The Cyber Security Body of Knowledge (CyBOK) Study Guide and Chapter Summaries – CyBOK Project (Version 1.2) BS Cyber Security

BS Cyber Security and BSIT students can use the Cyber Security Body of Knowledge, known as CyBOK, free from the official website cybok.org. It is not one textbook. It is a set of 22 PDF documents: an Introduction and 21 Knowledge Areas that range from risk, law and human factors to cryptography, malware, operating systems, web security and hardware. Version 1.2 was released on 22 September 2026 under the Open Government Licence v3.0.

Think of it as a map of the whole subject, not a classroom text with exercises. It is the best place to see how cryptography, network security, software security and incident response fit together, and every Knowledge Area ends with a table that points to the textbooks and papers behind it. Use it beside your course book for revision, to choose electives and to find a reliable source for a term paper. It is dense, so this page gives a plain summary, key points and a study route for each document.

Book Overview

CourseCyber Security Fundamentals / Introduction to Cyber Security
Degree ProgramsBS Cyber Security, BSIT
LevelUniversity / Undergraduate (reference for all levels, up to postgraduate)
EditionCyBOK Version 1.2, released 22 September 2026; individual Knowledge Areas carry their own versions, from 1.0.1 to 2.0.1
AuthorsThe CyBOK project: one author per Knowledge Area, with academic editors and reviewers; the Introduction is by Andrew Martin, Awais Rashid, Howard Chivers, Steve Schneider, Emil Lupu, George Danezis and Yulia Cherdantseva
CopyrightCrown Copyright, The National Cyber Security Centre 2026
StructureAn Introduction plus 21 Knowledge Areas in five categories, roughly 1,150 pages in total including reference lists
ExercisesNone; CyBOK is a reference framework and each Knowledge Area ends with a cross-reference table to textbooks and papers
LanguageEnglish
Total Chapters22 documents (1 Introduction + 21 Knowledge Areas), each a separate PDF
LicenseOpen Government Licence v3.0 (OGL v3.0)
FormatFree PDF documents, one per Knowledge Area, from the official CyBOK site

Chapter List

Introduction to CyBOK (Version 1.2.0)

Introductory Concepts · Difficulty: Easy · Key topics: definition of cyber security, 21 Knowledge Areas, five categories, adversary models, controls, vulnerabilities, risk, design principles, supply chain security

The Introduction explains why CyBOK exists. Cyber security knowledge is scattered, so students and teachers find it hard to see a clear path through the subject. Following the model of the software engineering body of knowledge (SWEBOK), CyBOK maps knowledge that already exists in textbooks, papers and standards instead of rewriting it. The document gives a working definition of cyber security, introduces the 21 Knowledge Areas in five categories, and shows how to use them together: adversary models, controls, vulnerabilities, failures and incidents, and risk. It then covers design principles and two crosscutting themes, security architecture and the lifecycle, and supply chain security.

Key Points:

  • Working definition (from the UK National Cyber Security Strategy 2016-2021): protecting information systems, the data on them and the services they provide from unauthorised access, harm or misuse, including harm caused by accident.
  • CyBOK has 21 Knowledge Areas in five categories, and the document says the categories overlap and other groupings would be equally valid.
  • Saltzer and Schroeder’s eight design principles are the starting set; the NIST principles are grouped into architecture and design, capability and intrinsic behaviour, and life cycle.
  • The “layer below” problem: attackers get around a control by attacking beneath the abstraction that the control relies on, for example through a side channel.
  • Reason’s Swiss Cheese model is used to explain latent design conditions that line up to cause a failure, especially in cyber-physical systems.
  • Version 1.2.0 (22 September 2026) removed security economics, which now has its own Knowledge Guide, and added supply chain security as a crosscutting theme.

Memory Tip: Read the 21 Knowledge Areas as a stack: people, law and risk on top, then attacks and defences, then systems, software and infrastructure underneath.

Common Mistake: Treating the five categories as separate boxes. The Introduction stresses that Knowledge Areas depend on each other, so a security architect needs more than one of them.

Important Questions:

  • What is CyBOK’s working definition of cyber security? Protecting information systems (hardware, software and infrastructure), the data on them and the services they provide from unauthorised access, harm or misuse, whether intentional or accidental.
  • What is the “layer below” problem? A security control depends on an abstraction, and an attacker who works below that abstraction (hardware, timing, memory) can bypass the control without breaking it directly.

Risk Management & Governance (Version 1.1.2)

Human, Organisational & Regulatory Aspects · Difficulty: Medium · Key topics: risk definition, ALARP, risk governance, threat, vulnerability, likelihood, impact, ISO/IEC 27005, NIST SP 800-30, FAIR, STRIDE, security metrics, incident response

This Knowledge Area starts by defining risk as the possibility that actions or events have consequences for things people value, and it separates risk assessment, risk management and risk governance. It explains why a risk may be judged intolerable, tolerable or acceptable, and why perceived risk often differs from actual risk. The governance section compares technocratic, decisionistic and transparent models and covers security culture and metrics. The longest section compares component-driven methods such as ISO/IEC 27005, NIST SP 800-30, FAIR and attack trees with system-driven methods such as STAMP and SABSA. It ends with business continuity and incident response.

Key Points:

  • The elements of a risk assessment are vulnerability, threat, likelihood and impact.
  • A risk can be intolerable, tolerable (reduced to a level that is as low as reasonably practicable, ALARP) or acceptable.
  • Renn’s four risk types are routine, complex, uncertain and ambiguous, and each needs a different governance approach.
  • STRIDE lists six threat types: spoofing, tampering, repudiation, information disclosure, denial of service and elevation of privilege.
  • ISO/IEC 27035-1 describes incident management as plan, detect, assess, respond and learn.
  • Tick-box compliance can create a false sense of security, because passing a checklist is not the same as managing a risk.
  • The KA cites a commonly quoted figure that about one in three breaches use a vulnerability for which a patch already existed.

Memory Tip: Say STRIDE out loud as a sentence: Spoofing, Tampering, Repudiation, Information disclosure, Denial of service, Elevation of privilege.

Common Mistake: Treating a compliance checklist as a risk assessment. A system can pass the checklist and still carry risks nobody has looked at.

Important Questions:

  • What is the difference between risk assessment, risk management and risk governance? Assessment identifies and measures risk, management chooses and applies controls, and governance sets the structures, responsibilities, policy and communication around both.
  • What do ALARP and ALARA mean? Both describe reducing a tolerable risk to a level that is as low as reasonably practicable (or allowable), instead of aiming for zero risk.

Law & Regulation (Version 1.0.3)

Human, Organisational & Regulatory Aspects · Difficulty: Hard · Key topics: jurisdiction, privacy and interception, GDPR, computer crime, Budapest Convention, contract, tort, intellectual property, intermediary liability, electronic signatures, export controls, ethics · Dated: the international data transfer examples stop at Safe Harbour and Privacy Shield

This is the longest Knowledge Area, written for readers with no legal training who work in more than one country. It states plainly that it is not legal advice. It begins with how law is made and researched, the difference between criminal and civil cases, and a simple risk function for legal exposure. Later sections cover jurisdiction, privacy and interception, data protection under the GDPR, computer crime, contract, tort, intellectual property, intermediary liability, electronic signatures, sector rules such as the NIS Directive, export controls on cryptography, public international law, ethics and vulnerability disclosure, and legal risk management.

