Course menu

Free preview. This lesson is open to everyone. See the full course.

Module 0 · Lesson 4 of 4 (step 5 of 8) · about 45 minutes

Self-check and roadmap: what the rest of the course assumes

A diagnostic across networking, operating systems and cryptography, with a map of which later modules lean on which idea, so you can decide what to review before Module 01 and what you can safely skim.

Checks completed: 0 of 3

In this lesson (10 sections)
  1. Why this matters
  2. Core concepts
  3. Deep dive
  4. Worked example
  5. In people systems
  6. Lab
  7. Common misconceptions
  8. Open questions and research frontiers
  9. Key terms
  10. References
Seen from above, a learner writes notes at a desk covered with open books and papers.
Photo: Jacob Bentzinger on Unsplash

Why this matters

This course is written for people who arrive from very different places. A security engineer may know handshakes by heart and never have heard of the . An technology lead may run an applicant tracking system () every day and never have read a file permission string. An consultant may be expert in validity evidence and new to both. Everyone needs the same foundations before Module 01, but nobody needs to relearn what they already know.

This lesson helps you decide where to spend your time. It does three things. First, it gives you a short diagnostic: a set of questions across the three earlier lessons, written so that a correct answer requires understanding rather than recall. Second, it maps each foundation idea to the later modules that depend on it, so you can see the cost of a gap. Third, it shows you how to judge your own readiness more accurately, because people are often poor judges of what they know.

The diagnostic is not graded. Its value comes from answering honestly before checking, then using the result to plan.

Core concepts

Prerequisite knowledge as a dependency graph

A prerequisite is knowledge a later topic assumes. Prerequisites form a dependency graph: a set of topics joined by arrows that mean "is needed for". If you have a gap in a topic near the start of many arrows, the gap spreads.

Drawing the diagram…

Dependency graph. Networking feeds modules 02, 04, 17 and 20. Operating systems feed modules 03, 05 and 20, and the Codespaces labs in 08, 11 and 15. Cryptography feeds modules 04, 17, 18 and 19. Module 01 sits after all three foundations and before every later module.
Diagram source
%% alt: Dependency graph. Networking feeds modules 02, 04, 17 and 20. Operating systems feed modules 03, 05 and 20, and the Codespaces labs in 08, 11 and 15. Cryptography feeds modules 04, 17, 18 and 19. Module 01 sits after all three foundations and before every later module.
flowchart LR
  N["L01 Networking"] --> M02["02 Recon and scanning"]
  N --> M04["04 Web application security"]
  N --> M17["17 Cloud and identity"]
  N --> M20["20 Detection and response"]
  O["L02 Operating systems"] --> M03["03 Active Directory"]
  O --> M05["05 Privilege escalation"]
  O --> LAB["Terminal labs in 08, 11, 15"]
  O --> M20
  C["L03 Cryptography"] --> M04
  C --> M17
  C --> M18["18 Secure ML engineering"]
  C --> M19["19 Privacy engineering"]

Text description of the diagram: the networking lesson feeds Module 02 (reconnaissance and scanning), Module 04 (web application security), Module 17 (cloud and identity security) and Module 20 ( and incident response). The operating systems lesson feeds Module 03 (Active Directory), Module 05 (exploitation and ), the terminal-based labs in Modules 08, 11 and 15, and Module 20. The cryptography lesson feeds Module 04, Module 17, Module 18 (secure engineering) and Module 19 ().

The diagram is a simplification. Almost every module uses a little of everything. The arrows show where a gap will hurt most.

What each foundation lesson unlocks

Networking (Lesson 01). Module 02's scanning lessons assume you can read blocks, know what an open means and understand the . Module 04 assumes you can read requests and responses, know what a header and a cookie are, and understand that the client controls everything it sends. Module 17 builds on , and proxies when it discusses cloud networks and identity federation. Module 20 reads network and proxy logs.

