OffSec's course that prepares learners for the OSCP+ exam. Appears in modules 23, 24, 99.
Penetration test
A time-boxed assessment that tries to find and demonstrate exploitable weaknesses in a defined scope. Appears in modules 1, 2, 6, 11, 14, 18, 25, 29, 30, 33, 40, 41, 43, 99.
PenTest+ — CompTIA PenTest+
A CompTIA penetration-testing certification with multiple-choice and performance-based questions; current exam PT0-003. Appears in modules 39, 99.
PEP — Policy Enforcement Point
The SP 800-207 component that enables, monitors and ends connections to a resource. Appears in module 17.
Percent agreement —
The share of cases where two raters gave the same label; it ignores agreement expected by chance. Appears in modules 13, 16, 40.
Performance-based question
An exam item that asks the candidate to perform a task in a simulated environment. Appears in module 99.
Permission-aware retrieval
Searching only among chunks the requesting user is allowed to see. Appears in module 8.
Permissions boundary
An AWS setting that caps the maximum permissions an identity policy can give a user or role. Appears in modules 17, 31.
Perplexity
How surprising a text is to a language model; gibberish has high perplexity. Appears in modules 8, 9, 20, 38.
Persistence
Techniques that keep an intruder's access across restarts and changes. Appears in modules 5, 10, 28, 33.
Personal data breach
Under the GDPR, a breach of security leading to accidental or unlawful destruction, loss, alteration, unauthorized disclosure of, or access to, personal data. Appears in modules 20, 36.
PERT — Program Evaluation and Review Technique
Here, a smooth distribution defined by a minimum, most likely and maximum value, used for expert estimates. Appears in module 12.
Perturbation budget —
The largest change an attacker is allowed to make to an input, measured in a chosen norm and written as epsilon. Appears in module 9.
PET — Parameter-Efficient Tuning
Another name used in some papers for parameter-efficient fine-tuning (PEFT). Appears in module 8.
Peter Principle
Promoting people on current job performance until they reach a role they are less suited to. Appears in module 41.
PF — Privacy Framework
Short for the NIST Privacy Framework, a voluntary tool for managing privacy risk. Appears in module 36.
PGD — Projected Gradient Descent
A many-step attack that takes small FGSM-like steps and pulls the input back into the allowed range after each one. Appears in modules 9, 10, 38.
PHI — Protected Health Information
Health information regulated by the HIPAA Privacy Rule. It excludes employment records held by a covered entity in its role as employer. Appears in module 36.
Phishing —
A message that impersonates a trusted sender to trick someone into revealing credentials, paying money or running malware. Appears in modules 0, 1, 2, 3, 4, 6, 8, 10, 12, 17, 25, 28, 29, 33, 35, 41, 42, 43, 46.
PIA — Privacy Impact Assessment
A general term for assessing a system's privacy risks before and during its use; the GDPR version is the DPIA. Appears in module 36.
Pickle — Python pickle
A Python serialization format that can execute code when loaded. Appears in modules 5, 8, 18, 27, 28.
PID — Process Identifier
The number an operating system assigns to a running process. Appears in module 30.
PIE — Position-Independent Executable
A program compiled so its own code can load at any address, letting ASLR randomize it too. Appears in module 27.
PII — Personally Identifiable Information
Information that identifies a person directly or in combination. Appears in modules 8, 11, 12, 14, 15, 16, 25, 29, 33, 37.
PIM — Privileged Identity Management
The Microsoft Entra service that makes people eligible for privileged roles and requires time-limited, approved activation. Appears in module 31.
PIN — Personal Identification Number
A short numeric secret used to unlock a device or confirm identity. Appears in module 35.
PipelineDP
A Python differential privacy library associated with Google's differential privacy project, published on PyPI as pipeline-dp. Appears in module 36.
PIR — Priority Intelligence Requirement
One of a small number of leadership-ranked questions that intelligence must answer, each tied to a single decision. Appears in module 33.
Pivoting —
Using one known fact, such as a domain, to discover related facts, such as other domains or victims. Appears in modules 30, 33.
PKCE — Proof Key for Code Exchange
An OAuth extension that binds an authorization code to the client that asked for it. Appears in modules 4, 17, 25.
PKCS — Public-Key Cryptography Standards
A family of standards for public-key cryptography, for example key file formats. Appears in modules 18, 34.
PKCS #11 — Public-Key Cryptography Standards #11
A standard interface for using cryptographic hardware such as hardware security modules and smart cards. Appears in module 18.
PKI — Public Key Infrastructure
The certificates, authorities and rules used to issue and trust public keys. Appears in modules 0, 34.
PKINIT — Public Key Cryptography for Initial Authentication in Kerberos
A Kerberos extension that lets users sign in with a certificate instead of a password. Appears in module 3.
Plaintext —
Readable, unencrypted data, such as data before encryption or after decryption. Appears in modules 0, 3, 34.
Planning fallacy
The tendency to underestimate how long one's own tasks will take. Appears in modules 40, 41.
platform — Platform
A shared set of internal services, tools and paved paths that product teams use instead of solving the same problem many times. Appears in module 44.
PLC — Programmable Logic Controller
A rugged industrial computer that repeatedly reads inputs, runs control logic and writes outputs to machines. Appears in module 35.
PLOS — Public Library of Science
A nonprofit publisher of open-access science journals. Appears in modules 32, 39.
Plugin (promptfoo) —
A promptfoo generator of adversarial inputs for one risk area. Appears in module 11.
PM — Program Management
SP 800-53 control family for organization-wide security program controls. Appears in modules 12, 25.
PMBOK — Project Management Body of Knowledge
PMI's guide and standard for project management, published in numbered editions. Appears in module 40.
PMI — Project Management Institute
A professional association that publishes the PMBOK Guide and awards the PMP certification. Appears in module 40.
PMP — Project Management Professional
A project-management certification; used in this module as an example of a credential a model might wrongly claim a candidate holds. Appears in modules 8, 40.
PNA — Public Network Access
The Microsoft Foundry setting that decides whether a project can be reached publicly or only through a private endpoint. Appears in module 31.
PNPT — Practical Network Penetration Tester
TCM Security's practical certification with a five-day test, two days for a report, and a live debrief. Appears in modules 23, 24, 39, 99.
PO — Policy
CSF 2.0 category GV.PO: the organization's cybersecurity policies. Appears in modules 12, 26, 43.
PO — Prepare the Organization
The SSDF practice group for people, processes and technology readiness. Appears in module 18.
Pod —
The smallest deployable unit in Kubernetes: one or more containers sharing network identity and storage. Appears in modules 17, 18.
Point of focus
An example characteristic an auditor may consider when judging a criterion; guidance, not a required control. Appears in module 12.
Policy as code —
Writing an organization's rules as code so they can be tested and run automatically. Appears in modules 4, 31.
Port —
A 16-bit number that identifies a program's endpoint on a host, such as 443 for HTTPS. Appears in modules 0, 1, 2, 3, 5, 11, 15, 17, 35.
Portable Executable — Portable Executable (PE)
The file format of Windows programs and libraries; it starts with an MS-DOS header and a 'PE' signature. Appears in module 28.
portfolio of proof —
A curated, dated set of work products, mapped to competencies and levels, that supports your ratings and career case. Appears in module 26.
Positive control —
A planted or emulated example of the behavior, used to prove a hunt or rule could have found it. Appears in module 32.
POSIX — Portable Operating System Interface
An IEEE standard set of operating system interfaces that Unix-like systems follow, so programs run across them. Appears in module 0.
POST — HTTP POST method
The web request type used to send data, such as a chat message, to a server. Appears in modules 0, 11, 25.
Post-hoc explanation —
A method applied after a model is trained to describe why it produced an output, without changing the model. Appears in module 16.
Posterior — Posterior distribution
In Bayesian analysis, the updated probability distribution for an unknown after combining a prior with the data. Appears in modules 2, 13, 21, 24.
Power — Statistical power
The probability that a test detects a true difference of a given size. Appears in modules 13, 21, 22, 37.
PPID — Parent Process Identifier
The PID of the process that started a given process. Appears in module 30.
PPL — Perplexity
A measure of how surprised a language model is by a text; unusually high values can flag odd inputs such as adversarial suffixes. Appears in module 8.
PPV — Positive predictive value
The share of alerts that are real problems. Appears in modules 2, 13, 25, 32.
PPVs — Positive Predictive Values
Plural of PPV: the share of flagged cases that are truly positive (precision). Appears in module 13.
PQ — Post-Quantum
Describes cryptography designed to stay secure even against a large quantum computer. Appears in module 34.
PQ/T — Post-quantum/traditional (hybrid)
A scheme that combines a post-quantum algorithm with a traditional one, secure if either holds (RFC 9794). Appears in module 34.
PQC — Post-quantum cryptography
Public-key cryptography designed to resist attacks by large quantum computers as well as classical ones. Appears in modules 34, 39.
PR — Privileges Required
CVSS metric for the account level the attacker needs first. Appears in module 6.
PR — Protect
The CSF 2.0 Protect function: safeguards to manage cybersecurity risk. Appears in modules 12, 20.
PRE — Pre-compromise (ATT&CK platform)
The ATT&CK platform label for attacker activity that happens before access, such as reconnaissance and resource development. Appears in module 10.
Pre-adverse action notice
The FCRA step of giving a person a copy of their consumer report and a summary of rights before acting against them based on it. Appears in module 36.
Pre-authentication — Kerberos pre-authentication
Proof, usually an encrypted timestamp, that a client knows its key before the KDC issues a TGT. Appears in module 3.
Pre-registration