Key Points:

  • Jurisdiction comes in three kinds: prescriptive (who makes the law), juridical (which court can hear the case) and enforcement (who can act on the ruling).
  • Computer crime offences cover improper access, interference with data or systems, interception and hacking tools; the Budapest Convention (2001) is a main international reference.
  • UK Computer Misuse Act 1990: up to two years for improper access and five years for data or system interference, rising to 14 years or life for serious damage after the 2015 amendment.
  • GDPR security obligations sit in Article 32, with privacy by design in Article 25 and impact assessments in Article 35.
  • In English contract law a contract usually needs offer, acceptance, consideration and an intention to create legal relations.
  • Section 13 on ethics covers client obligations, codes of conduct, and vulnerability testing and disclosure.

Memory Tip: For every legal scenario ask three questions in order: where is it happening, which law applies, and who is able to enforce it.

Common Mistake: Assuming one country’s law applies everywhere. The KA stresses that laws differ because they rest on territorial sovereignty, and this book does not cover Pakistani law.

Important Questions:

  • What is the difference between criminal and civil liability? A criminal case is brought by the state and needs proof beyond reasonable doubt, while a civil case is a private claim for compensation decided on the balance of probabilities (preponderance of evidence).
  • Why does the Budapest Convention matter to security researchers? It is the main international treaty on computer crime, so it shapes which acts such as improper access or interference with systems are treated as offences and how states cooperate on them.

Human Factors (Version 1.0.2)

Human, Organisational & Regulatory Aspects · Difficulty: Easy · Key topics: usable security, compliance budget, alarm fatigue, human error, Swiss cheese model, awareness, positive security, developer usability

Human Factors argues that security fails when people cannot or will not use it, and rejects the idea that users are the weakest link. It uses email encryption and password policies as running examples, then builds the basics of usable security: fit the task to the human by looking at capabilities, goals, context and the device. Next it covers human error with Reason’s Swiss cheese model, awareness and education, positive security, and stakeholder engagement with employees and developers. A theme throughout is that security is a secondary task, so every extra step spends a limited compliance budget.

Key Points:

  • Usability in ISO 9241-11 means effectiveness, efficiency and satisfaction.
  • The compliance budget is the limited effort people will spend on security tasks before they start to cut corners.
  • Even a 10 percent false alarm rate can produce alarm fatigue, and SSL certificate warnings are reported to have false positive rates of 50 percent or more.
  • The Fogg Behaviour Model says behaviour needs motivation, ability and a trigger.
  • In a developer study, two thirds of those using Stack Overflow or a textbook met their functional goals against 40 percent using official documentation, but official documentation produced the most secure code.
  • NCSC password guidance recommends two-factor authentication and password managers, and no forced expiry of strong passwords.

Memory Tip: Remember Fogg’s three ingredients as M-A-T: Motivation, Ability, Trigger. If a secure action fails, one of the three is missing.

Common Mistake: Blaming users. When people bypass a policy, the KA treats it as a design failure of the policy, not a character flaw.

Important Questions:

  • What is the compliance budget? The limited amount of effort and time people are willing to give to security tasks. Once a policy asks for more, people bypass it.
  • Why do security warnings stop working? If most warnings are false alarms, people learn to click through them. This alarm fatigue means the one real warning is also ignored.

Privacy & Online Rights (Version 1.0.3)

Human, Organisational & Regulatory Aspects · Difficulty: Hard · Key topics: confidentiality, control, transparency, anonymisation, k-anonymity, differential privacy, Tor, mix networks, censorship resistance, e-voting, privacy engineering

This Knowledge Area argues that privacy is more than secrecy: leaks can lead to theft, manipulation and damage to democracy. It organises privacy technologies into three paradigms. Privacy as confidentiality covers encryption-based techniques, anonymisation and metadata protection such as Tor. Privacy as control covers privacy settings and policy negotiation. Privacy as transparency covers feedback tools and audit logs. A section on democratic values looks at electronic voting and censorship resistance, and a final section on privacy engineering covers design strategies and how to evaluate a privacy mechanism against a threat model.

Key Points:

  • The three paradigms are privacy as confidentiality, privacy as control and privacy as transparency.
  • k-anonymity, l-diversity and t-closeness are measures for anonymised data sets; the KA illustrates them by generalising ZIP codes in a table.
  • Differential privacy adds calibrated noise; with epsilon = 3 the probability ratio can be about 20.08 (e to the power 3), so values above 1 need scrutiny.
  • Tor sends traffic through three onion routers; mix networks and Loopix add delays and Sphinx packets to hide who talks to whom.
  • Electronic voting needs ballot secrecy, verifiability and coercion resistance.
  • Randomising browser fingerprint features can make a user more unique, not less.

Memory Tip: Match each paradigm to a verb: confidentiality hides, control lets you choose, transparency lets you see.

Common Mistake: Thinking that removing names makes data anonymous. The KA shows that quasi-identifiers such as ZIP code can still single people out.

Important Questions:

  • What are the three privacy paradigms in CyBOK? Privacy as confidentiality (hide data and metadata), privacy as control (let users decide what is exposed and how it is used) and privacy as transparency (show users their footprint and audit data use).
  • What does the epsilon value in differential privacy tell you? It measures how much one person’s data can change the output. A smaller epsilon means stronger privacy; epsilon = 3 allows a probability ratio of about e cubed, roughly 20.

Malware & Attack Technologies (Version 1.0.2)

Attacks & Defences · Difficulty: Medium · Key topics: malware taxonomy, potentially unwanted programs, cyber kill chain, botnets, APT, static and dynamic analysis, fuzzing, symbolic execution, evasion, sinkholing, attribution

The chapter defines malware and classifies it along six dimensions: standalone or host-program, persistent or transient, the layer of the system stack, auto-spreading or user-activated, static or dynamically updated, and solo or coordinated. It maps malicious activity onto the seven-step Cyber Kill Chain and describes botnets, APTs and the underground economy. Most of the chapter is about analysis: static and dynamic analysis, fuzzing, symbolic and concolic execution, analysis environments, and the anti-analysis tricks malware uses. It then covers detection with indicators of compromise and machine learning, and response through takedown, sinkholing and attribution.

Key Points:

  • The Cyber Kill Chain has seven steps: reconnaissance, weaponisation, delivery, exploitation, installation, command and control, and actions on objectives.
  • Static analysis reads code without running it and is defeated by packing; dynamic analysis watches behaviour but only sees the behaviour that gets triggered.
  • Malware can run at any layer from firmware and boot sector to kernel and application, and lower layers are harder to detect and remove.
  • Evasion tricks include packing, polymorphism, trigger-based behaviour and checks for virtual machines or missing user activity.
  • Sinkholing redirects command-and-control domains to defenders’ servers; attackers answer with fast-flux DNS and domain generation algorithms.
  • Potentially unwanted programs such as bundled adware should be treated as malware because adware can turn into spyware.

Memory Tip: Learn the kill chain in order as a story: Recon, Weaponise, Deliver, Exploit, Install, Command, Act.

Common Mistake: Trusting signatures alone. Polymorphism and mutation defeat signatures, so the KA also relies on behaviour-based detection.

Important Questions:

  • What is the difference between static and dynamic malware analysis? Static analysis examines the code without running it, so it is safe but can be defeated by packing. Dynamic analysis runs the sample in an isolated environment to see what it does, but only sees behaviour that is actually triggered.
  • What is sinkholing? Redirecting the domains a botnet uses for command and control to servers run by defenders, so infected machines can no longer receive instructions and defenders can count them.

