Overview
Security Architecture and Engineering covers how security is built into systems from the first requirement to retirement. Its 10 sub-objectives span secure design principles such as least privilege, zero trust and secure access service edge (SASE); security models including Bell-LaPadula and Biba; control selection; hardware security capabilities like the Trusted Platform Module (TPM); vulnerabilities across client, server, cloud, IoT, container and embedded systems; cryptographic solutions and key management; cryptanalytic attacks; site and facility design and controls; and the information system lifecycle.
Candidates must explain which mechanism enforces which property: a model enforcing confidentiality versus one enforcing integrity, a hash versus a signature, symmetric speed versus asymmetric key distribution. The domain is the engineering backbone for Domain 4 network design, Domain 5 authentication and Domain 8 secure development, and it implements the risk-based control choices from Domain 1.
How it's tested
Questions mix conceptual identification (which model, which principle, which attack) with design decisions (which algorithm, which control, which facility measure). Many hinge on one keyword: 'no read up' signals confidentiality, 'no write up' signals integrity, 'sign' implies the sender's private key. Read the property the scenario is protecting before evaluating options.
-
Model and principle matching
A rule or design is described, and you name the security model or design principle it implements, such as Biba's simple integrity property or fail securely.
-
Which key, which algorithm
A requirement for confidentiality, integrity, authenticity or nonrepudiation is stated, and you choose the key or primitive that satisfies it.
-
Physical design choice
A facility scenario asks for the right fire suppression, power protection, or placement of a data center or wiring closet.
Alice must send Bob a contract that only Bob can read and that Bob can prove came from Alice. Encrypting with Bob's public key gives confidentiality; signing a hash with Alice's private key gives authenticity, integrity and nonrepudiation. An option offering only encryption with Bob's public key fails the proof requirement, and one using Alice's public key fails both, because anyone holds it.
Sub-objectives 10
Official sub-objectives from the ISC2 outline, in outline order.
3.1 Research, implement, and manage engineering processes using secure design principlesThreat modeling, least privilege, defense in depth, secure defaults, fail securely, SoD, simplicity, zero trust, privacy by design, shared responsibility and SASE.
- What it means
- Secure design principles guide engineering decisions so security is inherent rather than bolted on. The outline lists threat modeling, least privilege, defense in depth, secure defaults, fail securely, segregation of duties, keep it simple and small, zero trust or trust but verify, privacy by design, shared responsibility and secure access service edge.
- What to understand
- Zero trust removes implicit trust based on network location and verifies every request using identity, device posture and context. Fail securely means a failure leaves the system in a secure state, but life-safety systems such as fire exit doors fail open. Shared responsibility divides duties between cloud provider and customer by service model.
- How it shows up
- A design is described, such as a firewall that blocks all traffic when its rule engine crashes, and you identify the principle it follows.
- Common confusion
- Treating zero trust as a product; it is an architecture principle built from identity, segmentation, policy enforcement and continuous verification.
3.2 Understand the fundamental concepts of security models (e.g., Biba, Star Model, Bell-LaPadula)Formal models for confidentiality and integrity.
- What it means
- Bell-LaPadula enforces confidentiality: no read up (simple security property) and no write down (star property). Biba enforces integrity: no read down (simple integrity) and no write up (star integrity). Clark-Wilson enforces integrity through well-formed transactions and separation of duties; Brewer-Nash prevents conflicts of interest.
- What to understand
- The 'star' rules govern writing in both Bell-LaPadula and Biba, and each model protects only one property. Bell-LaPadula stops secrets leaking downward but does nothing to prevent a low-level user corrupting high-level data, which is Biba's job.
- How it shows up
- A rule such as 'a subject cannot write to a lower classification' is given and you name the model and property.
- Common confusion
- Reversing the directions; remember confidentiality worries about information flowing down, integrity worries about untrusted data flowing up.
3.3 Select controls based upon systems security requirementsChoosing controls from requirements, evaluation criteria and risk.
- What it means
- Controls are selected from documented security requirements, which derive from risk assessment, classification, legal obligations and the system's mission. Evaluation schemes such as the Common Criteria express assurance through protection profiles, security targets and evaluation assurance levels (EAL1 to EAL7).
- What to understand
- Requirements come first, then controls; buying a product and then writing requirements around it reverses the process. A Common Criteria rating states how rigorously a product was evaluated against its security target, not that it is secure in your environment.
- How it shows up
- A team proposes a product before defining requirements, and the BEST answer is to document security requirements derived from risk and classification first.
- Common confusion
- Treating a higher EAL as proof of more security features; EAL measures assurance depth, not functionality.
3.4 Understand security capabilities of Information Systems (e.g., memory protection, Trusted Platform Module (TPM), encryption/decryption)Hardware and operating system protections.
- What it means
- Systems protect themselves with memory protection and process isolation, protection rings, the reference monitor implemented by a security kernel within the trusted computing base, TPM chips for measured boot and sealed keys, hardware security modules, and hardware-assisted encryption.