Operating systems (Lesson 02). Module 03 builds on Windows accounts, and access control lists. Module 05 is largely about how permission mistakes such as programs, writable service files and over-privileged lead to privilege escalation. Modules 08, 11 and 15 run tools in a Linux terminal inside your own GitHub Codespace, so you need basic shell skills. Module 20 assumes you can read authentication events and service logs.

Cryptography (Lesson 03). Module 04's lesson on authentication and sessions assumes you know why passwords are hashed with slow functions and how session tokens must be random. Module 17's lesson on federation, , OpenID Connect and (Security Assertion Markup Language) relies on signatures, certificates and keyed (message authentication codes). Module 18 uses hashes and signatures to protect the model and data supply chain. Module 19 uses keyed and encryption in privacy engineering.

Shell basics the course assumes

Several labs ask you to type commands into a terminal. You do not need to be fluent, but you should be comfortable with this short list. Each one appeared in Lesson 02's lab.

Command What it does
pwd, ls -l, cd show where you are, list files with permissions, change directory
cat, less, head print or page through a file
mkdir, cp, mv, rm make a directory, copy, move, remove
chmod, umask, getfacl change and inspect permissions
whoami, id, sudo show your identity; run a command as another user
ps, grep list processes; search text
curl make an HTTP request from the terminal
` (pipe) and>` (redirect)

If any line here is unfamiliar, repeat Lesson 02's lab before Module 02.

Calibration: knowing what you know

(just "" in this lesson) is how well your confidence matches your accuracy. It is a different idea from the score calibration used in fairness work in Module 13. If you are well calibrated, then of all the answers you are 80% sure of, about 80% are right. Poor calibration has two forms: overconfidence, where you are surer than your accuracy justifies, and underconfidence, the reverse. Overconfidence is the more expensive error in this course, because it leads you to skip foundations and then struggle in later labs without knowing why.

In this lesson you will rate your confidence on each diagnostic item before checking. That makes the diagnostic measure two things: what you know, and how well you know what you know.

Check your understanding

A learner rates 10 answers at 90% confidence and gets 6 of them right. What does this suggest?

Deep dive

Measuring calibration with a score

One simple way to summarize across many items is the . For each item ii, let fif_i be your stated probability that your answer is correct (for example, 0.8) and let oio_i be 1 if the answer was correct and 0 if not. Over NN items, the score is

B=1N∑i=1N(fi−oi)2.B = \frac{1}{N}\sum_{i=1}^{N} \left(f_i - o_i\right)^{2}.

In words: for each item, take the gap between your confidence and what happened, square it, and average across items. Here NN is the number of items, fif_i is your confidence on item ii as a number between 0 and 1, and oio_i is the outcome. Lower is better. A perfect, fully confident and always correct learner scores 0. Answering every item at 0.5 confidence scores 0.25 regardless of outcomes. A learner who is often confidently wrong can score worse than 0.25.

The score mixes two things: how often you are right, and how well your confidence tracks it. That is useful here, because both matter for planning. The worked example below computes it.

Why answering before checking matters

Answering a question from memory, before seeing the answer, is a form of . The course brief asks for it in every lesson for this reason: recalling an idea strengthens later recall more than rereading it. Rereading, in contrast, creates a feeling of familiarity that is easily mistaken for understanding, which is one cause of overconfidence. The checks in each lesson and the cumulative review at the start of each later module are both built on this idea. (Unverified here: this lesson states the principle as the course brief's design basis and does not cite a specific study.)

Fair use of a diagnostic

A diagnostic like this one is a low-stakes measurement. It is reasonable to use it to plan your own study. It would not be reasonable to use it to rank people or make decisions about them, because a dozen items cannot cover a domain, the items have not been validated against any criterion, and a single attempt is noisy. This is the same reasoning an consultant applies to any assessment: the uses a score can support depend on the validity evidence for those uses (Module 13). Keep that in mind if anyone proposes reusing course quizzes to screen job candidates.