Adversarial Behaviours (Version 1.0.2)

Attacks & Defences · Difficulty: Medium · Key topics: cyber-enabled and cyber-dependent crime, hacktivists, state actors, affiliate programmes, exploit kits, bulletproof hosting, attack trees, kill chains, situational crime prevention, attribution

This Knowledge Area explains who attacks and how their operations are organised. It first classifies adversaries by motivation and capability: interpersonal offenders, organised criminals, hacktivists and state actors, and it separates crimes that moved online from crimes that only exist because of technology. It then describes the parts of a malicious operation: affiliate programmes, infection vectors, hosting and command-and-control infrastructure, exploit kits, pay-per-install services, human services such as money mules, and payment methods. The last section gives models for analysis: attack trees, kill chains, environmental criminology, the underground economy as capital flow, and attack attribution.

Key Points:

  • Cyber-enabled crime is an existing crime moved online (such as fraud), while cyber-dependent crime (such as ransomware) exists only because of technology.
  • Clough’s five reasons criminals move online are reach, lower cost, speed, crossing borders easily and lower risk of capture.
  • Attack trees put the goal at the root and use AND and OR branches, so likelihood and cost can be calculated up the tree.
  • The Hutchins kill chain has seven phases and is mapped to defensive actions: detect, deny, disrupt, degrade and deceive.
  • Situational crime prevention applies criminology ideas such as routine activity theory and rational choice to cyber attacks, and warns about displacement.
  • Attribution works at tactical (how), operational (what) and strategic (why) levels, and exploit kits and leaked tools make exploit-based attribution weaker.

Memory Tip: Remember the five defensive verbs as D-D-D-D-D: Detect, Deny, Disrupt, Degrade, Deceive.

Common Mistake: Picturing attackers as lone hackers. The KA shows most criminal operations are built from bought services such as pay-per-install, bulletproof hosting and money mules.

Important Questions:

  • What is the difference between cyber-enabled and cyber-dependent crime? Cyber-enabled crime is a traditional crime carried out with technology, for example fraud. Cyber-dependent crime, such as ransomware or DDoS-for-hire, could not exist without computers and networks.
  • What is an attack tree? A diagram that puts the attacker’s goal at the root and breaks it into sub-goals joined by AND and OR nodes, so the cheapest or likeliest attack path can be worked out.

Security Operations & Incident Management (Version 1.0.3)

Attacks & Defences · Difficulty: Medium · Key topics: MAPE-K loop, IDS, SIEM, SOAR, NetFlow, Syslog, misuse and anomaly detection, base-rate fallacy, CVE, CVSS, MITRE ATT&CK, honeypots, CSIRT, incident response

Security Operations and Incident Management (SOIM) begins with James Anderson’s 1981 report and treats the field as an application of the MAPE-K loop: Monitor, Analyse, Plan, Execute, with Knowledge in the middle. It describes the architecture of intrusion detection systems, SIEM and SOAR, then walks through each stage. Monitoring covers packet capture, NetFlow, DNS, logs and Syslog. Analysis compares misuse and anomaly detection and how detectors are tested. Planning covers SIEM correlation, execution covers mitigation, and knowledge covers CVE, CVSS, CWE, CAPEC, ATT&CK, honeypots and threat intelligence. The last section is about people: preparing for, handling and following up incidents.

Key Points:

  • Misuse detection matches known signatures; anomaly detection flags deviations from a model of normal behaviour and raises more false alarms.
  • Detectors are measured with precision, recall and ROC curves, and Axelsson’s base-rate fallacy shows that rare attacks produce many false alarms even from an accurate detector.
  • IDS, SIEM and SOAR are three nested loops: detection, correlation of alerts with asset and vulnerability data, and automated response.
  • Knowledge bases include CVE (identifiers), CVSS (severity scores), CWE (weakness types), CAPEC (attack patterns) and MITRE ATT&CK (attacker techniques).
  • Incident management follows prepare, handle and follow up, drawing on NIST SP 800-61; handling includes analysis, mitigation (containment) and communication.
  • Honeypots cannot promise “no false positives”, because background Internet noise also hits them; the KA finds fake email addresses and fake database rows the most useful.

Memory Tip: Spell the loop as M-A-P-E-K and match it to a SOC job: monitor the data, analyse alerts, plan with the SIEM, execute the fix, keep the knowledge base current.

Common Mistake: Judging a detector by its detection rate alone. Without the false alarm rate and the base rate of real attacks, the number means very little.

Important Questions:

  • What is the difference between misuse detection and anomaly detection? Misuse detection looks for known bad patterns (signatures) and is accurate for known attacks. Anomaly detection learns normal behaviour and can find new attacks but produces more false alarms.
  • What does a SIEM do? It collects and normalises alerts and logs from many sources, correlates them with asset inventories, vulnerability scans and threat intelligence, and presents the result to analysts.

Forensics (Version 1.0.2)

Attacks & Defences · Difficulty: Medium · Key topics: digital evidence, Daubert, ACPO principles, write blockers, file carving, memory forensics, browser artifacts, cloud forensics, hashing, approximate matching

Forensics treats digital forensics as the use of scientific methods to identify, collect and analyse digital artifacts for legal proceedings, and leaves jurisdiction-specific law to the Law and Regulation KA. It starts with definitions and conceptual models, including evidence standards and the shift from state-centric to log-centric reconstruction. Then it works up the system: operating system and storage analysis (acquisition, file systems, file carving), main memory forensics, application forensics using web browsers as the case study, and cloud forensics. It ends with artifact analysis techniques such as cryptographic hashing, block-level hashing and approximate matching.

Key Points:

  • The legal background includes the UK Computer Misuse Act 1990, the US Daubert standard (which replaced Frye) and the ACPO good-practice principles for digital evidence.
  • Storage forensics works through layers: physical media, block device, file system and application artifacts; hardware write blockers and tools such as dd protect the original.
  • File carving recovers files from raw data using header and footer signatures, for example a JPEG starts FF D8 FF and ends FF D9.
  • Pulling the plug on a running machine was largely abandoned because RAM holds evidence; memory forensics must bridge the semantic gap between raw memory and objects.
  • Cloud forensics is mostly logical acquisition through APIs, because the provider holds the authoritative record and data is spread across many machines.
  • Approximate matching comes in bytewise, syntactic and semantic classes, and NIST’s National Software Reference Library supplies hash sets of known files.

Memory Tip: Think of evidence as four layers from bottom to top: media, block device, file system, application. Each layer hides or reveals different traces.

Common Mistake: Examining the original disk directly. The KA’s process starts with preserving integrity: image first with a write blocker, hash it, and analyse the copy.

Important Questions:

  • Why are cryptographic hashes used in forensics? A hash proves that an image or file has not changed since acquisition, and hash sets of known files let investigators filter out irrelevant material quickly.
  • Why is main memory forensics important? RAM holds running processes, network connections and keys that never reach the disk, and this evidence is lost once the machine is powered off.

Cryptography (Version 1.0.2)

Systems Security · Difficulty: Hard · Key topics: provable security, one-time pad, secret sharing, block ciphers, hash functions, modes of operation, MAC, RSA, ECDSA, key agreement, zero-knowledge, MPC, homomorphic encryption, side channels

This Knowledge Area covers the cryptography a cyber security expert should know, and it assumes first-year undergraduate mathematics. It sets up the maths and the security models used to define security, then moves to information-theoretic constructions such as the one-time pad and Shamir secret sharing. Next come symmetric primitives (block ciphers, stream ciphers, hash functions), modes of operation and MACs, public key encryption and signatures, standard protocols for authentication and key agreement, and advanced protocols such as oblivious transfer, zero-knowledge proofs and secure multi-party computation. It closes with schemes with special properties and implementation issues such as side channels and fault attacks.