Writing down estimands, thresholds and analysis plans before collecting data, to prevent choosing results after the fact. Appears in modules 13, 16, 21, 22, 24, 29, 38, 40.
Precision —
The share of alerts that are real attacks. Appears in modules 10, 20, 29, 32.
Precision–recall curve —
A plot of precision against recall across thresholds; more informative than the ROC curve when positives are rare. Appears in module 32.
Predictability
A NIST privacy engineering objective: people and operators can make reliable assumptions about data and its processing. Appears in module 36.
Predictive bias — Predictive bias (differential prediction)
When the relationship between scores and performance differs across groups. Appears in module 13.
Preference tuning
Training a model on judgments that one answer is better than another for the same prompt, as in RLHF and DPO. Appears in module 37.
Prefill
The first phase of generation, in which the whole prompt is processed in parallel. Appears in module 37.
Premortem
A group exercise that imagines a plan has already failed and lists the reasons, using prospective hindsight to surface risks. Appears in module 46.
presentation attack —
Fooling a biometric sensor with an artifact, such as a printed photo or fake fingerprint. Appears in modules 25, 35.
Preventive control —
A control that stops a non-compliant action from happening at all. Appears in modules 3, 5, 12, 31.
Primitive — Exploit primitive
A small capability gained from a bug, used as a building block in a larger exploit. Appears in modules 5, 27.
Privacy budget
The total privacy loss (such as a total epsilon) allowed across all releases from a dataset; each release spends some of it. Appears in modules 8, 9, 19, 36.
Privacy design strategies
Hoepman's eight design approaches: minimize, hide, separate, aggregate, inform, control, enforce and demonstrate. Appears in module 36.
Privacy engineering
Building systems so that the way they handle personal data does not cause problems for people; NIST treats it as a specialty of systems engineering. Appears in modules 0, 12, 19, 35, 36, 41, 45.
Privacy metric
A measure that tells whether a privacy program is working, such as the share of access requests answered on time. Appears in module 36.
Privacy unit
What one person's data means in a DP guarantee, such as one row or one employee; protecting a row is weaker when a person has many rows. Appears in module 36.
PrivateLink — AWS PrivateLink
AWS's private connectivity that lets a VPC reach a service through an interface endpoint without the public internet. Appears in module 31.
Privilege change —
Any change to what an account may do, such as a role assignment or a new API key. Appears in module 32.
Privilege creep —
The build-up of access rights as people move roles without old rights being removed. Appears in modules 25, 35.
Privilege escalation
Gaining more rights than you were granted, vertically (more power) or horizontally (another account's data). Appears in modules 0, 4, 5, 8, 10, 17, 23, 24, 28, 31, 33, 35, 99.
Privilege escalation path —
A chain of permitted actions by which a principal gains more privilege than it was directly granted. Appears in module 31.
PRM — Professional Risk Manager
A credential from PRMIA covering finance theory, instruments, mathematics and risk practice. Appears in module 46.
PRMIA — Professional Risk Managers' International Association
The association that awards the Professional Risk Manager credential. Appears in module 46.
PRNG — Pseudorandom number generator
An algorithm that expands a seed into random-looking numbers; only cryptographically secure ones are safe for keys and tokens. Appears in module 34.
Proactive control —
A control that checks resources before they are provisioned and blocks non-compliant ones. Appears in module 31.
Probe — Probe (garak)
A garak plugin that creates attack inputs for one weakness family. Appears in modules 11, 24.
Probe —
A simple classifier trained on a model's internal activations to test whether a property can be read from them. Appears in module 38.
Probe (garak) —
A garak plugin that creates attack inputs for one weakness family. Appears in module 11.
Problematic data action
A data action that could cause a problem for individuals, such as embarrassment, discrimination or loss of autonomy. Appears in module 36.
Procedure —
A specific, documented way a real group or tool carried out a technique. Appears in modules 10, 29.
Processing pipeline —
A set of pySigma transformations, such as field renaming, applied to a rule before a backend converts it. Appears in module 32.
Proctored exam
An exam supervised in person or remotely to prevent cheating. Appears in modules 23, 24, 27, 28, 30, 31, 33, 35, 39, 99.
product operating model — Product operating model
An arrangement in which long-lived teams own a product and its outcomes and are funded as ongoing capacity rather than as fixed-scope projects. Appears in module 44.
productivity J-curve — Productivity J-curve
The pattern in which measured productivity dips before rising after a general-purpose technology arrives, because intangible complements are costs before they are output. Appears in module 44.
Profile — AI RMF or CSF Profile
A tailored selection of framework outcomes describing an organization's current or target state. Appears in module 12.
Profiling — Profiling (GDPR sense)
Automated processing of personal data to evaluate personal aspects of a person, such as work performance. Appears in modules 12, 19, 36.
Prompt infection
Prompt injection that copies itself from one agent to another in a multi-agent system. Appears in module 15.
Prompt injection
Input that makes an LLM follow instructions the application designer did not intend; indirect when carried in processed content. Appears in modules 0, 1, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 18, 19, 20, 22, 23, 24, 26, 27, 28, 29, 30, 32, 33, 37, 38, 39, 43, 99.
promptfoo —
An open-source tool for evaluating and red teaming LLM applications, often used in CI. Appears in modules 11, 12, 14, 23, 24, 26, 29, 39.
Propensity evaluation —
A test of whether a model tends to use its capabilities in harmful ways under realistic conditions; also called an alignment evaluation. Appears in module 38.
Proper scoring rule —
A way of scoring forecasts in which you get the best expected score by reporting your true belief. Appears in module 33.
Proportionality —
The principle that monitoring must be no more intrusive than its legitimate purpose requires. Appears in modules 19, 32.
Protected Users — Protected Users security group
A built-in AD group whose members get fixed protections such as no NTLM, no RC4 and no delegation. Appears in module 3.
Provenance
A verifiable record of where an artifact came from: its inputs, the process that built it, and who ran it. Appears in modules 8, 17, 18.
Provider — Provider (EU AI Act)
The organization that develops an AI system, or has it developed, and puts it on the market under its own name. Appears in modules 12, 45.
Proxy feature
An input correlated with a protected characteristic, which can reproduce group differences even when the characteristic is excluded. Appears in modules 13, 23, 38.
Proxy interview —
An interview in which someone other than the applicant answers, by appearing in their place or by feeding answers in real time. Appears in module 25.
PS — Personnel Security
SP 800-53 control family on screening and managing staff with system access. Appears in modules 12, 25.
PS — Protect the Software
The SSDF practice group for protecting code and releases from tampering and unauthorized access. Appears in module 18.
PSA — Pod Security Admission
The built-in Kubernetes controller that applies the Pod Security Standards in enforce, audit or warn mode. Appears in module 17.
PSA — Public Service Announcement
A notice to the public from an agency, such as the FBI IC3's warnings about fraud schemes. Appears in module 25.
Pseudonymization
Replacing identifiers so data cannot be linked to a person without separately held extra information; the data remain personal data. Appears in modules 0, 19, 36, 37.
PSI — Population Stability Index
A number that measures how much a variable's distribution has shifted between a baseline and recent data. Appears in modules 18, 37.
PSS — Pod Security Standards
Kubernetes' three cumulative pod security levels: Privileged, Baseline and Restricted. Appears in module 17.
Psychological safety
A shared belief in a team that it is safe to take interpersonal risks, such as admitting mistakes. Appears in modules 41, 46.
PT — PII Processing and Transparency
SP 800-53 control family on lawful, transparent processing of personal data. Appears in module 12.
PTEF — Purple Team Exercise Framework
An open framework, maintained by SCYTHE and contributors, for planning, running and measuring purple team exercises. Appears in module 29.
PTES — Penetration Testing Execution Standard
A community penetration-testing methodology. Appears in modules 1, 6.
Purdue model —
A layered reference model for manufacturing systems, widely used to organize OT network segmentation from Level 0 (process) to Levels 4-5 (enterprise). Appears in module 35.
Purple team exercise
A full-knowledge exercise in which attackers and defenders collaborate technique by technique to test and improve detection and response. Appears in module 29.
purpose creep —
Using data collected for one purpose, such as safety, for another, such as performance evaluation. Appears in module 35.
PW — Produce Well-Secured Software
The SSDF practice group for design, coding, review and testing. Appears in module 18.
PWG — Public Working Group
An open NIST working group; the Generative AI Public Working Group (GAI PWG) informed NIST AI 600-1. Appears in module 12.
Pyodide
A version of Python compiled to run inside a web browser, with packages such as NumPy and scikit-learn. Appears in modules 9, 13, 14, 15, 16, 18, 19, 21, 23.
PyPI — Python Package Index
The public registry from which Python packages are installed with pip. Appears in modules 8, 11, 18, 27, 28, 30, 31, 32, 34, 36, 37, 38.
Pyramid of Pain —
David Bianco's ranking of indicator types by how costly they are for an adversary to change, from hash values up to TTPs. Appears in modules 28, 33.
PyRIT — Python Risk Identification Tool for generative AI
Microsoft's open-source framework and scanner for red teaming generative AI systems. Appears in modules 11, 14, 29, 39.
pySigma —
SigmaHQ's Python library that parses Sigma rules and converts them into query languages through backends and pipelines. Appears in module 32.
Python
A general-purpose programming language widely used for data analysis and machine learning; the course's notebooks are written in it. Appears in modules 5, 6, 7, 8, 9, 10, 11, 13, 15, 16, 18, 20, 21, 25, 27, 28, 29, 30, 31, 32, 34, 35, 36, 37, 38.