Worked example

Scenario. A learner, Sam, is an technology lead. Before reading the answers, Sam answers five diagnostic items and records confidence for each:

Item Topic Sam's confidence fif_i Correct? oio_i
1 Size of a /22 block 0.9 1
2 What does and does not protect against 0.8 0
3 Reading -rw-r----- 0.6 1
4 Why must be unique 0.5 0
5 Which password hash to choose 0.9 1

Step 1: accuracy. Sam got 3 of 5 right, 60%.

Step 2: average confidence. (0.9+0.8+0.6+0.5+0.9)/5=3.7/5=0.74(0.9 + 0.8 + 0.6 + 0.5 + 0.9)/5 = 3.7/5 = 0.74. Sam was, on average, 74% sure and 60% right: mildly overconfident.

Step 3: . The squared gaps are (0.9−1)2=0.01(0.9-1)^2 = 0.01, (0.8−0)2=0.64(0.8-0)^2 = 0.64, (0.6−1)2=0.16(0.6-1)^2 = 0.16, (0.5−0)2=0.25(0.5-0)^2 = 0.25 and (0.9−1)2=0.01(0.9-1)^2 = 0.01. The sum is 1.071.07, so B=1.07/5=0.214B = 1.07/5 = 0.214.

Step 4: read the pattern. The single biggest contribution is item 2, where Sam was 80% sure and wrong. Confident errors are the most valuable to find, because they are the ones Sam would never have checked.

Step 5: plan. Item 2 maps to Lesson 01 (NAT and ), which feeds Modules 02, 04, 17 and 20. Item 4 maps to Lesson 03 (-GCM), which feeds Modules 04, 17, 18 and 19. Sam's plan: reread Lesson 01's sections on NAT and firewalls, then redo Lesson 03's lab Part C, then retake those two items in a week.

Faded example

Alex, a security engineer, answers five items:

Item Confidence fif_i Correct? oio_i
1 1.0 1
2 1.0 1
3 0.9 1
4 1.0 0
5 0.7 1

Accuracy is 80%. Average confidence is (1.0+1.0+0.9+1.0+0.7)/5=0.92(1.0 + 1.0 + 0.9 + 1.0 + 0.7)/5 = 0.92. Now finish: compute the five squared gaps and the Brier score, identify which item contributes most, and say what Alex should review.

Check your work: the squared gaps are 0, 0, 0.01, 1.00 and 0.09, so B=1.10/5=0.22B = 1.10/5 = 0.22. Item 4, answered with complete confidence and wrong, contributes almost all of it.

Check your understanding

A learner answers four items with confidences 0.8, 0.6, 0.9 and 0.5. The outcomes are correct, wrong, correct, correct. What is the ? Give three decimal places.

Independent problem

Take the full diagnostic in the Lab section. Record your confidence for each item before you check it. Compute your accuracy, your average confidence and your Brier score. Then write a study plan: list your two most confident errors, the lesson section each maps to, the later modules that depend on it, and the date you will retest.

In people systems

Mixed teams need shared foundations. Audits of hiring tools work best with mixed teams: security testers, technology owners, legal or privacy staff and consultants. Each brings knowledge the others lack. A short common foundation, like this module, lets them talk about the same request, the same file and the same key without talking past each other.

matters in audit judgments too. Auditors make many judgments under uncertainty: is this finding exploitable, is this adverse- meaningful, is this vendor claim credible? Overconfident auditors miss things; underconfident ones bury clients in caveats. Practicing calibration on low-stakes items is practice for that professional skill.

Do not turn learning checks into hiring tests. It can be tempting for an employer to reuse a training diagnostic as a screening test for security or HR technology roles. Without , validation and adverse-impact monitoring, that would repeat the problems this course teaches you to find in AI hiring tools. The diagnostic here is for learners' own planning only.