Key Points:

  • Concrete security is judged against roughly 2^128 operations; “negligible” advantage is an asymptotic notion for schemes with growing key sizes.
  • AES has a 128-bit block with 128, 192 or 256-bit keys; DES has a 56-bit key and a 64-bit block and is too weak.
  • Shamir secret sharing is Q2 when the threshold t is below n/2 and Q3 when t is below n/3, in which case invalid shares can be corrected.
  • Modes of operation include CBC, CTR and GCM; MACs include CBC-MAC, HMAC and KMAC; hash designs include Merkle-Damgard and the SHA-3 sponge.
  • Public key encryption is built as a KEM-DEM hybrid; RSA-OAEP, RSA-KEM and ECIES are standard examples.
  • For DSA, ECDSA and Schnorr signatures, leaking the per-signature nonce is very damaging.
  • Paillier is the standard linearly homomorphic example, and such schemes cannot be IND-CCA secure.

Memory Tip: Keep two lists apart: primitives (AES, SHA-3, RSA) are the bricks, and schemes and protocols (GCM, ECDSA, TLS) are the buildings made from them.

Common Mistake: Mixing up encryption and authentication. A mode such as CBC gives confidentiality only; you also need a MAC or an authenticated mode such as GCM.

Important Questions:

  • What is the difference between a block cipher and a mode of operation? A block cipher such as AES encrypts one fixed-size block. A mode such as CBC, CTR or GCM defines how to use it on longer messages, and it decides which security properties you get.
  • Why must a signature nonce never leak or repeat? In discrete-log schemes such as DSA, ECDSA and Schnorr, the nonce is mixed with the private key, so a leaked or reused nonce lets an attacker work out the private key.

Operating Systems & Virtualisation Security (Version 1.0.2)

Systems Security · Difficulty: Hard · Key topics: isolation, attacker model, Rowhammer, side channels, microkernel, TCB, Bell-LaPadula, Biba, ACLs, capabilities, protection rings, SGX, ASLR, CFI

The Knowledge Area frames operating system security as isolating mutually distrusting security domains: processes, the kernel and virtual machines. It first describes the attacker model, including malicious kernel modules, memory errors, Rowhammer, double fetches and side channels. It compares OS designs (monolithic kernels, microkernels, library OSes and unikernels, virtualisation and containers) and applies Saltzer and Schroeder’s principles and the Bell-LaPadula and Biba models. A long section covers primitives for isolation: ACLs, capabilities, memory protection, hardware extensions such as TrustZone and SGX, and protection rings. The last section covers hardening: information hiding, control-flow restrictions, partitioning and integrity checks.

Key Points:

  • Studies cited in the KA suggest 1 to 6 bugs per 1000 lines of code, which is the case for a small trusted computing base.
  • Bell-LaPadula protects confidentiality (no read up, no write down) and Biba protects integrity with the mirror rules.
  • Isolation mechanisms have a history: chroot appeared in Version 7 Unix (1979), FreeBSD Jails in 2000 and Capsicum in FreeBSD 9.0 (2012); seL4 is formally verified.
  • Capabilities help avoid the confused deputy problem that ACLs can suffer from.
  • ASLR was introduced by PaX in 2001; the KA’s KASLR example has only 9 bits of entropy, so about 512 guesses are enough.
  • Control-Flow Integrity limits indirect jumps and calls to legitimate targets, with hardware help from Intel CET and ARM pointer authentication, but it cannot stop non-control-data attacks.
  • W xor X and the NX bit stop code injection, and SMEP and SMAP stop the kernel from executing or reading user memory.

Memory Tip: The five hardening families spell a story: hide, restrict, partition, check, watch (information hiding, control-flow restrictions, partitioning, integrity checks, anomaly detection).

Common Mistake: Assuming randomisation is strong protection. KASLR and coarse-grained ASLR break once one pointer leaks or a side channel reveals the layout.

Important Questions:

  • Why do operating system designers want a small trusted computing base? Every line of trusted code can contain bugs, and bug density is roughly 1 to 6 per 1000 lines, so less trusted code means fewer ways to break isolation.
  • What do Bell-LaPadula and Biba protect? Bell-LaPadula protects confidentiality by forbidding reads of higher levels and writes to lower levels. Biba protects integrity with the reverse rules: no reading down and no writing up.

Distributed Systems Security (Version 1.0.2)

Systems Security · Difficulty: Hard · Key topics: peer-to-peer overlays, DHT, Sybil attack, Eclipse attack, consensus, Paxos, Raft, Byzantine fault tolerance, CAP theorem, two-phase commit, blockchain

The Knowledge Area defines a distributed system as dispersed resources that act as one coordinated service, built on access control, data transport, coordination services and data management. It sorts systems into decentralised peer-to-peer designs and coordinated resource clustering. For peer-to-peer systems it explains unstructured, structured, hybrid and hierarchical overlays, then catalogues attacks such as Sybil, Eclipse and routing table poisoning with their mitigations. For coordinated systems it covers consensus, consistency models, replication, the CAP theorem and commit protocols, and finally how timing and value disruptions affect cloud, PKI, key-value stores, databases and blockchains.

Key Points:

  • Structured overlays such as Chord, Pastry, Tapestry, Kademlia and CAN use a distributed hash table, with identifiers drawn from 2^w values where w is typically 128 or 160.
  • Peer-to-peer attacks include Sybil (many fake identities), Eclipse (isolating a victim, often built on Sybil and routing table poisoning), pollution and white-washing.
  • Consensus needs agreement, validity and termination, and the FLP result says deterministic consensus is impossible in a fully asynchronous system.
  • Paxos and Raft provide crash-fault consensus; Byzantine fault tolerance needs more than 3f nodes to tolerate f malicious ones.
  • Consistency ranges from strong (strict, linearisability) to weak (sequential, causal, eventual); two-phase commit is widely used and three-phase commit less so.
  • In a blockchain each block holds the hash of the previous block; Proof-of-Work resists spam but uses a lot of energy, and Proof-of-Stake is less mature.

Memory Tip: Remember Byzantine fault tolerance as 3f + 1: to survive one traitor you need at least four nodes.

Common Mistake: Believing identical replicas give Byzantine fault tolerance. Identical copies share the same vulnerabilities, so the KA says diversity is needed too.

Important Questions:

  • What is a Sybil attack? An attacker creates many fake identities in an open peer-to-peer network to gain influence, which can then support attacks such as Eclipse.
  • Why does Byzantine fault tolerance need more than 3f nodes? Malicious nodes can send conflicting answers, so honest nodes need a clear majority even when f nodes lie and f others are slow or absent. This is why the standard rule is n of at least 3f + 1.