Q

Q&A — Question and Answer
A task or interface where a user asks a question and the system returns an answer. Appears in modules 7, 8, 15.
QA — Question Answering
A task in which a model answers questions, often about a given passage. Appears in module 37.
QLoRA — Quantized Low-Rank Adaptation
LoRA trained on top of a frozen base model stored in 4-bit numbers (NF4), so large models can be fine-tuned in much less memory. Appears in module 37.
Quality-of-service harm
Harm from a system working worse for some groups, such as misreading foreign degree names. Appears in module 29.
Quantization
Storing a model's numbers with fewer bits than they were trained in, trading some rounding error for less memory and faster generation. Appears in modules 24, 37.
Quasi-identifier
An attribute such as ZIP code or birth date that can identify a person when combined with others. Appears in modules 9, 19, 36.
QUIC —
A transport protocol built on UDP that adds reliability, multiplexed streams and built-in encryption; HTTP/3 runs over it. The name is not an acronym in its RFC. Appears in module 0.

R

RA — Risk Assessment
CSF 2.0 category ID.RA, and the SP 800-53 control family RA. Appears in module 12.
RACI — Responsible, Accountable, Consulted, Informed
A table that assigns roles for each task or decision. Appears in modules 12, 26, 40, 41, 42, 45.
RAF — Risk appetite framework
The whole system of policies, processes, roles and limits used to set, communicate and monitor how much risk an organization will take. Appears in module 46.
RAG — Retrieval-Augmented Generation
Giving a language model relevant documents fetched at question time so it can answer from them. Appears in modules 6, 7, 8, 10, 26, 99.
RAI — Responsible AI
Harms such as harmful, abusive or discriminatory outputs, as distinct from classic security impacts. Appears in modules 11, 29.
RAM — Random-Access Memory
A computer's working memory. Appears in modules 28, 30.
Randomized smoothing —
A defense that classifies many noisy copies of an input and takes a vote, which allows a proof of robustness within a radius. Appears in modules 9, 11, 38.
Range restriction
Reduced score spread in a selected group, which makes correlations look smaller than in the full applicant pool. Appears in modules 13, 14, 26, 44, 47.
Ransomware
Malware used for extortion, typically by encrypting the victim's data and demanding payment. Appears in modules 5, 6, 12, 28, 29, 33, 35, 43, 47.
RAPID — Recommend, Agree, Perform, Input, Decide (unverified)
A decision-rights model that names who recommends, who must agree, who carries out, who gives input and who decides. Appears in module 41.
RAPPOR — Randomized Aggregatable Privacy-Preserving Ordinal Response
Google's local differential privacy technique for collecting statistics from Chrome users, based on randomized response. Appears in module 36.
RAS — Risk appetite statement
The written statement at the core of a risk appetite framework, combining qualitative commitments and quantitative limits. Appears in module 46.
Rate limiting —
Capping how often one client may call an endpoint in a time window, often answered with HTTP status 429. Appears in modules 2, 4, 8, 18, 24, 34, 35.
rationalization — Rationalization
Reducing duplication and waste in the application and data estate by retiring, merging or replacing systems and fixing inconsistent data copies. Appears in module 44.
RBAC — Role-Based Access Control
Granting access by assigning roles (sets of permissions) to people or workloads. Appears in module 17.
RBCD — Resource-based constrained delegation
Delegation configured on the back-end resource, listing which front-end accounts may act for users. Appears in module 3.
RC — Recover
The CSF 2.0 Recover function: restoring assets and operations after an incident. Appears in modules 12, 20.
RC4 — Rivest Cipher 4
An old stream cipher; its Kerberos variant uses an unsalted password hash as the key and is being removed. Appears in modules 3, 4.
RCA — Root-Cause Analysis
Finding the earliest point where a program's state first became wrong, rather than where it crashed. Appears in module 27.
RCNY — Rules of the City of New York
The compiled rules of NYC agencies; DCWP's AEDT rules are in Title 6. Appears in modules 12, 24, 40.
RCP — Resource Control Policy
An AWS Organizations policy that caps the permissions available on resources in member accounts. Appears in modules 17, 31.
RDP — Remote Desktop Protocol
Microsoft's protocol for remote graphical logins. Appears in module 5.
Re-test
Running the same tests after a fix under a new baseline to see whether the pre-set criterion is met. Appears in modules 6, 7, 8, 12, 14, 17, 23, 24, 26, 27, 46.
Re-wrapping —
Decrypting data keys with an old key encryption key and encrypting them with a new one, without touching the data. Appears in module 34.
Reachable set —
Every identity a principal can turn into, directly or through a chain of permitted actions. Appears in modules 17, 31.
ReAct — Reasoning and Acting
A prompting pattern that interleaves reasoning steps and tool actions. Appears in modules 7, 8, 15.
Readiness for change
People's beliefs, attitudes and intentions about whether a change is needed and whether they can carry it out. Appears in module 42.
README — Read me file
The introductory document at the top of a software project that explains what it is and how to use it. Appears in modules 0, 2, 5, 7, 8, 10, 11, 15, 17, 18, 24, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 39, 40.
Recall —
The share of real attacks that a detection catches. Appears in modules 10, 20, 32.
Red team engagement
An objective-based exercise in which a team emulates a realistic adversary to test whether the organization can prevent, detect and respond. Appears in module 29.
Reference class forecasting
Forecasting a project from the actual outcomes of comparable past projects, to correct optimism bias. Appears in module 42.
Reference model
A frozen copy of the model, usually the SFT model, that preference tuning measures changes against so the trained model does not drift too far. Appears in module 37.
Rego —
The declarative policy language of Open Policy Agent. Appears in module 31.
Regression suite —
Fixed tests built from confirmed findings and re-run on every change. Appears in modules 8, 11, 14, 29, 37.
Regression test —
A test re-run after a change to show that what worked before still works. In detection engineering, recorded events that once triggered a rule are replayed to prove the rule still matches them. Appears in modules 11, 13, 27, 29, 32, 34, 37.
Regression to the mean
Extreme results tend to be followed by more average ones, because part of any result is noise. Appears in modules 26, 41.
Regulation FD — Regulation Fair Disclosure
SEC rule (17 CFR 243.100) requiring public disclosure when a company shares material nonpublic information with certain market professionals or holders. Appears in module 47.
rehosting —
Running firmware in an emulator instead of on its original hardware, so it can be analyzed. Appears in module 35.
Relationship conflict
Disagreement that is personal, about people rather than the work. Appears in module 41.
Reliability
How consistent scores are when a measurement is repeated under conditions that should not change the result. Appears in modules 14, 16, 47.
Remote attestation —
A hardware-signed statement from a trusted execution environment proving which code it is running. Appears in modules 31, 34.
REP — Reply
The response half of a Kerberos exchange, as in AS-REP or TGS-REP. Appears in module 3.
RepE — Representation Engineering
An approach that reads and steers population-level directions in a model's activations to monitor or change high-level behaviors. Appears in module 38.
Representational harm
Harm from how a group is portrayed, such as stereotyped language in candidate summaries. Appears in module 29.
Reproducibility
The ability to re-create a result; for ML it may be bit-for-bit, statistical (within measured noise) or documentary. Appears in modules 6, 7, 9, 10, 11, 14, 16, 21, 23, 24, 37, 38, 39.
REQ — Request
The request half of a Kerberos exchange, as in AS-REQ or TGS-REQ. Appears in module 3.
Reservation point
The worst deal you would accept in a negotiation. Appears in module 41.
Residual risk
Risk remaining after controls, which someone accountable must accept or reduce. Appears in modules 5, 7, 8, 11, 12, 15, 18, 19, 27, 31, 36, 37, 42, 43, 45, 46, 47.
Resource-based view
The theory that lasting advantage comes from resources that are valuable, rare, hard to imitate and hard to substitute. Appears in modules 42, 44.
REST — Representational State Transfer
A common style of web API in which clients send HTTP requests to URLs. Appears in modules 11, 29.
RESTful — Following Representational State Transfer
Describes an API designed in the REST style, using HTTP methods on resources. Appears in module 28.
Retention schedule
A documented rule for how long each category of data is kept and what happens at the end. Appears in modules 19, 30, 36, 44, 45.
Retired content —
Practice-platform content that no longer counts toward rankings or events; some platforms allow write-ups only for it. Appears in modules 23, 39.
Retired machine
A Hack The Box machine no longer in the active rotation; platforms generally allow write-ups only for retired content. Appears in modules 23, 39.
retranslation —
A check in which a second group of experts sorts anchors back to their competencies and levels; anchors they disagree on are rewritten or dropped. Appears in modules 26, 47.
Retrieval practice —
Recalling an idea from memory before checking it, as opposed to rereading. Appears in modules 0, 2, 9, 12, 23, 24, 39.
Reverse proxy —
A server in front of other servers that receives client requests for them, often to terminate TLS, balance load or filter requests. Appears in module 0.
Reverse stress testing
Starting from an unacceptable outcome and working back to the events that could cause it. Appears in module 46.
Revoked —
An ATT&CK item replaced by another; its old page redirects to the replacement. Appears in module 99.
Reward hacking —
Optimizing an imperfect proxy reward in a way that lowers performance on the true goal. Appears in modules 7, 38.
RFC — Request for Comments
The numbered document series that defines internet standards, such as RFC 9293 for TCP. Appears in modules 0, 2, 3, 4, 7, 8, 10, 15, 17, 22, 25, 30, 34, 39.
RFI — Request for Information
A public notice asking for information, without proposing a rule. Appears in modules 24, 40.
RFOA — Reasonable Factors Other Than Age
A defense under U.S. age-discrimination law: a practice that hurts older workers is lawful if based on a reasonable non-age factor. Appears in modules 21, 22.
RFP — Request for Proposal
A formal invitation for vendors to propose a solution and price against stated requirements. Appears in module 40.
RFQ — Request for Quotation
A request for vendors' prices against a precisely defined need. Appears in module 40.
RICE — Reach, Impact, Confidence, Effort
A product prioritization score: reach times impact times confidence, divided by effort. Appears in module 41.
RIMS — RIMS, the risk management society
A professional association for risk managers that awards the Certified Risk Management Professional credential. Appears in module 46.
RISC — Reduced Instruction Set Computer
A processor design that uses a small set of simple instructions, as ARM and RISC-V chips do. Appears in module 27.
Risk acceptance
A documented decision by an accountable person not to fix a finding. Appears in modules 6, 29, 41, 42, 43, 45, 47.
Risk appetite
The amount and type of risk an organization is willing to accept in pursuit of its objectives. Appears in modules 12, 26, 39, 40, 42, 43, 44, 45, 46, 47.
Risk culture
The shared assumptions, values and perceptions in an organization about risk and how people should act on it. Appears in module 46.
Risk limit
A hard, measurable boundary that keeps exposure inside a tolerance. Appears in module 46.
Risk matrix — Risk matrix (heat map)
A grid that maps ordinal likelihood and impact ratings to colored priority levels; easy to read but often misleading. Appears in modules 12, 46.
Risk register
A repository of risk information, including what is understood about each risk over time. Appears in modules 12, 15, 26, 41, 43, 44, 46.
Risk taxonomy
A shared, structured classification of risk events, so risks can be compared and added up without double counting. Appears in modules 12, 20, 46.
Risk tolerance
A measurable limit on the remaining risk an organization will bear to achieve an objective. Appears in modules 12, 36, 42, 43, 46.
RL — Reinforcement Learning
Training by trial and reward: the model's outputs are scored and it is updated to make high-scoring outputs more likely. Appears in modules 7, 8.
RLAIF — Reinforcement Learning from AI Feedback
Like RLHF, but with preferences generated by an AI guided by written principles. Appears in module 7.
RLHF — Reinforcement Learning from Human Feedback
Training a model to produce outputs people prefer, using human comparisons and a reward model. Appears in modules 7, 11, 29, 37, 38.
RM — Risk Management Strategy
CSF 2.0 category GV.RM: the organization's risk priorities and tolerances. Appears in modules 12, 26, 42, 43.
RMF — Risk Management Framework
Short for a structured approach to managing risk; in this module it refers to the NIST AI RMF. Appears in modules 12, 26, 36.
Robust statistic —
A statistic, such as the median, that a few extreme values do not move much. Appears in module 32.
ROC — Receiver Operating Characteristic
A curve that shows a detector's true positive rate against its false positive rate as the threshold changes. Appears in modules 9, 32, 38.
ROE — Rules of engagement
The operating rules for a security test: scope, times, contacts and incident handling. Appears in modules 1, 6, 8.
RoE — Rules of Engagement
The written limits of a test: targets, methods, data, timing, stop conditions and contacts. Appears in modules 11, 14, 29.
ROI — Return on Investment
A ratio of gain to cost; simpler than NPV and usually ignores timing. Appears in modules 29, 40, 45.
Rollback
Returning a system to a previous known-good version of everything that decides its outputs. Appears in modules 18, 37.
ROM — Read-Only Memory
Memory whose contents are fixed at manufacture; boot code stored there acts as a root of trust. Appears in module 35.
ROME — Rank-One Model Editing
A technique for editing specific factual associations in a model's weights. Appears in modules 8, 10.
Root cause
The underlying reason a class of weaknesses exists. Appears in modules 4, 5, 6, 8, 14, 15, 16, 22, 23, 24, 27, 30, 32, 36, 39, 40, 41.
Root key —
The top key in a key hierarchy, normally kept inside a hardware security module. Appears in modules 3, 13, 34, 35.
ROP — Return-Oriented Programming
A code-reuse technique that chains short existing instruction sequences ending in return instructions. Appears in module 27.
ROPA — Record of Processing Activities
The written record of processing that GDPR Article 30 requires, listing purposes, categories of people and data, recipients, transfers, erasure time limits and security measures. Appears in module 36.
ROPC — Resource Owner Password Credentials
An OAuth grant where the app collects the user's password; RFC 9700 says it must not be used. Appears in module 17.
RoT — Root of Trust
The component that provides basic security functions and must be trusted because nothing beneath it can check it. Appears in module 35.
RP — Incident Recovery Plan Execution
CSF 2.0 category RC.RP: carrying out recovery plans. Appears in modules 12, 20.
RP — Relying Party
In OIDC, the application that relies on the OpenID Provider's ID Token. Appears in module 17.
RPC — Remote Procedure Call
A way for one program to ask another to run a function and return the result. Appears in module 15.
RPO — Recovery Point Objective
The point in time to which data can be recovered after an outage, set by how recent the last backup is. Appears in modules 12, 46.
RR — Roles, Responsibilities, and Authorities
CSF 2.0 category GV.RR: who is accountable for cybersecurity risk. Appears in modules 12, 26, 42, 43, 47.
RS — Respond
The CSF 2.0 Respond function: actions taken about a detected incident. Appears in modules 12, 20, 30.
RSA — Rivest-Shamir-Adleman
A public-key algorithm used for digital signatures and, in older protocols, for key transport. Appears in modules 0, 30, 34.
RSP — Responsible Scaling Policy
Anthropic's published policy that ties capability thresholds to required safeguards; it has been revised many times. Appears in module 38.
RST — Reset
A TCP message refusing or ending a connection. Appears in module 2.
RSU — Restricted stock unit
A grant of company shares delivered once conditions, usually continued employment over a vesting schedule, are met. Appears in module 26.
RTO — Recovery Time Objective
The longest a system can stay unavailable before unacceptable impact; it must fit within the MTD. Appears in modules 12, 46.
RTOS — Real-Time Operating System
An operating system designed to respond within guaranteed time limits, common in control devices. Appears in module 35.
RTU — Remote Terminal Unit
A field device in industrial control systems that reads sensors and drives equipment for a central control system. Appears in module 35.
Rule of three —
After zero successes in n attempts, the 95% upper bound on the rate is about 3/n. Appears in modules 11, 12, 14, 16, 19, 30.
RV — Respond to Vulnerabilities
The SSDF practice group for finding, fixing and learning from vulnerabilities after release. Appears in module 18.