Role-based paths. HR and governance readers may decide to read the foundations for meaning and skip the optional terminal parts of labs, then focus on Modules 01, 06, 12, 13 and 14. Security readers may skim Module 00 and spend extra time on Modules 07, 13 and 21, where the hiring-specific and statistical content is new. Both paths still reach the same capstone.

Lab Lab map

Goal: take a diagnostic, measure your and produce a written study plan. Everything happens in your browser and on paper.

Tools: this page, a notebook or spreadsheet (any online spreadsheet you own works), and optionally CyberChef for item D2.

Part A: the diagnostic (25 minutes). For each item, write your answer and a confidence between 0.5 and 1.0 before looking anything up. Then check against the lesson indicated.

Networking (Lesson 01)

  • N1. How many addresses are in 172.16.32.0/20?
  • N2. Is 172.32.5.4 inside the 1918 private range 172.16.0.0/12?
  • N3. Name the three segments in the handshake, in order.
  • N4. An vendor's receives a request with X-Forwarded-For: 198.51.100.7. What, if anything, does that prove about the client's address?
  • N5. Which part of an request to an AI vendor can a network observer usually still see?

Operating systems (Lesson 02)

  • O1. Write the octal mode for -rwxr-x---.
  • O2. With 077, what mode does a file requested as 0666 get?
  • O3. On Linux, you own a file with mode ---rw-rw-. Can you read it?
  • O4. What is the difference between a Windows and ?
  • O5. Which Windows event records a failed logon?

Cryptography (Lesson 03)

  • C1. Why is a poor choice for storing passwords?
  • . What goes wrong if - reuses an with the same key?
  • C3. A and a both detect tampering. What can a signature do that a MAC cannot?
  • C4. In 1.3, which message proves that the server holds the private key for its certificate?
  • C5. Name 's post-quantum key-encapsulation standard and its number.

Shell basics

  • S1. What does ls -l | grep csv do?
  • S2. What does sudo -l show?

Part B: score it (10 minutes). Mark each item 1 or 0. Compute accuracy, average confidence and the . In a spreadsheet, add a column for (fi−oi)2(f_i - o_i)^2 and average it.

Part C: plan (10 minutes). List your three most confident errors. For each, write the lesson section, the later modules that depend on it (use the dependency graph above), and one action: reread, redo a lab part, or ask in office hours. Set a retest date about a week away.

Answer key (check only after Part A). N1: 4,096. N2: no; the block ends at 172.31.255.255. N3: , SYN-, ACK. N4: nothing by itself; a client can write any value, so it is meaningful only if your own proxy set it. N5: addresses, , sizes and timing, usually the lookup and the server name. O1: 750. O2: 0600. O3: no; as owner the owner bits apply, and they grant nothing. O4: the DACL decides access; the SACL decides which access attempts are audited. O5: 4625. C1: it is fast and, unless salted, identical passwords share hashes; offline guessing is cheap. C2: an attacker can likely recover the hash subkey and forge , and relationships leak. C3: anyone with the public key can verify it, and only the private-key holder can create it, so it convinces third parties. C4: CertificateVerify. C5: , FIPS 203. S1: lists files in long form and keeps only lines containing "csv". S2: which commands your account may run with sudo on this machine.

Deliverable: your scored table and the study plan.

Safety: this lab involves no systems at all.

Check your understanding

Match each gap to the later module where it would cause the most trouble.

Common misconceptions

  • "I have worked near these systems for years, so I can skip the foundations." Familiarity is not the same as being able to explain a mechanism. The diagnostic tests explanation.
  • "A high score on a self-check means I am ready." Only if your confidence was also accurate. A high score with several confident errors still points to specific gaps.
  • "Being unsure means I don't know it." Underconfidence is real. If you were right but unsure, a short review will usually fix the confidence, not the knowledge.
  • "Foundations are for beginners." Experts revisit them too. Many serious findings, such as world-readable secrets or reused , are foundation-level mistakes made by experienced teams.
  • "A diagnostic score can be reused to compare people." Not without validity evidence for that use.