Formal Methods for Security (Version 1.0.1)

Systems Security · Difficulty: Hard · Key topics: specification, trace properties, hyperproperties, noninterference, Dolev-Yao, Tamarin, ProVerif, Isabelle/HOL, model checking, information flow control, policy analysis

This Knowledge Area presents formal methods as mathematically grounded ways to specify a system, the adversary and the properties it must meet. It argues that testing alone, the “penetrate and patch” approach, is not enough, and also discusses the limits of models. It then covers foundations: trace properties, hyperproperties such as noninterference, logics, and property checking with theorem provers, SAT and SMT solvers, static analysis and runtime verification. Application sections cover hardware and side channels, cryptographic protocols (symbolic, stochastic and computational methods), software and large systems (information flow control, cryptographic libraries, kernels, web applications, full-stack verification) and configuration of access control policies.

Key Points:

  • Trace properties include safety and liveness; hyperproperties such as noninterference (Goguen and Meseguer, 1982) compare pairs of executions.
  • The Dolev-Yao adversary controls the network; Tamarin and ProVerif check protocols symbolically, while EasyCrypt and CryptoVerif support game-based computational proofs.
  • Lowe found an attack on the Needham-Schroeder Public Key protocol using such methods in the 1990s.
  • Named tools include Isabelle/HOL, Coq, Z3, SPIN, Tamarin, ProVerif and EasyCrypt; Amazon uses TLA+ on security-critical components.
  • A security type system labels data as high or low and rejects assigning high data to low variables, including through an implicit flow in an if statement.
  • A verified result is only as good as its model: the adversary model and the specification can both be wrong, and Rice’s theorem limits what can be decided automatically.

Memory Tip: For any proof ask three things first: what is the system model, what is the adversary model, and what is the property. All three must be right for the result to mean something.

Common Mistake: Reading “formally verified” as “secure”. Verification only shows the system meets the stated specification against the stated attacker.

Important Questions:

  • Why is testing alone not enough for security? Tests show that expected behaviour works, but attackers look for unexpected behaviour. Formal methods can show the absence of whole classes of attack within a model, which testing cannot.
  • What is noninterference? A property saying that secret (high) inputs have no effect on public (low) outputs, so an observer of the public outputs learns nothing about the secrets.

Authentication, Authorisation & Accountability (Version 1.0.3)

Systems Security · Difficulty: Medium · Key topics: access control, reference monitor, DAC, MAC, RBAC, ABAC, XACML, Bell-LaPadula, Kerberos, SAML, OAuth, OpenID Connect, password hashing, audit logs, Certificate Transparency

The Knowledge Area warns that its three core terms are overloaded, so authorisation can mean setting rules or checking requests against them. Authorisation comes first: principals, subjects and objects, the reference monitor, access control matrices, ACLs and capabilities, discretionary and mandatory control, role-based and attribute-based access control, code-based control, mobile permissions, digital rights management, usage control, enforcement with XACML, and the formal models. Then it covers access control in distributed systems (tokens, federation, web origin policies), authentication (passwords, biometrics, tokens, protocols and formal properties) and accountability through protected audit logs.

Key Points:

  • The reference monitor mediates every access; XACML splits enforcement into PAP, PDP, PIP and PEP components.
  • Bell-LaPadula forbids reading up and writing down; Biba forbids reading down and writing up; Clark-Wilson and Chinese Wall address integrity procedures and conflicts of interest.
  • NIST RBAC has flat, hierarchical, constrained and symmetric variants, and separation of duties is a constraint on roles.
  • Web mechanisms include the same-origin policy, HttpOnly cookies, Content Security Policy and CORS headers.
  • Distributed protocols covered include Needham-Schroeder, Kerberos, SAML 2.0 and OAuth 2.0 with OpenID Connect; formal analysis uses TAMARIN and ProVerif.
  • Accountability relies on tamper-evident records: hash-chained logs, WORM storage and Certificate Transparency (RFC 6962).
  • Passwords should be salted and hashed, and NIST advice favours length over complexity rules and avoids forced expiry.

Memory Tip: Say it in order: Authentication asks who you are, Authorisation asks what you may do, Accountability asks what you did.

Common Mistake: Using authentication and authorisation as if they were the same. Proving who you are does not decide what you are allowed to do.

Important Questions:

  • What is the difference between authentication and authorisation? Authentication checks that a user or system is who it claims to be. Authorisation decides which actions that identity is allowed to perform on which objects.
  • What does a reference monitor do? It sits between subjects and objects and checks every access request against the security policy, so no access can bypass the check.

Secure Software Lifecycle (Version 2.0.0)

Software & Platform Security · Difficulty: Medium · Key topics: Microsoft SDL, Touchpoints, SAFECode, NIST SSDF, DevSecOps, software supply chain, SBOM, SLSA, AI code generation, SAMM, BSIMM, Cyber Resilience Act

This is the most heavily revised Knowledge Area in version 1.2. It argues against “penetrate and patch” and then compares three prescriptive secure development processes: Microsoft SDL, McGraw’s Touchpoints and the SAFECode practices, followed by the NIST Secure Software Development Framework (SSDF). Separate sections cover agile, DevOps and DevSecOps, software supply chain security, and new material on AI-generated code and AI-native systems. The second half adapts the lifecycle to mobile, cloud, IoT, road vehicles, payment cards and APIs, explains how to assess maturity with SAMM, BSIMM and the Common Criteria, summarises regulation such as the EU Cyber Resilience Act, and ends with advice on adopting a lifecycle in practice.

Key Points:

  • NIST SSDF (SP 800-218) has 19 practices and 34 tasks in four groups: prepare, protect, produce well-secured software and respond to vulnerabilities; SP 800-218A adds an AI profile with 21 practices.
  • McGraw’s Touchpoints has seven practices, including code review, architectural risk analysis, penetration testing, abuse cases and security requirements.
  • Supply chain attack vectors are malicious or vulnerable dependencies, compromised build infrastructure (SolarWinds) and social engineering; SLSA and Sigstore address provenance.
  • New AI risks include insecure generated code, “slopsquatting” (attackers registering package names that AI tools invent), prompt injection and the OWASP Top 10 for LLM Applications.
  • The EU Cyber Resilience Act requires security by design and by default and an SBOM for the top-level dependencies of products sold in the EU.
  • Assessment models include OWASP SAMM, BSIMM and the Common Criteria.

Memory Tip: Do not memorise three lifecycles separately. Place their practices on the same six phases, as the KA’s comparison does, and see where the models agree.

Common Mistake: Adopting a model as a checklist to follow whole. The KA advises starting with low-friction automated practices: static analysis in the IDE, dependency scanning in CI/CD and secrets detection at pre-commit.

Important Questions:

  • Why is “penetrate and patch” not enough? Fixing flaws after release is slow and expensive, breaches can go undetected for a long time, and patches themselves carry risk. A secure lifecycle tries to prevent and find defects before release.
  • What is an SBOM and why does it matter? A software bill of materials lists the components inside a product. It lets customers and regulators see vulnerable dependencies, which is why supply chain rules and the EU Cyber Resilience Act require it.

Web & Mobile Security (Version 1.0.2)

Software & Platform Security · Difficulty: Medium · Key topics: appification, webification, same-origin policy, CSP, app stores, permissions, Web PKI, cookies, clickjacking, SQL injection, XSS, CSRF, password storage

This Knowledge Area is about the place where web and mobile security meet. Its frame is two trends: appification, with many small apps, and webification, where apps and browsers share HTTP, HTML and JavaScript. Fundamentals cover app stores and vetting, sandboxing and the same-origin policy, permission dialogs, Web PKI and HTTPS, authentication, cookies, passwords and their alternatives, and frequent updates. Client-side topics include phishing, clickjacking, client-side storage and physical attacks. Server-side topics include SQL, command and file injection, cross-site scripting, CSRF, misconfiguration and vulnerable components such as Heartbleed.