S

S256 — PKCE code challenge method S256
The PKCE method in which the client sends the SHA-256 hash of its secret code verifier, encoded in base64url. Appears in module 17.
S4U — Service for User
Microsoft Kerberos extensions that let a service obtain tickets on behalf of a user. Appears in module 3.
SA — Subsequent System Availability
CVSS v4.0 metric for availability impact on systems other than the vulnerable one. Appears in modules 6, 12, 35, 39.
SaaS — Software as a service
Software run by a vendor and used over the internet. Appears in modules 1, 2, 4, 5, 6, 10, 12, 14, 17, 29, 30, 31, 32, 34, 40, 42, 44.
SACL — System access control list
The part of an object's security settings that says which accesses are written to the audit log. Appears in modules 0, 3.
SAE — Sparse Autoencoder
A small network trained to rewrite a model's internal activations as a sparse mix of many learned directions, so that each direction is easier to interpret. Appears in module 38.
SAFe — Scaled Agile Framework
A commercial framework for running agile methods across many teams; it popularized WSJF. Appears in modules 40, 41.
Safe harbor
A client's agreement not to pursue claims against testers for authorized, good-faith testing. Appears in modules 1, 13, 14, 22, 36, 38, 39, 43.
Safetensors
A model weight format that stores only tensors and metadata, not executable code. Appears in modules 8, 18, 27, 28.
SAIF — Secure AI Framework
Google's framework for AI security, which HTB says its AI curriculum aligns with. Appears in module 99.
Salt —
A unique random value mixed into each password hash so that identical passwords give different stored values. Appears in modules 0, 3, 34.
SAML — Security Assertion Markup Language
An XML-based standard that lets an identity provider tell an application that a user signed in. Appears in modules 0, 17, 25, 34.
SAMM — Software Assurance Maturity Model
An OWASP model for measuring and improving how an organization builds secure software. Appears in modules 26, 43.
Sampling variability —
The natural, chance wobble in a rate or average from one sample to the next, even when nothing real has changed. Appears in module 13.
Sandbox
An environment that restricts what code can reach: files, network, processes and credentials. Appears in modules 1, 2, 8, 13, 15, 16, 17, 18, 22, 24, 25, 27, 28, 30, 31, 35.
SANS — SANS Institute
A security training company whose courses align with GIAC certifications. Appears in modules 27, 28, 33, 35, 39.
SAST — Static Application Security Testing
Analyzing source code for vulnerabilities without running it. Appears in modules 8, 18.
SaTML — Secure and Trustworthy Machine Learning
The IEEE Conference on Secure and Trustworthy Machine Learning, a venue for ML security research (details not verified from the authoring environment). Appears in modules 7, 8, 38, 39.
SB — Senate Bill
A bill introduced in a state Senate, such as Colorado SB24-205; the digits give the year and number. Appears in modules 12, 19.
SBOM — Software Bill of Materials
A list of the components inside a piece of software, used to find known vulnerabilities. Appears in modules 6, 8, 18, 34.
SBOMs — Software Bills of Materials
Plural of SBOM: machine-readable lists of the components in software. Appears in modules 6, 8, 18.
SC — Subsequent System Confidentiality
CVSS v4.0 impact on confidentiality of other affected systems. Appears in modules 6, 12, 26, 31, 39, 43.
SCA — Software Composition Analysis
Identifying third-party components and matching them against known vulnerabilities. Appears in module 18.
SCADA — Supervisory Control and Data Acquisition
Control systems that supervise assets spread over large distances, such as pipelines, grids and water networks, from a central control center. Appears in module 35.
SCC — Standard Contractual Clause
One of the model contract terms adopted by the European Commission to safeguard transfers of personal data outside the EEA. Appears in module 36.
SCCs — Standard Contractual Clauses
Model contract terms adopted by the European Commission (currently Decision 2021/914) that exporters and importers sign to safeguard transfers of personal data outside the EEA. Appears in module 36.
SCCT — Situational Crisis Communication Theory
Coombs's theory linking how much responsibility the public attributes for a crisis to the response strategies that best protect reputation. Appears in module 47.
Scenario planning
Preparing for an uncertain future by testing a strategy against several plausible, internally consistent futures. Appears in module 42.
SCHUFA
A German credit agency; also the short name of CJEU case C-634/21 on automated scoring (Module 19). Appears in modules 19, 36.
SCIM — System for Cross-domain Identity Management
An IETF standard (RFC 7643 and RFC 7644) for creating, updating and removing user accounts across systems over HTTP, usually driven from an HRIS through an identity provider. Appears in module 25.
SCO — STIX Cyber-observable Object
A STIX object for an observable fact such as a file, domain or IP address. Appears in module 28.
Scorer (PyRIT) —
A PyRIT component that judges whether a response met the attack objective. Appears in module 11.
Scoring rate — Scoring rate (LL144)
The share of a category scoring above the median score of everyone assessed. Appears in modules 12, 13, 14, 23.
SCP — Service Control Policy
An AWS Organizations policy that sets the maximum permissions for accounts in an organization. Appears in modules 17, 31.
ScreenLab
The course's reference specification for a deliberately vulnerable LLM résumé screener. Learners do not build it; they use it as a hypothesis list when auditing the hosted résumé-screener lab. Appears in module 14.
scrypt —
A memory-hard password hashing and key-derivation function, recommended when Argon2id is unavailable. Appears in module 0.
SCTP — Stream Control Transmission Protocol
A transport protocol, like TCP, that delivers messages reliably and can use several network paths at once. Appears in module 2.
SD — Standard Deviation
A measure of how spread out values are around their average. Appears in modules 13, 14, 21, 24.
SDK — Software Development Kit
A package of libraries and tools for building software against a platform. Appears in modules 15, 17, 20.
SDLC — Software Development Life Cycle
The stages software passes through, from requirements to retirement. Appears in modules 1, 18.
SDO — STIX Domain Object
A STIX object for a high-level concept such as Indicator, Malware or Attack Pattern. Appears in module 28.
SE — Standard Error
How much an estimate, such as a success rate, would vary from sample to sample; smaller is more precise. Appears in modules 6, 8, 11, 13, 16, 21, 38, 44.
SEAL — Simple Encrypted Arithmetic Library (unverified)
Microsoft's open-source library for homomorphic encryption, which lets programs compute on encrypted data. Appears in module 34.
SEC — Securities and Exchange Commission
The U.S. regulator of securities markets; its 2023 rule requires public companies to disclose material cybersecurity incidents and describe cybersecurity risk management and governance. Appears in modules 14, 42, 43, 44, 47.
Seccomp — Secure computing mode
A Linux kernel feature that limits which system calls a process may make. Appears in modules 5, 17, 27.
Secret sharing —
Splitting a value into random-looking shares that reveal nothing alone but reconstruct the value together. Appears in module 34.
Secrets management —
Storing, distributing, rotating and auditing passwords, tokens and keys through a central system. Appears in modules 17, 34, 44.
Secrets manager
A service that stores secrets encrypted, controls and logs access, and rotates them. Appears in modules 5, 8, 17, 25, 31, 34.
secure boot —
A startup process in which each stage checks a digital signature on the next and refuses to run unsigned code. Appears in module 35.
Secure code review —
A person reading code to find security problems, guided by a threat model and secure coding standards. Appears in module 18.
Secure Enclave —
Apple's hardware-based key manager, isolated from the main processor, that never releases plain-text keys. Appears in modules 30, 34, 35.
security.txt —
A text file defined in RFC 9116 that a website publishes to say how to report security issues. Appears in modules 22, 39.
Seeded defect
A known vulnerability planted in a test system so you can measure whether an audit finds it. Appears in module 16.
Segmentation — Network segmentation
Dividing a network so each system can reach only what it needs. Appears in modules 5, 17, 27, 35.
SEI — Software Engineering Institute
A federally funded research center at Carnegie Mellon University that hosts CERT/CC. Appears in modules 22, 25.
Selection procedure
Any measure or procedure used as a basis for an employment decision, including AI tools. Appears in modules 1, 2, 5, 6, 7, 8, 12, 13, 14, 15, 16, 17, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 32, 33, 37, 39, 40, 41, 42, 44, 46, 47, 99.
Selection rate —
The number of people who passed a stage divided by the number who were assessed at that stage. Appears in modules 1, 7, 11, 12, 13, 14, 18, 19, 20, 21, 23, 24, 25, 26, 27, 33, 34, 35, 36, 37, 38, 41, 43, 45, 46.
SEM — Standard Error of Measurement
The typical distance between a single observed score and its true score, in score units. Appears in module 16.
Semantic entropy
Uncertainty measured over groups of sampled answers that mean the same thing. Appears in module 8.
Semantic versioning
A MAJOR.MINOR.PATCH version scheme signaling breaking changes, additions and fixes. Appears in module 24.
Semi-honest adversary —
A corrupted party that follows the protocol but tries to learn from what it sees. Appears in module 34.
Sensitivity — Sensitivity (differential privacy)
The most that one person's record can change the output of a calculation. Appears in modules 9, 36.
Sensitivity — Sensitivity (detection rate)
The probability that a test or audit flags a problem when the problem really exists. Appears in module 16.
Sensitivity (DP) —
The most that one person's record can change the output of a calculation. Appears in module 9.
Server-side request forgery —
Making a server send requests to destinations an attacker chooses, often inside the internal network. Appears in modules 0, 4, 5, 8, 17, 25, 99.
Service account
An account that an application, service or scheduled job runs as, rather than a person. Google Cloud calls its workload identities service accounts. Appears in modules 0, 3, 5, 12, 15, 17, 23, 29, 30, 31, 32, 33.
Session ID — Session identifier
A token that stands for a logged-in user across requests; whoever holds it is treated as that user. Appears in modules 4, 20, 30, 32.
Set cover — Set cover problem
Choosing the cheapest set of items whose combined coverage includes every requirement. Appears in module 12.
set point —
The target value a controller tries to maintain, such as a temperature. Appears in module 35.
Setuid —
A Linux permission bit that makes a program run with its file owner's identity, often root, whoever starts it. Appears in modules 0, 5.
Severity axes — Security and people-impact severity axes
Rating each finding separately for technical security impact and for harm to people, then reporting the higher. Appears in module 14.
SFT — Supervised Fine-Tuning
Further training on example instructions and desired responses. Appears in modules 7, 37.
SFTP — SSH File Transfer Protocol
File transfer over an SSH connection, often used for payroll and benefits feeds. Appears in module 34.
SFU — Service for User
Microsoft's Kerberos extensions (S4U2Self and S4U2Proxy) that support delegation. Appears in module 3.
SGD — Stochastic Gradient Descent
The standard training method that updates model weights step by step using gradients from small batches of data. Appears in modules 8, 9.
SGX — Software Guard Extensions
Intel's processor feature for creating isolated enclaves; a widely studied trusted execution environment. Appears in module 34.
SHA — Secure Hash Algorithm
A family of standard hash functions; SHA-256 is used to fingerprint evidence files. Appears in modules 0, 6, 30, 35.
SHA-1 — Secure Hash Algorithm 1
An older 160-bit hash function that NIST has retired because of collision attacks. Appears in module 30.
SHA-256 — Secure Hash Algorithm 256-bit
A cryptographic hash used to show a file has not changed. Appears in modules 0, 6, 8, 14, 17, 18, 20, 21, 22, 23, 27, 28, 30, 34, 35, 37.
Shadow stack
A protected second copy of return addresses that is compared with the normal stack on every return. Appears in module 27.
SHAP — SHapley Additive exPlanations
An explanation method that splits a prediction among features using Shapley values. Appears in modules 16, 38.
Shapley value —
A fair-division rule from cooperative game theory, used by SHAP to split a prediction among features. Appears in modules 16, 38.
Shared responsibility model
The division of security tasks between a cloud provider and its customer, which changes with the service model. Appears in modules 15, 17, 31.
Shared services
A central unit providing a standard service to many business units, often with service levels and chargeback. Appears in modules 35, 42, 44, 45.
Sharing group —
In MISP, an explicit list of organizations allowed to receive an event. Appears in modules 28, 33.
Shor's algorithm —
A quantum algorithm that factors integers and computes discrete logarithms in polynomial time, breaking RSA and elliptic-curve cryptography. Appears in module 34.
Shortfall
How many more people in a group would have passed if the group had passed at the reference group's rate. Appears in modules 13, 14, 21.
Showback —
Reporting a business unit's share of technology cost without moving money. Appears in modules 40, 44.
Shrinkage — Shrinkage (partial pooling)
Pulling noisy group estimates toward a shared average, more strongly for groups with less data. Appears in module 21.
SHRM — Society for Human Resource Management
A professional association for HR practitioners, with local chapters that host events. Appears in module 24.
SHS — Secure Hash Standard
NIST's FIPS 180-4, the standard that defines the SHA-1 and SHA-2 hash functions. Appears in module 0.
SI — Subsequent System Integrity
CVSS v4.0 metric for integrity impact on systems other than the vulnerable one. Appears in modules 6, 12.
SIAM — Society for Industrial and Applied Mathematics
A professional society that publishes applied mathematics journals. Appears in module 34.
SID — Security identifier
The unique number Windows uses to identify a user, group or computer. Appears in modules 0, 3, 5.
Side channel —
Leakage of secret information through timing, errors, power, caches or other effects rather than the intended output. Appears in modules 8, 17, 34, 35, 36.
SIEM — Security information and event management
A system that collects, stores and queries security logs. Appears in modules 3, 10, 20, 28, 29, 30, 32, 42.
SIF — Safety Instrumented Function
One specific safety action performed by a safety instrumented system, such as closing a valve on high pressure. Appears in module 35.
SIG — Special Interest Group
A FIRST working group on one topic, such as cyber insurance. Appears in modules 6, 17, 22, 33.
SIGCAS — Special Interest Group on Computers and Society
The ACM group concerned with the social effects of computing. Appears in module 34.
SIGCSE — ACM Special Interest Group on Computer Science Education
The ACM group whose annual symposium publishes computing-education research. Appears in modules 27, 39.
SIGKDD — Special Interest Group on Knowledge Discovery and Data Mining
The ACM group that runs the KDD data-mining conference. Appears in modules 8, 16.
Sigma —
An open, YAML-based format for writing log-detection rules that tools convert into many query languages. Appears in modules 10, 20, 28, 29, 31, 32, 39.
Significance level — alpha
The p-value cut-off chosen before the analysis, usually 0.05, below which a result is called statistically significant. Appears in modules 13, 14, 21, 24, 37, 45.
Signpost
An observable event chosen in advance that would make a scenario more or less likely and trigger a planned response. Appears in modules 42, 46.
SIGOPS — ACM Special Interest Group on Operating Systems
The ACM group that publishes Operating Systems Review. Appears in module 15.