Open questions and research frontiers

  • Adaptive diagnostics. Could the platform choose the next question based on earlier answers, to estimate readiness with fewer items? makes this possible in principle; how well it works for a fast-changing technical field, where item banks age quickly, is open.
  • Transfer. Does strong performance on foundation items predict success in later hands-on labs? The course could study this with learners' consent. It is the same predictive-validity question that asks about selection tests.
  • training for auditors. Can calibration practice on technical questions improve the accuracy of real audit judgments, such as severity ratings? There is little direct evidence for security work specifically.
  • Fair prerequisites. Do assumed foundations disadvantage learners from non-traditional backgrounds, and does an open foundation module reduce that gap? Free access is a design choice that this module makes partly for that reason; its effect is an empirical question.

Key terms

  • Prerequisite: knowledge a later topic assumes.
  • Dependency graph: topics joined by "is needed for" arrows.
  • Diagnostic: a low-stakes test used to plan learning, not to judge people.
  • : how well confidence matches accuracy.
  • Overconfidence / underconfidence: confidence above / below actual accuracy.
  • : the average squared gap between stated probability and outcome.
  • : recalling an idea from memory before checking it.
  • Shell: a program that reads typed commands and runs them, such as bash.
  • Pipe / redirect: sending a command's output to another command, or to a file.
  • : applicant tracking system.
  • : Security Assertion Markup Language, a federation standard covered in Module 17.

References

  1. Braden, R. (Ed.). Requirements for Internet Hosts: Communication Layers. 1122, October 1989. https://www.rfc-editor.org/rfc/rfc1122.html (checked 2026-11-08). Basis for Lesson 01's layer model used in items N3–N5.
  2. Fuller, V., and Li, T. Classless Inter-domain Routing (). RFC 4632, August 2006. https://www.rfc-editor.org/rfc/rfc4632.html (checked 2026-11-08). Items N1–N2.
  3. Rekhter, Y., and others. Address Allocation for Private Internets. RFC 1918, February 1996. https://www.rfc-editor.org/rfc/rfc1918.html (checked 2026-11-08). Item N2.
  4. Linux man-pages project. path_resolution(7) and (2). https://raw.githubusercontent.com/mkerrisk/man-pages/master/man7/path_resolution.7 and https://raw.githubusercontent.com/mkerrisk/man-pages/master/man2/umask.2 (checked 2026-11-08). Items O1–O3.
  5. Microsoft Learn. "Access Control Lists" and "4625(F): An account failed to log on". https://learn.microsoft.com/en-us/windows/win32/secauthz/access-control-lists and https://learn.microsoft.com/en-us/previous-versions/windows/it-pro/windows-10/security/threat-protection/auditing/event-4625 (checked 2026-11-08). Items O4–O5.
  6. Dworkin, M. Recommendation for Block Cipher Modes of Operation: Galois/Counter Mode () and . 800-38D, November 2007, Appendix A. https://csrc.nist.gov/pubs/sp/800/38/d/final (checked 2026-11-08). Item .
  7. Rescorla, E. The Transport Layer Security () Protocol Version 1.3. RFC 8446, August 2018. https://www.rfc-editor.org/rfc/rfc8446.html (checked 2026-11-08). Item C4.
  8. National Institute of Standards and Technology. Module--Based Key-Encapsulation Mechanism Standard. 203, 13 August 2024. https://csrc.nist.gov/pubs/fips/203/final (checked 2026-11-08). Item C5.
  9. Cheat Sheet Series. Password Storage Cheat Sheet. https://cheatsheetseries.owasp.org/cheatsheets/Password_Storage_Cheat_Sheet.html (checked 2026-11-08). Item C1.
  10. Anthony Hammock. Hiring Audit Lab (companion lab repository; and setup). https://github.com/anthonyhammock/ai-hiring-audit-lab (checked 2026-11-08). Basis for the terminal-lab dependencies in Modules 08, 11 and 15.