Key Points:

  • The same-origin policy (1995) compares protocol, host and port; Content Security Policy restricts which origins may supply scripts and media.
  • Android separates normal and dangerous permissions, and runtime dialogs replaced install-time lists, but users get used to approving them.
  • SQL injection is prevented with prepared statements or an ORM, not with hand-written escaping; XSS needs input validation, output encoding and CSP.
  • CSRF is stopped by unpredictable per-session tokens; secret cookies and HTTPS alone do not stop it.
  • Basic HTTP authentication sends Base64, which is encoding, not encryption.
  • Store passwords hashed with Argon2 or PBKDF2 and a unique salt, and give database accounts minimal privileges.
  • The X-Frame-Options: DENY header stops clickjacking more reliably than frame-busting scripts.

Practice Tip: Open your browser’s developer tools on a site you can test, look at the response headers, and check for Content-Security-Policy, X-Frame-Options and Secure and HttpOnly flags on cookies.

Common Mistake: Treating HTTPS as the fix for every web problem. It protects data in transit, but it does not stop injection, XSS or CSRF.

Important Questions:

  • What is the difference between XSS and CSRF? XSS makes a victim’s browser run attacker script inside a trusted site. CSRF makes the victim’s browser send an unwanted request to a site where they are already logged in.
  • How do you prevent SQL injection? Use prepared statements or an ORM so that user input is passed as data and never as part of the query text, and run the database account with least privilege.

Software Security (Version 1.1.0)

Software & Platform Security · Difficulty: Medium · Key topics: memory errors, injection, race conditions, TOCTOU, API vulnerabilities, side channels, memory-safe languages, static analysis, fuzzing, stack canaries, CFI, ASLR

Software Security focuses on implementation vulnerabilities, bug classes that an attacker can trigger, and leaves human factors, deployment and web and mobile specifics to other areas. The first section describes the main categories: memory management errors (spatial and temporal), structured output generation (injection), race conditions, API vulnerabilities and side channels. The remaining sections follow prevention, detection and mitigation. Prevention uses language and API design, such as memory-safe languages and prepared statements. Detection compares static, dynamic and hybrid techniques, including taint analysis, fuzzing and symbolic execution. Mitigation limits exploitation of remaining bugs with canaries, NX, CFI, ASLR, sandboxing and trusted boot.

Key Points:

  • Spatial errors are out-of-bounds accesses; temporal errors are use-after-free; both occur in memory-unsafe languages such as C and C++.
  • A TOCTOU race happens when a condition is checked and then used and an attacker changes it in between.
  • Rice’s theorem forces a trade-off: sound tools miss no vulnerabilities but raise false alarms, while complete tools raise none but can miss real bugs.
  • Fuzzing can be black-box, grey-box or white-box, and dynamic symbolic execution is the main white-box technique.
  • Java and Rust prevent memory errors through runtime checks and ownership rules; garbage collection and ownership are two ways to get temporal memory safety.
  • Mitigations such as stack canaries, NX memory, CFI and ASLR do not remove bugs, they make exploitation harder.
  • The CVE list contained over 300,000 entries at the time of writing, and the classic injection example is the input John' --.

Practice Tip: Write a tiny C program with an out-of-bounds array write and compile it once normally and once with -fsanitize=address to see the bug reported at run time.

Common Mistake: Believing a mitigation fixes the vulnerability. A stack canary or ASLR only raises the cost of exploiting a bug that is still there.

Important Questions:

  • What is the difference between spatial and temporal memory errors? A spatial error reads or writes outside the bounds of an object. A temporal error uses memory after it has been freed or reused, for example use-after-free.
  • Why can no static analysis tool be both sound and complete? By Rice’s theorem, deciding non-trivial program properties is undecidable in general, so a tool must either accept some false alarms or miss some real vulnerabilities.

Hardware Security (Version 1.0.2)

Infrastructure Security · Difficulty: Hard · Key topics: root of trust, Y-chart, FIPS 140-2, Common Criteria, HSM, TPM, TrustZone, SGX, side-channel attacks, fault attacks, PUF, random number generation, hardware Trojans

The chapter organises hardware security by the abstraction layers of the Gajski-Kuhn Y-chart, asking at each layer what is trusted, what the threat model is and what designers do. After the design cycle it covers how hardware security is measured: FIPS 140-2 levels, Common Criteria, EMVCo and SESIP. It then describes secure platforms such as HSMs, smart cards and TPMs, hardware support for software security (isolation, attestation, TrustZone, protected module architectures such as SGX), and cryptographic hardware. Later sections explain side-channel and fault attacks with countermeasures, entropy sources including PUFs, and the trust problems in the hardware design process itself.

Key Points:

  • A TPM provides three roots of trust: measurement, storage and reporting; TPM 2.0 comes as discrete chips, integrated, firmware and software TPMs, with protection falling in that order.
  • FIPS 140-2 level 1 can be met in software, level 2 adds tamper evidence, level 3 tamper resistance and level 4 protection for physically unprotected environments.
  • TrustZone separates secure and normal worlds with a Non-Secure bit on bus transactions.
  • Simple power analysis can work from one trace, while differential power analysis may need tens of traces for unprotected devices and millions for protected ones.
  • The Boneh-DeMillo-Lipton fault attack shows that one faulty RSA-CRT signature can reveal the signing key.
  • Weak PUFs such as SRAM PUFs need error correction, and strong PUFs such as the arbiter PUF can be broken by machine learning.
  • Spectre, Meltdown and Foreshadow exploit micro-architectural behaviour, and performance optimisation is a major source of leakage.

Memory Tip: Group attacks into passive and active: side channels only listen (time, power, EM, cache), fault attacks interfere (glitching, laser, Rowhammer).

Common Mistake: Assuming hardware is fixed and trustworthy. The KA notes that hardware is becoming more “soft” (FPGA fabric) and that foundries and EDA tools can also be sources of Trojans.

Important Questions:

  • What is a root of trust? A component that the rest of the system must rely on and whose trustworthiness cannot itself be verified from outside, such as a TPM measurement root or a secure boot ROM.
  • What is the difference between a side-channel attack and a fault attack? A side-channel attack passively observes leakage such as timing, power or EM emissions. A fault attack actively disturbs the device, for example by glitching or laser injection, and uses the faulty result.

Physical Layer & Telecommunications Security (Version 1.0.2)

Infrastructure Security · Difficulty: Hard · Key topics: channel reciprocity, MIMO, jamming, spread spectrum, radio fingerprinting, distance bounding, compromising emanations, sensor spoofing, NFC, ADS-B, cellular security, GNSS

The Knowledge Area reviews security at the wireless physical layer, covering signals from a few hertz up to optical frequencies, mainly radio. It begins with schemes that use channel randomness for confidentiality, integrity and access control, such as key establishment from channel reciprocity, MIMO beamforming and friendly jamming. Then it covers jamming and anti-jamming spread spectrum, radio fingerprinting of devices, and distance bounding and secure positioning. Further sections cover compromising emanations and sensor spoofing, and then apply the ideas to specific technologies: NFC, aviation networks (ADS-B), cellular generations from GSM to 5G, and satellite navigation (GNSS).