SIGSAC — Special Interest Group on Security, Audit and Control
The ACM group that runs the CCS computer security conference. Appears in modules 7, 8, 9, 34.
Sigstore
An open-source project for signing software artifacts with short-lived certificates and a public transparency log. Appears in modules 8, 17, 18.
SIOP — Society for Industrial and Organizational Psychology
The professional body that publishes the Principles for the Validation and Use of Personnel Selection Procedures. Appears in modules 7, 12, 13, 16, 19, 22, 24, 37, 99.
SIR — Specific Intelligence Requirement
A narrower question that, together with others, answers a priority intelligence requirement. Appears in module 33.
SIS — Safety Instrumented System
An independent system that brings a process to a safe state when dangerous conditions are detected. Appears in module 35.
SISA — Sharded, Isolated, Sliced, Aggregated (training)
A training method that splits data into shards with separate sub-models, so deleting a record needs only one shard to be retrained. Appears in module 36.
SIV — Synthetic initialization vector
A design in which the nonce used internally is derived from the message, so a repeated external nonce does not cause catastrophic failure; used in AES-GCM-SIV (RFC 8452). Appears in module 34.
Skill decay —
The loss of a trained skill or knowledge over a period of non-use. Appears in module 39.
Skip-level meeting
A conversation between a leader and people two or more levels below them. Appears in module 41.
SLA — Service-Level Agreement
A committed time limit; in vulnerability management, the deadline to remediate a finding of a given priority. Appears in modules 6, 40.
SLH-DSA — Stateless Hash-Based Digital Signature Algorithm
NIST's post-quantum signature standard built only on hash functions, FIPS 205 (2024). Appears in modules 0, 34.
SLI — Service-Level Indicator
The actual measurement used to judge whether a service-level objective was met. Appears in module 40.
Slice — Evaluation slice
A subset of test items, such as one language or one group, whose results are reported separately. Appears in module 37.
Sliding window —
A time window anchored on each event, such as the hour before it, which catches bursts that straddle clock boundaries. Appears in module 32.
SLO — Service-Level Objective
The target value for a service measure, such as 99.9% monthly availability. Appears in module 40.
SLSA — Supply-chain Levels for Software Artifacts
A framework of build levels for software supply-chain integrity. Appears in modules 17, 18.
SMB — Server Message Block
The Windows file-sharing protocol. Appears in modules 3, 5, 23.
SME — Subject-Matter Expert
A person with expert knowledge of a job who helps define its content and requirements. Appears in module 13.
SMEs — Subject Matter Experts
People with deep knowledge of a job who help define its tasks and requirements. Appears in modules 13, 26.
SMT — Satisfiability Modulo Theories
A family of automated solvers that decide whether logical formulas over data types such as strings and numbers can be true. Appears in module 31.
SNMP — Simple Network Management Protocol
A protocol for monitoring and managing network devices, often on UDP port 161. Appears in module 2.
SOC — Security Operations Center
The team and facility that monitors and responds to security events. Appears in modules 6, 10, 20, 28, 29, 30, 33.
SOC — System and Organization Controls
The AICPA's family of independent accountant reports on controls at an organization; SOC 2 covers security and related criteria for service organizations. Appears in module 12.
SOC — Standard Occupational Classification
The U.S. federal system that groups jobs into occupations with numeric codes, used by the Bureau of Labor Statistics and O*NET. Appears in modules 26, 40, 41, 47.
SOC 1 — SOC 1 report
An AICPA report on a service organization's controls relevant to its customers' internal control over financial reporting. Appears in module 12.
SOC 2 — SOC 2 report
An AICPA attestation report on a service organization's controls against the Trust Services Criteria (security, plus optionally availability, processing integrity, confidentiality and privacy). It is a report, not a certificate. Appears in modules 1, 12, 27, 40, 43, 46.
SOC 3 — SOC 3 report
A short, general-use report on the same criteria as SOC 2, without the detailed tests and results. Appears in module 12.
SoD — Separation of Duties
Splitting a sensitive task so that no single person can complete it alone, such as creating an employee and approving the payroll that pays them. Appears in module 25.
Softmax
A function that turns a list of scores into probabilities that sum to one. Appears in modules 7, 38.
SoK — Systematization of Knowledge
A research paper type that organizes and evaluates the existing work in a field. Appears in modules 27, 35.
SOSP — Symposium on Operating Systems Principles
A leading research conference on operating systems. Appears in module 37.
Source reliability —
A rating (A to F) of how trustworthy a source has been, judged separately from any single report. Appears in module 33.
SOW — Statement of Work
A contract document describing one engagement's tasks and deliverables. Appears in modules 6, 14, 26.
SP — Special Publication
NIST's series name for its guidance documents, as in SP 800-115. Appears in modules 0, 1, 5, 6, 8, 12, 18, 20, 25, 26, 29, 30, 31, 33, 34, 35, 36, 40, 42, 43, 46, 99.
SP — Service Provider
In SAML, the application that relies on the identity provider's assertion. Appears in module 17.
SP 800-115 — NIST Special Publication 800-115
NIST's 2008 Technical Guide to Information Security Testing and Assessment, including an ROE template. Appears in modules 1, 2, 5, 6, 14, 29.
SP 800-53 — NIST Special Publication 800-53
NIST's catalog of security and privacy controls, organized into 20 families; Revision 5 is current. Appears in modules 0, 1, 12, 14, 25.
SPA — Single-Page Application
A web app that runs mostly in the browser and loads data without full page reloads. Appears in module 17.
SPACE — Satisfaction and well-being, Performance, Activity, Communication and collaboration, Efficiency and flow
A framework for developer productivity that says to measure across several of these dimensions and levels rather than with one metric. Appears in module 44.
Spacing effect —
Spreading practice sessions over time gives better long-term retention than cramming the same practice into one block. Appears in module 39.
Span
One timed operation within a trace, with attributes describing it. Appears in module 20.
Span of control
The number of people who report directly to one manager. Appears in modules 26, 41, 42.
Sparse cell
A table cell with very few people or selections, where large-sample methods become unreliable. Appears in module 21.
Spearman-Brown formula —
A formula giving the reliability of an average of k raters or items from the reliability of one: k r1 / (1 + (k - 1) r1). Appears in module 47.
Special categories — Special categories of personal data
GDPR Article 9 data, such as health, ethnic origin, trade union membership and identifying biometrics, whose processing needs an extra condition. Appears in modules 12, 19, 28, 36, 45, 99.
Specificity
The probability that a test or audit reports no problem when none exists. Appears in module 16.
SPF — Sender Policy Framework
A DNS record listing which servers may send email for a domain. Appears in module 2.
SPI — Schedule Performance Index
Earned value divided by planned value; below 1 means behind schedule. Appears in module 40.
SPL — Search Processing Language
Splunk's query language for searching and summarizing log data. Appears in modules 3, 32.
SPN — Service principal name
A label linking a network service to the account that runs it. Appears in modules 3, 4.
Spotlighting
Transforming untrusted input so the model can tell where it came from (Hines et al., 2024). Appears in modules 6, 8, 20, 39.
SQL — Structured Query Language
The language most relational databases use to read and change data. Appears in modules 1, 4, 7, 8, 15, 27, 32, 34, 39, 99.
SQS — Simple Queue Service
Amazon Web Services' managed message queue for passing work between programs. Appears in module 31.
SquashFS —
A compressed, read-only file system format common in embedded Linux firmware. Appears in module 35.
SR — Supply Chain Risk Management
SP 800-53 control family on risks from suppliers and components. Appears in modules 12, 40.
SR — Selection Rate
The number of a group who pass a stage divided by the number who reached it. Appears in module 13.
SR — Supervision and Regulation letter
A Federal Reserve guidance letter series; SR 11-7 (2011) is U.S. banking supervisors' guidance on model risk management. Appears in module 45.
SR 11-7 — Supervision and Regulation Letter SR 11-7
The Federal Reserve's 2011 guidance on model risk management; superseded on April 17, 2026, by SR 26-2. Appears in modules 45, 46.
SR 26-2 — Supervision and Regulation Letter SR 26-2
The Federal Reserve, OCC and FDIC's Revised Guidance on Model Risk Management (April 17, 2026), which replaced SR 11-7 and leaves generative and agentic AI models outside its scope. Appears in modules 45, 46.
SSAE — Statements on Standards for Attestation Engagements
The AICPA's standards for attestation engagements, codified in AT-C sections. Appears in module 12.
SSCP — Systems Security Certified Practitioner
An ISC2 certification for hands-on security operations and administration. Appears in module 39.
ssdeep
A widely used fuzzy hashing tool based on context triggered piecewise hashing. Appears in module 28.
SSDF — Secure Software Development Framework
NIST's outcome-based set of secure development practices, SP 800-218. Appears in module 18.
SSH — Secure Shell
An encrypted protocol for remote command-line logins and file transfer. Appears in modules 0, 5, 15, 17, 31, 34.
SSL — Secure Sockets Layer
The obsolete predecessor of TLS. Appears in module 34.
SSN — Social Security number
The nine-digit U.S. identifier used for tax and employment records; a high-value target for identity theft. Appears in module 34.
SSO — Single sign-on
Logging in once to reach many applications. Appears in modules 1, 10, 17, 25, 30, 34.
SSRF — Server-Side Request Forgery
A flaw that lets an attacker make a server send requests to places the attacker chooses, often internal systems. Appears in modules 1, 4, 5, 8, 17, 23, 25, 99.
SSRN — Social Science Research Network
An online repository for working papers and preprints, mainly in the social sciences and law. Appears in modules 22, 24.
SSTI — Server-side template injection
Injection of template syntax that the server's template engine executes. Appears in module 4.
Stack canary
A secret guard value placed before a function's return address and checked before the function returns. Appears in modules 5, 27.
Stack string
A string a program builds in memory from values inside instructions, so tools like strings cannot see it in the file. Appears in module 28.
Stakeholder salience
How much attention a stakeholder gets, based on their power, legitimacy and urgency. Appears in module 41.
Standard deviation —
A measure of how spread out values are around their average; the square root of the variance. Appears in modules 9, 12, 13, 14, 16, 17, 19, 21, 24, 26, 32, 36, 38, 39, 45, 46, 47.
Standard error —
The typical size of the chance wobble in an estimate; it shrinks as the sample grows. Appears in modules 6, 8, 11, 13, 16, 21, 24, 29, 37, 38, 44, 45.
STAR — Situation, Task, Action, Result
A way to structure answers to behavioral interview questions. Appears in module 26.
Static analysis
Examining a file without running it: its format, hashes, strings, imports and disassembled code. Appears in modules 18, 28, 31, 35, 39.
Statistical power
The probability that a test detects a difference of a stated size when that difference truly exists. Appears in modules 13, 21, 37, 45.
STD — Internet Standard
The label the RFC series gives to a specification that has become a full internet standard. Appears in modules 0, 2.
Step-up authorization
Asking for more permissions only at the moment a privileged action is first attempted. Appears in module 15.
Sticky bit —
A Linux directory bit that lets only a file's owner (or root) delete or rename it, as on /tmp. Appears in module 0.
STIX — Structured Threat Information Expression
A JSON-based standard format in which ATT&CK data is published. Appears in modules 10, 28, 33.
Stop condition
An event that requires testing to halt immediately, such as exposure of real personal data. Appears in modules 4, 6, 11, 14, 29.
Strategy (promptfoo) —
A promptfoo technique that wraps generated inputs in an attack pattern. Appears in module 11.
Strategy kernel
Rumelt's three parts of a good strategy: a diagnosis, a guiding policy and coherent actions. Appears in module 42.
STRIDE — Spoofing, Tampering, Repudiation, Information disclosure, Denial of service, Elevation of privilege
A checklist of six threat types used to find what could go wrong in a design. Appears in modules 1, 2, 5, 7, 18, 23, 25, 31, 36, 42, 99.
structured interview —
An interview in which every candidate gets the same job-related questions and answers are rated against predefined scales. Appears in modules 1, 11, 13, 16, 19, 23, 25, 26, 31, 40, 41, 44, 47.
STS — Security Token Service
The AWS service that issues temporary credentials, for example when a role is assumed. Appears in module 31.
Sub-technique —
A more specific version of a technique, shown with a suffix such as .001. Appears in modules 2, 8, 9, 10, 17, 28, 29, 32, 33, 99.
Subadditivity
The property that the measured risk of a combined portfolio is no more than the sum of the measured risks of its parts. Appears in module 46.
Subnet —
A block of consecutive IP addresses that share a common prefix, used to divide networks into zones. Appears in modules 0, 2, 4, 5, 31.
Subservice organization
Another service organization a vendor relies on, such as a cloud host or model-hosting provider. Appears in module 12.
substitutability — Substitutability
How easily a supplier could be replaced, in time, cost and disruption. Appears in module 44.
Succession planning
Identifying and developing people who could fill critical roles if the holder leaves. Appears in modules 41, 47.
Summary of results — Summary of results (LL144)
The public disclosure of an LL144 bias audit's data, rates and impact ratios. Appears in modules 12, 14, 37.
Super-timeline —
A single chronological sequence merging timestamped events from many sources, such as file systems, logs and the registry. Appears in module 30.
Superposition —
The hypothesis that a network stores more features than it has dimensions by using nearly orthogonal directions. Appears in module 38.
Supervised fine-tuning
Further training on example prompts and the completions you want, usually scoring only the completion tokens. Appears in modules 7, 8, 37, 38.
Surprisal —
The negative logarithm of a probability; summed over steps, it measures how unexpected a sequence is under a model. Appears in module 32.
SUT — System Under Test
The application or system being audited. Appears in module 23.
SWOT — Strengths, Weaknesses, Opportunities and Threats
A simple planning grid that lists internal strengths and weaknesses and external opportunities and threats. Appears in module 42.
Symbol table —
A description of an operating system kernel's data structures that memory analysis tools need to interpret raw memory. Appears in modules 27, 28, 30.
SYN — Synchronize
The first message of a TCP handshake. Appears in modules 0, 2.
Synthetic data
Generated records that represent no real person. Appears in modules 1, 7, 8, 9, 11, 13, 14, 18, 19, 20, 21, 22, 23, 24, 25, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39.
Synthetic identity —
A made-up person assembled from a mix of real and invented details, such as a real identity number with a different name and date of birth. Appears in module 25.
SYSTEM — Local System account
The most privileged built-in account on a Windows machine. Appears in module 5.
system of record — System of record
The single authoritative source for an important data entity, with documented definitions and owners. Appears in modules 15, 29, 40, 44.

T

t-closeness
A strengthening of l-diversity requiring each group's sensitive-value distribution to be within distance t of the whole table's. Appears in module 36.
T&Cs — Terms and Conditions
The contract terms a service provider sets for using its product. Appears in module 8.
T4 — NVIDIA T4
A data-center GPU model from NVIDIA that Google Colab has offered on its free runtime; availability varies and was not verified for this module. Appears in modules 12, 23, 37.
TA — Tactic (ID prefix)
The prefix of ATT&CK and ATLAS tactic identifiers, as in TA0001 or AML.TA0001. Appears in modules 10, 32.
TA — Trust Algorithm
The process a policy engine uses to decide access from identity, device, behavior and resource signals. Appears in module 17.
Table plan —
In Azure Monitor Logs, a table's storage class (Analytics, Basic or Auxiliary / Lake), which sets its cost and query and alert features. Appears in module 32.
Tabletop exercise
A discussion-based rehearsal of a realistic incident scenario to test plans, roles and decisions. Appears in modules 5, 12, 20, 41, 43, 46, 47, 99.
Tactic —
In ATT&CK and ATLAS, the attacker's goal for a step: the 'why'. Appears in modules 2, 5, 10, 28, 29, 33.
Tail dependence
The tendency of extreme losses in different risks to happen together. Appears in module 46.
Taint tracking
Labeling data with its source so untrusted data can be restricted. Appears in module 7.
TAP — Tree of Attacks with Pruning
An automated jailbreak search that explores many attack prompts and drops weak branches. Appears in module 11.
TARa — Total Agreement Rate (parsed answer)
Atil et al.'s measure of how often N repeated runs give the same parsed-out answer. Appears in module 16.
TARr — Total Agreement Rate (raw)
Atil et al.'s measure of how often N repeated runs give identical raw model output. Appears in module 16.
Task conflict
Disagreement about the content of the work, such as which option is best. Appears in module 41.
Task exposure —
The share of a job's tasks that a technology could affect; not a forecast of job loss. Appears in module 45.
TAXII — Trusted Automated Exchange of Intelligence Information
An OASIS standard application layer protocol and REST API for exchanging cyber threat intelligence, usually STIX. Appears in modules 28, 33.
TBM — Technology Business Management
A practitioner discipline and cost taxonomy, maintained by the TBM Council, for sorting technology costs into cost pools, towers and services so they can be compared and explained. Appears in module 44.
TC — Technical Committee
A standards body's working committee that develops a specification. Appears in module 17.
TCC — Theory of Cryptography Conference
An annual research conference on the theoretical foundations of cryptography. Appears in module 9.
TCM — TCM Security
The company that issues the PNPT certification. Appears in modules 23, 24, 99.
TCO — Total Cost of Ownership
All costs of owning and running something over its life, including implementation, staff time, compliance and exit, not just the price. Appears in modules 40, 44.
TCP — Transmission Control Protocol
A connection-based transport protocol that guarantees delivery. Appears in modules 0, 2, 35.
Team Topologies
A team design model by Skelton and Pais naming stream-aligned, platform, enabling and complicated-subsystem teams and three interaction modes. Appears in modules 41, 44.
Technical assistance —
Non-binding agency documents that explain how the agency reads existing law. Appears in modules 13, 19, 24, 25, 99.
Technical debt —
The future extra effort caused by choosing a quick solution now; it grows with the number of future changes to the affected area. Appears in modules 40, 42, 44.
Technique —
In ATT&CK and ATLAS, how an attacker achieves a tactic. Appears in modules 10, 28, 29, 33.
TEE — Trusted Execution Environment
An isolated area of a processor that runs sensitive code and protects keys from the main operating system. Appears in modules 34, 35.
TEF — Threat Event Frequency
How often a threat acts against an asset in a given time frame. Appears in module 12.
Temperature
A sampling setting: lower makes output more predictable, higher more random. Appears in modules 6, 7, 8, 14, 16.
template —
The stored mathematical summary of a person's biometric, used for later comparison instead of the raw image. Appears in module 35.
Tenant —
One customer organization in a shared, multi-customer service. Appears in modules 1, 2, 3, 4, 5, 6, 8, 10, 11, 14, 17, 18, 22, 23, 24, 25, 29, 30, 31, 32, 34, 35, 37, 39.
Terraform state —
Terraform's record of the resources it manages and all their attributes, which can include secrets and must be protected. Appears in module 31.
Test-retest reliability —
How consistently a measure gives the same result when the same thing is measured again later. Appears in modules 7, 16.
TGS — Ticket-Granting Service
The KDC function that issues service tickets in exchange for a TGT. Appears in module 3.
TGT — Ticket-granting ticket
A Kerberos ticket proving to the KDC that you already signed in; encrypted with the krbtgt key. Appears in module 3.
Third country
Under the GDPR, any country outside the EEA. Appears in modules 31, 36.
Threat actor —
A person, group or organization that carries out or intends to carry out an attack. Appears in modules 1, 2, 10, 29, 30, 33.
Threat hunting —
A proactive, human-led search for threats that existing detections did not catch. Appears in module 32.
Threat landscape —
A strategic product describing the threats relevant to an organization, their likelihood, impact and trends. Appears in module 33.
Threat model —
A statement of who might attack a system, what they know and can change, and what they want. Appears in modules 1, 3, 4, 8, 9, 10, 11, 16, 17, 18, 19, 23, 24, 25, 27, 29, 30, 31, 35, 36, 38, 39, 99.
Threat modeling —
Structured analysis of what can go wrong with a design, often organized by four questions and STRIDE. Appears in modules 1, 4, 7, 18, 36, 38, 99.
Threat-informed defense —
Prioritizing defenses according to how relevant attackers actually behave. Appears in modules 10, 29, 33.
Three Lines Model
The IIA's description of first-line (owns risk), second-line (expertise and challenge) and third-line (independent assurance) roles under a governing body. Appears in modules 39, 46.
Three-valued logic —
SQL's rule that a comparison with a missing (NULL) value is neither true nor false, so rows can silently drop out of results. Appears in module 32.
Three-way handshake —
TCP's connection set-up: SYN from the client, SYN-ACK from the server, ACK from the client. Appears in module 0.
Throughput
The total number of tokens a server produces per second across all users. Appears in module 37.
TIBER-EU — Threat Intelligence-Based Ethical Red Teaming (European Union framework)
The European Central Bank's framework for intelligence-led red team tests of financial institutions (details unverified in this course). Appears in module 29.
Tiering — Tiered administration
Separating admin accounts and devices by level of control so higher-tier credentials never touch lower-tier machines. Appears in module 3.
Time bucketing —
Rounding timestamps down to fixed intervals, such as the clock hour, to count events per interval. Appears in module 32.
Title VII — Title VII of the Civil Rights Act of 1964
US law prohibiting employment discrimination based on race, color, religion, sex and national origin. Appears in modules 6, 12, 13, 14, 19, 21, 22, 25, 30, 47, 99.
TLP — Traffic Light Protocol
FIRST's labels (RED, AMBER, AMBER+STRICT, GREEN, CLEAR) that tell recipients how far they may share information. Appears in modules 28, 33.
TLS — Transport Layer Security
The protocol that encrypts web traffic. Appears in modules 0, 2, 17, 25, 34, 35.
TLSH — Trend Micro Locality Sensitive Hash
A fuzzy hash for similarity comparison that needs at least 50 bytes of sufficiently varied input. Appears in module 28.
TOGAF — The Open Group Architecture Framework
A widely used enterprise architecture framework published by The Open Group, built around the Architecture Development Method. Appears in module 40.
Token
The unit of text a model reads and writes, usually a word piece. Appears in modules 7, 8, 37, 44.
Token bucket —
A rate limiter with a burst size and a refill rate; a client can make at most burst plus rate times time requests. Appears in module 4.
Token exchange — OAuth 2.0 Token Exchange
Swapping one token for another with a different audience or scope, optionally recording delegation (RFC 8693). Appears in module 17.
Token passthrough
An anti-pattern where a server forwards a client's access token downstream without checking it was issued to that server. Appears in module 15.
Tokenization —
Replacing a sensitive value with a random surrogate whose mapping is kept in a protected vault. Appears in module 34.
Tokenizer —
The component that splits text into the tokens a language model reads, and turns tokens back into text. Appears in modules 7, 8, 18, 27, 37.
Tombstone
A minimal record that a deletion happened, kept so that restored backups can be cleaned again. Appears in module 36.
tone at the top —
The ethical climate senior leaders set through what they say, do, reward and tolerate. Appears in modules 12, 47.
Tool call —
An action an LLM agent takes through a function the application exposes, such as sending an email. Appears in modules 7, 8, 10, 11, 12, 15, 20, 23, 30, 32, 33, 39.
Tool calling
A model outputting a structured request for application code to run a function. Appears in modules 7, 15.
Tool-call anomaly —
A tool call that is unusual for its session: an unexpected tool, argument, destination, sequence or rate. Appears in module 32.
TPM — Trusted Platform Module
A security chip that generates and protects keys and records boot measurements in registers. Appears in module 35.