Key Points:

  • Distance bounding protects against four frauds: distance fraud, mafia fraud, terrorist fraud and distance hijacking.
  • A timing error of 1 ns corresponds to about 30 cm, so tiny measurement errors destroy distance estimates.
  • Wim van Eck demonstrated compromising emanations against CRT monitors in 1985.
  • NFC works over a few centimetres, yet relay (wormhole) attacks can make devices believe they are much closer.
  • ADS-B aircraft messages are often unauthenticated and can be jammed, spoofed or injected; multilateration cross-checks position using time of arrival.
  • 2G authentication used 56 to 64-bit keys and the weak A3, A5 and A8 algorithms; 3G added 128-bit keys and mutual authentication.
  • Civilian GNSS signals are unauthenticated and their spreading codes are public, so spoofing is possible, and seamless takeover is among the strongest attacks.

Memory Tip: Remember the speed of light in everyday units: radio travels about 30 cm per nanosecond, which is why distance bounding is so sensitive.

Common Mistake: Trusting proximity. Because relay attacks can forward signals, a short-range technology like NFC does not by itself prove the device is physically close.

Important Questions:

  • What is distance bounding? A protocol in which a verifier measures round-trip time of challenge and response messages to put an upper limit on how far away a prover can be.
  • Why can GNSS be spoofed? Civilian GNSS signals carry no authentication and their spreading codes are public, so an attacker can transmit fake signals that a receiver accepts as real.

Cyber-Physical Systems Security (Version 1.0.2)

Infrastructure Security · Difficulty: Medium · Key topics: CPS, sensors and actuators, transduction attacks, Purdue model, ICS, Stuxnet, smart grid, vehicles, medical devices, IoT, remote attestation, regulation · Dated: the industry snapshot in the conclusion dates from 2018

This Knowledge Area defines cyber-physical systems as engineered systems that depend on integrating computation with physical components, and it focuses on sensing, control and actuation, the lower layers of the Purdue model. It explains why safety, reliability and robust control do not protect against deliberate attackers, and reviews real incidents. A section on crosscutting security groups defences into prevention, detection and mitigation, including physics-based detection. Next it applies them to industrial control systems, power grids, vehicles, robotics, medical devices and consumer IoT. The last section looks at policy: market incentives, regulation, procurement, cyber-insurance, cyber-conflict and industry standards.

Key Points:

  • Transduction attacks inject false signals through the physical environment rather than through software.
  • Maroochy Shire (2000): a former contractor used stolen SCADA software and radios to release over 750,000 gallons of sewage.
  • Stuxnet (2010) changed motor speeds on PLCs at the Natanz plant while reporting normal operation; the 2016 Ukraine attack used Industroyer, which crafted industrial protocol commands automatically.
  • Defence categories are prevention (zoning, firewalls, data diodes), detection (remote attestation, physics-based detection) and mitigation (resilient state estimation, safe-control monitors).
  • Remote attestation of embedded memory takes about 5 to 7 seconds, which limits its use in real-time systems.
  • After the August 2003 blackout the NERC power-system standards became mandatory; California SB-327 requires reasonable security for IoT devices.

Memory Tip: Remember the three defence verbs for CPS: prevent, detect, mitigate. Mitigation matters most here because a control system must keep working safely during an attack.

Common Mistake: Using normal IT reflexes on control systems. Scanning or patching a fragile legacy controller can interrupt the physical process, so safety comes first.

Important Questions:

  • What is a transduction attack? An attack that uses physical signals, such as acoustic or electromagnetic interference, to make a sensor report false readings, without compromising any software.
  • How did Stuxnet hide its effect from operators? It changed the speeds of motors at the plant while reporting normal values to the operators, so the control system’s own reporting hid the manipulation.

Network Security (Version 2.0.1)

Infrastructure Security · Difficulty: Medium · Key topics: CIA triad, Dolev-Yao attacker, Wi-Fi security, TLS 1.3, DNSSEC, BGP, IPsec, VPN, QUIC, Tor, firewalls, IDS, SDN, network access control

Network Security assumes familiarity with the layered Internet stack and starts with security goals (confidentiality, integrity, availability, authenticity, non-repudiation, anonymity) and attacker models such as Dolev-Yao, active or passive, on-path or off-path. It looks at how requirements differ for LANs, wireless, bus networks, peer-to-peer and SDN. The largest part walks down the stack: application layer (email security, HTTPS, DNS, NTP, Tor), transport layer (TLS 1.2 and 1.3, TCP and QUIC), Internet layer (IPsec, VPNs, NAT, BGP) and link layer (802.1X, ARP spoofing, Wi-Fi, CAN bus). It then covers firewalls, IDS, monitoring, SDN and NFV security, and network access control.

Key Points:

  • WEP used RC4 with a 24-bit IV that can be exhausted in under 2 hours on a fast link; WPA3 (2018) replaced the pre-shared key handshake with SAE and added forward secrecy.
  • TLS 1.3 needs one round trip, drops RSA key exchange, and allows 0-RTT resumption at the cost of forward secrecy and replay protection.
  • DNSSEC gives authenticity and integrity but no confidentiality; DNS over TLS and DNS over HTTPS add encryption.
  • TCP SYN cookies let a server avoid keeping state during a SYN flood by encoding it in the initial sequence number.
  • RPKI route origin validation filters bogus origins but cannot detect a malicious AS in the middle of a path; BGPsec addresses paths but has low deployment.
  • QUIC servers must limit responses to three times what they received before the client’s address is validated.
  • VLANs allow about 4096 segments, while VXLAN allows about 16 million.

Memory Tip: Walk the stack top to bottom and attach one example to each layer: PGP at application, TLS at transport, IPsec at Internet, 802.1X at link.

Common Mistake: Assuming DNSSEC hides your lookups. It only proves the answers are genuine; use DoT or DoH if you also want the queries encrypted.

Important Questions:

  • What does the Dolev-Yao attacker model assume? The attacker controls the network completely: it can read, block, modify and inject messages, but cannot break the cryptography itself.
  • Why is TLS 1.3 considered an improvement over TLS 1.2? It removes outdated algorithms and RSA key transport, always provides forward secrecy except in 0-RTT data, and completes the handshake in one round trip.

Applied Cryptography (Version 1.0.1)

Infrastructure Security · Difficulty: Medium · Key topics: authenticated encryption, key sizes, cryptographic agility, post-quantum cryptography, libraries and APIs, side channels, key management, PKI, TLS, secure messaging, contact tracing

Applied Cryptography is the practical companion to the Cryptography KA. Its introduction points out that standard algorithms rarely fail when used correctly; failures come from API misuse, poor key management, unsafe composition and side channels. Section 1 reviews algorithms, schemes and protocols such as hash functions, AES, MACs, authenticated encryption, RSA, Diffie-Hellman and signatures, plus key sizes, agility, standardisation and post-quantum cryptography. Section 2 covers implementation, libraries and API design. Section 3 covers the key life cycle, derivation, storage and PKI. Section 4 explains why people misuse cryptography, and Section 5 studies TLS, secure messaging and DP-3T contact tracing.

Key Points:

  • Encrypt-then-MAC is the sound composition; MAC-then-Encrypt and Encrypt-and-MAC have caused failures in TLS and SSH, and AES-GCM is used in about 90 percent of web TLS connections.
  • A 3072-bit RSA modulus gives roughly 128-bit security, the same level as a 256-bit elliptic curve; 256-bit security needs 15360-bit RSA against 512-bit ECC.
  • Cryptographic agility helps migration but invites downgrade attacks, as seen in SSL and TLS EXPORT cipher suites; WireGuard takes an opinionated, fixed design instead.
  • Shor’s algorithm threatens RSA and discrete-log schemes, while Grover’s algorithm is handled by doubling symmetric key sizes; NIST is running a post-quantum standardisation process.
  • The key life cycle runs from generation and distribution through use, update and revocation to archiving and secure deletion.
  • The DP-3T contact tracing design uses a chain of daily keys and 96 pseudo-random beacons a day, one for each 15-minute interval.
  • Homomorphic encryption (FHE) still costs 10^6 to 10^8 times more than plaintext computation.