TPR — True Positive Rate
The share of real positives a test correctly flags; for membership inference, the share of true members detected. Appears in modules 9, 13, 14, 20, 25, 28, 32, 38.
TPRM — Third-Party Risk Management
Identifying, assessing, contracting for and monitoring the risks that suppliers bring. Appears in modules 40, 46.
Trace
The record of one request's path through a system, made of spans. Appears in module 20.
TRAIGA — Texas Responsible Artificial Intelligence Governance Act
Texas's AI law, enacted as HB 149 in the 89th Legislature. Appears in modules 12, 19.
Training data extraction —
Getting a generative model to output pieces of its training data word for word. Appears in module 9.
Transfer of training —
Applying what was learned in training to a new setting such as the job, and keeping it over time. Appears in module 39.
Transformation modifier —
A Sigma value modifier that changes the value itself, such as base64; backend-agnostic. Appears in module 32.
Transformer
The attention-based neural network architecture behind modern LLMs. Appears in modules 7, 8, 11, 13, 26, 37, 38.
Transition services agreement
An agreement in which a seller keeps providing services to a divested business for a limited time after closing. Appears in module 42.
Transparency log
A public, append-only record of signing events, so unexpected signatures can be noticed. Appears in modules 2, 8, 18, 30.
TRL — Transformers Reinforcement Learning
Hugging Face's open-source (Apache-2.0) library for post-training language models, with trainers for SFT, DPO and other methods. Appears in module 37.
Trojan horse
A non-replicating program that looks harmless but has a hidden malicious purpose. Appears in module 28.
True positive rate —
The share of truly positive cases that a test or model correctly flags; also called sensitivity or recall. Appears in modules 2, 9, 10, 13, 14, 20, 25, 28, 32, 33, 38.
Trust boundary
A place where data passes between parts of a system with different levels of trust. Appears in modules 0, 1, 3, 5, 7, 10, 14, 15, 23, 24, 25, 27, 28, 36.
Trusted account —
In AWS cross-account access, the account that holds the principal making the request. Appears in module 31.
Trusted agent
A person on the defending side who knows about a covert test and helps the white cell keep it safe. Appears in module 29.
Trusted referee —
In NIST SP 800-63A-4, a vetted, trained agent who makes risk-based identity proofing decisions for applicants who cannot complete the normal process. Appears in module 25.
Trusting account —
In AWS cross-account access, the account that holds the resource being accessed. Appears in module 31.
TSA — Timestamp Authority
A service that signs a trusted timestamp proving data existed at a certain time. Appears in modules 18, 42.
TSC — Trust Services Criteria
The AICPA criteria a SOC 2 is measured against, in five categories: security, availability, processing integrity, confidentiality and privacy. Appears in modules 12, 43.
TTFT — Time To First Token
How long a user waits before the first output token appears; driven mostly by queueing and processing the prompt. Appears in module 37.
TTL — Time to live
A counter that limits how long something lasts: the number of router hops left for an IP packet, or how long a DNS answer may be cached. Appears in module 0.
TTP — Tactics, Techniques and Procedures
How an attacker operates: their goals, methods and specific steps. Appears in modules 2, 28, 29, 33.
TTPs — Tactics, Techniques and Procedures
Shorthand for attacker behavior at three levels: goals, methods and specific observed implementations. Appears in modules 2, 10, 11, 29, 33.
Tukey fences —
Outlier limits set at a multiple of the interquartile range below the first quartile and above the third. Appears in module 32.
Two-proportion z-test
A test of whether two success rates differ by more than chance would explain. Appears in modules 6, 8, 12, 13, 22, 23, 35.
TXT — Text record
A DNS record holding free text, often used to prove domain ownership or publish email policy. Appears in modules 0, 2.
Type 1 — SOC 2 Type 1 report
A SOC 2 report on whether controls were suitably designed as of a specific date. Appears in module 12.
Type 2 — SOC 2 Type 2 report
A SOC 2 report on whether controls were suitably designed and operated effectively throughout a period. Appears in module 12.
Type confusion
Accessing a resource as a type incompatible with the one it was created as (CWE-843). Appears in modules 5, 27.
Type modifier —
A Sigma value modifier that tells the backend to treat a value differently, such as re or cidr; the backend must support it. Appears in module 32.

U

UA — User Agreement
A platform's contract with its users; HTB's UA is cited in Module 23. Appears in module 23.
UAC — userAccountControl
An AD attribute of bit flags controlling account behavior, such as whether pre-authentication is required. Appears in module 3.
UAC — User Account Control
Windows' consent prompt before elevated actions; Microsoft treats it as a security feature, not a boundary. Appears in module 5.
UAF — Use After Free
Using memory through a pointer after that memory has been freed (CWE-416). Appears in module 27.
UB — Undefined Behavior
Operations whose result the C or C++ standard does not define, such as reading past an array; the compiler may assume they never happen. Appears in module 27.
UBSan — UndefinedBehaviorSanitizer
A compiler instrumentation that reports undefined behavior, and optionally unsigned integer wraparound, at run time. Appears in module 27.
UC — Universally composable (security)
Canetti's framework in which a protocol proven secure stays secure when combined with any other protocols. Appears in module 34.
UDP — User Datagram Protocol
A connectionless transport protocol with no delivery guarantee. Appears in modules 0, 2.
UEBA — User and Entity Behavior Analytics
Tools that build behavioral baselines for users and devices and score deviations from them. Appears in module 32.
UEFI — Unified Extensible Firmware Interface
The modern firmware interface on PCs and servers that starts the operating system and supports Secure Boot. Appears in module 35.
UGESP — Uniform Guidelines on Employee Selection Procedures
Common abbreviation for the Uniform Guidelines, 29 CFR Part 1607. Appears in modules 13, 19, 47.
UI — User Interaction
CVSS metric for whether another person must act: None, Passive or Active. Appears in modules 4, 6, 8, 11, 32.
UID — User Identifier
The number Linux uses to identify a user account. Appears in modules 0, 5, 35.
UK — United Kingdom
The United Kingdom of Great Britain and Northern Ireland. Appears in modules 37, 38.
Umask —
A per-process mask of permission bits that are removed from newly created files. Appears in module 0.
Unawareness and unintervenability
A LINDDUN threat: people do not know about, or cannot influence, how their data is processed. Appears in modules 1, 36, 39.
Uniform Guidelines — Uniform Guidelines on Employee Selection Procedures (1978)
Federal guidance at 29 CFR Part 1607 on how agencies evaluate adverse impact and validity of selection procedures. Appears in modules 1, 2, 10, 12, 13, 14, 16, 19, 21, 22, 23, 24, 26, 33, 36, 37, 38, 39, 40, 41, 42, 43, 45, 46, 47, 99.
Unit economics —
Cost expressed per unit of business value, such as cloud cost per application processed. Appears in modules 40, 42, 44, 45.
Unit test —
A small test of one piece of logic with a known expected result. For a detection rule, it is a hand-written event that should or should not match. Appears in modules 21, 32.
Upper echelons theory
Hambrick and Mason's proposal that strategic choices and performance are partly predicted by top managers' backgrounds. Appears in modules 42, 47.
URI — Uniform Resource Identifier
A string that identifies a resource, such as a web address. Appears in modules 4, 11, 15, 17, 25.
URL — Uniform Resource Locator
A web address, such as https://example.com/jobs. Appears in modules 0, 1, 4, 5, 7, 8, 10, 12, 14, 17, 18, 20, 24, 25, 26, 30, 31, 32, 33, 37, 45, 99.
URLs — Uniform Resource Locators
Plural of URL: web addresses. Appears in modules 1, 4, 6, 7, 8, 9, 10, 14, 17, 18, 19, 25.
US — United States
The United States of America. Appears in modules 12, 13, 19, 29, 30, 31, 33, 36, 37, 38, 39, 40, 43, 45, 46, 47, 99.
USB — Universal Serial Bus
The standard connector and protocol for removable drives and peripherals, a common route for copying data off a computer. Appears in module 30.
USCIS — U.S. Citizenship and Immigration Services
The U.S. agency that administers Form I-9 guidance and the E-Verify program. Appears in module 25.
USD — United States Dollar
The U.S. currency, used here for illustrative cost figures. Appears in modules 8, 24, 28, 31, 33, 36, 39, 45.
Use-after-free —
A memory bug in which a program keeps using memory after it has been freed, so the memory may already hold other data (CWE-416). Appears in modules 5, 27.
User Account Control —
The Windows feature that runs even administrators' programs with standard rights until an elevation prompt is approved. Appears in modules 0, 5.
UTAUT — Unified Theory of Acceptance and Use of Technology
A model of what drives people's intention to use, and actual use of, a new technology. Appears in module 41.
UTC — Coordinated Universal Time
The world time standard used for log timestamps. Appears in modules 20, 30.
UTF — Unicode Transformation Format
A family of text encodings; UTF-8 is the standard encoding for text in files and on the web. Appears in modules 18, 34.
utilization —
The share of an independent practitioner's working time that is billable to clients. Appears in module 26.
UUID — Universally Unique Identifier
A 128-bit label generated so that two systems almost never create the same one. Appears in module 4.

V

VA — Vulnerable System Availability
CVSS v4.0 metric for availability impact on the vulnerable system. Appears in module 6.
Validity
The degree to which evidence and theory support interpretations of scores for their proposed uses. Appears in modules 13, 16, 21, 26, 41, 47.
Value modifier —
A Sigma keyword after a field name and a pipe, such as contains or gte, that changes how the value is matched. Appears in module 32.
Vanity metric
A number that looks impressive but does not track the outcome you care about. Appears in modules 24, 26, 36, 43.
VaR — Value at Risk
The loss level that will not be exceeded with a stated probability over a stated period, such as the 99% one-year VaR. Appears in module 46.
Variance explained —
The share of differences between people on one measure that lines up with another measure; for a correlation r it is r squared. Appears in modules 39, 41.
VAST — Visual, Agile and Simple Threat modeling (unverified)
A threat-modeling method built to scale across many teams by using process flow diagrams. Appears in module 1.
VC — Vulnerable System Confidentiality
CVSS v4.0 impact on confidentiality of the vulnerable system. Appears in module 6.
VDP — Vulnerability disclosure policy
An organization's published rules for reporting security flaws. Appears in modules 1, 22, 23, 39.
VDR — Virtual data room
A secure online repository where a seller shares documents with potential buyers during due diligence. Appears in module 42.
VDSO — Virtual Dynamic Shared Object
A small library the Linux kernel maps into every process to speed up some system calls. Appears in module 27.
Vector
An ordered list of numbers, such as a candidate's feature values or a model's weights. Appears in modules 7, 9.
Vector database
A store for embeddings that finds the most similar ones quickly. Appears in modules 7, 8, 17.
Vector store —
A store for embeddings that finds the most similar ones quickly. Appears in modules 7, 8, 20, 31, 34.
verification (1:1) —
Checking a sample against the one template of the identity a person claims. Appears in module 35.
VI — Vulnerable System Integrity
CVSS v4.0 metric for integrity impact on the vulnerable system. Appears in modules 6, 12, 36, 47.
Virtual camera —
Software that presents itself to applications as a webcam but supplies arbitrary video, which can be used to inject deepfakes. Appears in module 25.
Virus — Computer virus
Malware that copies itself by inserting copies into other programs or files, usually when a person opens or runs them. Appears in module 28.
vLLM
An open-source (Apache-2.0) server for running LLMs on GPUs at high throughput, built on PagedAttention and continuous batching. Appears in module 37.
VM — Virtual Machine
A software computer that runs inside another computer or in the cloud. Appears in modules 2, 5, 6, 15, 17, 28, 30.
VMs — Virtual Machines
Plural of VM. Appears in modules 5, 15, 17, 28.
VP — Vice President
A senior executive title below the C-suite; in this module, the executive accountable for a technology or security function across the enterprise. Appears in modules 26, 42, 44.
VPC — Virtual Private Cloud
A logically isolated private network inside a cloud provider. Appears in module 31.
VPC Service Controls —
Google Cloud perimeters around specified services that reduce the risk of data being copied out. Appears in module 31.
VPN — Virtual Private Network
An encrypted tunnel that connects a device to a private network over the internet. Appears in modules 0, 2, 17, 24, 30, 34, 43, 99.
VSL — Volatility Software License
The license under which Volatility 3 is released. Appears in module 30.

W

W3C — World Wide Web Consortium
The standards body for the web; it publishes the Verifiable Credentials data model that Open Badges 3.0 builds on. Appears in module 0.
WAF — Web application firewall
A filter in front of a web application that blocks some malicious requests. Appears in modules 0, 4.
Wald interval —
A simple confidence interval: the estimate plus or minus 1.96 standard errors. It misbehaves with small samples or rates near 0 or 1. Appears in modules 11, 13, 16, 21.
Wargame —
A permanent, non-competitive series of security challenges in which each level unlocks the next. Appears in module 39.
Webhook —
An HTTP callback in which a system sends a message to a registered URL when an event happens, such as a candidate being hired. Appears in modules 0, 25, 32, 34.
WEP — Words of Estimative Probability
A fixed set of terms, such as likely or unlikely, each tied to a probability range. Appears in module 33.
WER — Word Error Rate
The share of words an automated speech recognition system gets wrong (substituted, deleted or inserted) compared with a reference transcript. Appears in module 25.
White cell
The small group that controls an exercise, answers deconfliction questions and can pause or stop it; also called a white team. Appears in module 29.
Wilson interval — Wilson score interval
A confidence interval for a proportion that behaves well near 0 and 1. Appears in modules 8, 9, 10, 11, 13, 14, 15, 16, 22, 23, 25, 29, 38, 39, 45.
Wilson score interval —
A confidence interval for a proportion that stays sensible with small samples and rates near zero. Appears in modules 8, 11, 16, 29.
WMDP — Weapons of Mass Destruction Proxy
A public benchmark of 3,668 multiple-choice questions used as a proxy for hazardous knowledge in biosecurity, cybersecurity and chemical security. Appears in module 38.
work profile —
On Android, a separately managed area of the phone controlled by the employer, kept apart from the personal profile. Appears in module 35.
work sample —
A selection exercise that reproduces part of the job, such as a take-home report or a live test of a lab application. Appears in modules 1, 6, 13, 16, 25, 26, 41, 44, 47.
Workload identity federation
Letting software prove its identity with a platform-issued token to get short-lived cloud credentials, with no stored key. Appears in modules 17, 31.
Worm — Computer worm
A self-contained program that copies itself to other hosts, usually without any user action. Appears in modules 28, 35.
WP — Working Party
Short for the Article 29 Working Party, the group of EU data protection authorities that came before the European Data Protection Board. Appears in module 19.
Write blocker —
A hardware or software tool that lets a computer read a storage device while preventing any writes to it. Appears in module 30.
Write-up —
A written account of how a challenge was approached and solved, ideally teaching the method and the defense. Appears in modules 23, 39.
WSJF — Weighted Shortest Job First
A prioritization rule: cost of delay divided by job duration; higher scores go first. Appears in modules 41, 42.
WSTG — Web Security Testing Guide
OWASP's catalog of web application test cases. Appears in modules 1, 4, 6.
WTLS — Wireless Transport Layer Security
An early security layer for mobile data protocols, studied in Vaudenay's padding-oracle paper. Appears in module 34.

X

X.509 —
The standard format for public-key certificates used on the internet, profiled in RFC 5280. Appears in module 0.
XML — Extensible Markup Language
A tag-based text format for structured data. Appears in modules 11, 17, 25, 32.
XMSS — eXtended Merkle Signature Scheme
A stateful hash-based signature scheme approved in NIST SP 800-208. Appears in module 34.
XOR — Exclusive or
A bitwise operation that outputs 1 where two bits differ and 0 where they match; widely used inside ciphers and MACs. Appears in modules 0, 34.
XSS — Cross-site scripting
Injection of script into a web page so it runs in other users' browsers. Appears in modules 4, 8, 23, 99.
XSW — XML Signature Wrapping
Attacks that trick an application into reading unsigned XML while the signature check passes on a different element. Appears in module 17.

Y

YAML — YAML Ain't Markup Language
A human-readable configuration file format based on indentation. Appears in modules 8, 10, 11, 20, 24, 25, 29, 31, 32.
YARA
A pattern-matching tool and rule language for identifying and classifying files, especially malware, using strings and Boolean conditions. Appears in modules 28, 39.
YARA-X
VirusTotal's re-implementation of YARA, intended to replace it as the default pattern-matching tool for malware researchers. Appears in module 28.

Z

z-test — two-proportion z-test
A large-sample test of whether two rates differ by more than chance would explain, using the normal (bell) curve. Appears in modules 13, 14, 21.
ZAP — Zed Attack Proxy
An open-source web application scanner. Appears in module 18.
ZBB — Zero-based budgeting
A budgeting method in which every activity is justified from zero each cycle instead of adjusting last year's budget. Appears in module 42.
zCDP — Zero-Concentrated Differential Privacy
A variant of differential privacy measured by a parameter rho that gives tighter accounting when many noisy releases are combined; used by the 2020 US Census. Appears in module 36.
Zelkova —
The automated-reasoning engine AWS IAM Access Analyzer uses to translate policies into logic and check them with SMT solvers. Appears in module 31.
ZeRO — Zero Redundancy Optimizer
A training technique that splits a model's weights, gradients and optimizer state across GPUs so each GPU stores only a share. Appears in module 37.
Zero trust —
Security paradigms that grant no implicit trust based on network location or asset ownership. Appears in modules 17, 35, 42.
ZIP — ZIP archive
A compressed file format that bundles many files into one. Appears in modules 5, 12, 13, 19, 28, 36, 38.
zone —
In ISA/IEC 62443, a group of assets that share security requirements. Appears in module 35.
ZT — Zero Trust
Security paradigms that grant no implicit trust based on network location or asset ownership. Appears in module 17.
ZTA — Zero Trust Architecture
An enterprise plan that applies zero trust principles to infrastructure and workflows (NIST SP 800-207). Appears in modules 17, 35.
ZTMM — Zero Trust Maturity Model
CISA's staged model of progress toward zero trust across pillars. Appears in module 17.