Memory Tip: Remember the three states of data: in transit (TLS, messaging), at rest (disk and database encryption) and under computation (FHE, MPC).

Common Mistake: Designing your own cipher or scheme. The KA lists the “kitchen sink”, “large keys” and “friendly cryptanalysis” fallacies and says there is no substitute for independent analysis.

Important Questions:

  • Why is Encrypt-then-MAC preferred over MAC-then-Encrypt? Checking the MAC before decrypting stops an attacker from using error messages or padding behaviour as an oracle, which is the weakness that broke MAC-then-Encrypt designs in TLS and SSH.
  • What is cryptographic agility? The ability to switch algorithms or parameters without redesigning a system. It eases upgrades, but supporting old and new options together can enable downgrade attacks.

How to Study This Book

CyBOK is a set of 22 separate PDFs, not one book, so you do not read it from cover to cover. Start with the Introduction, because every Knowledge Area uses its vocabulary, and then pick the Knowledge Areas that match your course. Each document follows the same pattern: an introduction, a set of numbered sections, a conclusion and a cross-reference table that points to the textbooks and papers behind each topic. That table is the most useful part when you need to go deeper.

For a first Introduction to Cyber Security or Information Security course, read the Introduction, Human Factors, Risk Management & Governance, Malware & Attack Technologies and the application and transport layer parts of Network Security (HTTPS, DNS, TLS, firewalls). These are the most readable documents and need little background. Add Adversarial Behaviours if your course covers cyber crime, and Privacy & Online Rights for privacy topics.

For a cryptography or applied security course, read Cryptography before Applied Cryptography. The first assumes first-year undergraduate maths and gives definitions and constructions. The second shows how those pieces fail in real systems through libraries, API design, key management, TLS and messaging. For operating systems and software security, pair Operating Systems & Virtualisation Security with Software Security, then add Web & Mobile Security for web application courses and Secure Software Lifecycle for software engineering or DevSecOps.

Some Knowledge Areas are advanced. Formal Methods for Security, Hardware Security, Physical Layer & Telecommunications Security, Distributed Systems Security and Privacy & Online Rights expect that you already know the basics of the area. Read them for a final-year project, an elective or a research proposal, and use the section headings to take only what you need. Law & Regulation is the longest document at about 120 pages including references. It is built around UK, US and EU law, so use it to learn the concepts and not as a guide to Pakistani law.

Some content has aged. Most Knowledge Areas were first published between 2019 and 2021 and were re-issued in September 2026, mostly with layout, glossary and bibliography fixes. Secure Software Lifecycle (2.0.0) and Software Security (1.1.0) have real content updates, including supply chain and AI material. Check recent sources for examples such as international data transfer rules in the Law document, and for fast-moving topics such as post-quantum cryptography and AI-generated code.

CyBOK has no exercises, labs or tool tutorials. Make your own practice: draw a kill chain for a malware case study, write a STRIDE table for an app you have built, capture traffic with Wireshark and match it to the Network Security sections, and test a small web app for injection. Use the CyBOK sections as the reference for what to check and why. It also does not teach step-by-step hacking tools or prepare you for a specific certification exam.


Used In These Programs

This book is used for the Cyber Security Fundamentals course in: BS Cyber Security · BS Information Technology. Browse all Cyber Security books or all Computer Science books.

Who Should Read This

CyBOK is written for BS Cyber Security and BSIT students from the second year onwards, for lecturers who plan a course, and for self-learners who want an authoritative map of the field. It suits courses in introduction to cyber security, information security, network security, secure software development and digital forensics, and it is a good source when you write a final-year project proposal. It is not the right choice if you want a single textbook with exercises, hands-on hacking labs or a certification guide. Many Knowledge Areas assume that you already know operating systems and networking, so if those feel new, read Operating Systems and Networks or Operating Systems: Three Easy Pieces and Think OS first, and use Computer Networking: Principles, Protocols and Practice beside the Network Security document.


Applicable Universities

This book is useful for students at Pakistani universities offering BS Cyber Security or BS Information Technology, including Punjab University, Virtual University, COMSATS, FAST, UET, NUST, and other HEC-recognized institutions.

Official Copy of CyBOK

CyBOK is free from its official website. The Knowledgebase page lists the Introduction and all 21 Knowledge Areas, each as its own PDF, so you can pick the areas your course needs. Click the button below to open the official CyBOK Knowledgebase page.

→ Visit Official Page

FAQs

Is CyBOK free?

Yes. All 22 documents are free to download from the official CyBOK site, and they are published under the Open Government Licence v3.0. The licence allows you to copy, share and reuse the material, including for education, as long as you give the attribution line that the documents provide. Use the official site for downloads instead of copies posted elsewhere.

Which version is this?

This page covers CyBOK Version 1.2, released on 22 September 2026. Each Knowledge Area carries its own version number: for example the Introduction is 1.2.0, Secure Software Lifecycle is 2.0.0, Network Security is 2.0.1 and Software Security is 1.1.0. Earlier releases were Version 1.0 in October 2019 and Version 1.1 in July 2021.

How many chapters and exercises does it have?

There are 22 documents: one Introduction and 21 Knowledge Areas, in five categories (human, organisational and regulatory aspects; attacks and defences; systems security; software and platform security; infrastructure security). None of them has exercise sets or labs. Each ends with a cross-reference table to textbooks and papers that you can use for further study.

Is it good for complete beginners?

Partly. The Introduction, Human Factors, Risk Management and Governance and Malware and Attack Technologies are readable with little background. Others are much harder: Cryptography assumes first-year undergraduate maths, and Formal Methods, Hardware Security and Physical Layer Security are advanced. If you are new to the subject, start with the easy documents and keep a networking or operating systems textbook beside you.

Does it teach hacking tools or Pakistani cyber law?

No to both. CyBOK explains attacker behaviour, malware, analysis methods and defences in concept, but it has no step-by-step tool tutorials or lab guides. The Law and Regulation document uses UK, US and EU law as its examples and does not cover Pakistani law, so use it to learn legal concepts and read local law separately.

Is it still current in 2026?

Largely yes. The set was re-issued in September 2026, and Secure Software Lifecycle and Software Security have real content updates, including supply chain security and AI. Most other Knowledge Areas were first written between 2019 and 2021 and received editorial fixes, so check newer sources for fast-moving topics such as international data transfer rules, post-quantum cryptography and current attack campaigns.

Related Books

CyBOK gives you one free, citable map of cyber security, from law and human factors to cryptography and hardware, in 22 official PDFs. Browse more Computer Science books for the rest of your semester.

The Cyber Security Body of Knowledge (CyBOK), Version 1.2, by the CyBOK project. CyBOK © Crown Copyright, The National Cyber Security Centre 2026, licensed under the Open Government Licence v3.0: http://www.nationalarchives.gov.uk/doc/open-government-licence/. Official site: https://www.cybok.org/. This page links to the official site and does not host or copy the documents.

Leave a Comment