- What to understand
- The reference monitor must be tamperproof, always invoked and small enough to verify. A TPM stores keys bound to a specific device and boot state, so disk encryption keys are released only if the platform is unmodified; an HSM is a separate, often network-attached device for high-volume key operations.
- How it shows up
- A laptop's disk must not decrypt if the bootloader is tampered with, pointing to TPM-sealed keys with measured boot.
- Common confusion
- Confusing a TPM (device-bound chip for that system) with an HSM (dedicated key-management appliance for many systems).
3.5 Assess and mitigate the vulnerabilities of security architectures, designs, and solution elementsWeaknesses in client, server, database, cryptographic, OT/ICS, cloud, distributed, IoT, microservice, container, serverless, embedded, HPC, edge and virtualized systems.
- What it means
- Each architecture has characteristic weaknesses: databases face aggregation and inference, ICS prioritizes availability and runs legacy protocols, containers share a host kernel, serverless shifts patching to the provider but expands function permissions, and virtualization risks VM escape and sprawl.
- What to understand
- Aggregation combines individually harmless data into something sensitive; inference deduces sensitive facts from accessible data. OT environments often cannot patch quickly, so segmentation and monitoring are primary mitigations.
- How it shows up
- A user with access to individual unclassified troop movements deduces a classified operation, and you identify inference or aggregation.
- Common confusion
- Applying IT patching cadences to ICS; safety and availability usually require tested, scheduled changes and compensating controls.
3.6 Select and determine cryptographic solutionsCryptographic lifecycle, methods, PKI, key management and digital signatures and certificates.
- What it means
- This covers the cryptographic lifecycle (key and algorithm selection through retirement), symmetric, asymmetric, elliptic curve and quantum methods, public key infrastructure, key management practices such as rotation, and digital signatures and certificates for nonrepudiation and integrity.
- What to understand
- Symmetric algorithms such as AES are fast but require secure key distribution and n(n-1)/2 keys for n users; asymmetric algorithms such as RSA and ECC solve distribution with 2n keys but are slower, so hybrid systems use asymmetric exchange to share a symmetric session key. To sign, the sender uses their own private key; to encrypt for a recipient, use the recipient's public key.
- How it shows up
- A requirement for nonrepudiation leads to a digital signature, while a requirement for bulk encryption of a large database leads to AES.
- Common confusion
- Believing hashing provides confidentiality or that encryption alone provides nonrepudiation.
3.7 Understand methods of cryptanalytic attacksBrute force, ciphertext-only, known and chosen attacks, frequency analysis, implementation, side-channel, fault injection, timing, MITM, pass the hash, Kerberos exploitation and ransomware.
- What it means
- Cryptanalytic attacks target algorithms (brute force, frequency analysis), the attacker's knowledge (ciphertext only, known plaintext, chosen plaintext or ciphertext), implementations (side-channel, timing, fault injection) or protocols and credentials (man-in-the-middle, pass the hash, Kerberos ticket attacks). Ransomware abuses cryptography offensively.
- What to understand
- Side-channel attacks extract keys from physical leakage such as power or timing without breaking the algorithm. Pass the hash reuses a captured NTLM hash without cracking it, so stronger passwords alone do not stop it; Kerberos golden ticket attacks forge tickets using the krbtgt account's key.
- How it shows up
- An attacker measures how long a smart card takes to process operations and recovers its key, which identifies a timing side-channel attack.
- Common confusion
- Thinking longer keys stop side-channel or pass-the-hash attacks; these bypass the math entirely.
3.8 Apply security principles to site and facility designCrime prevention through environmental design and secure site selection.
- What it means
- Facility design applies security principles to location and layout: site selection away from flood plains and hazards, crime prevention through environmental design (CPTED) using natural surveillance, natural access control and territorial reinforcement, and layered physical defense.
- What to understand
- Critical rooms such as data centers belong in the core of the building, not on top floors, basements or exterior walls. Good design reduces reliance on guards by making unauthorized approach visible and difficult.
- How it shows up
- A new data center plan places servers in a basement near water pipes, and you identify the design flaw and better location.
- Common confusion
- Treating CPTED as fences and guards; it uses landscaping, lighting, sightlines and layout to deter.
3.9 Design site and facility security controlsWiring closets, data centers, media and evidence storage, work areas, HVAC, environment, fire and power.
- What it means
- This covers wiring closets and intermediate distribution frames, server rooms, media and evidence storage, restricted work areas, utilities and HVAC, environmental issues such as natural and man-made disasters, fire prevention, detection and suppression, and redundant and backup power.
- What to understand
- Fire classes drive suppression: Class A (common combustibles), B (liquids), C (electrical), D (metals) and K (cooking oils). Pre-action sprinklers are preferred in data centers because they need a detector alarm before water fills the pipes. A UPS covers short outages; generators cover longer ones.
- How it shows up
- A server room needs water-based suppression with minimal accidental discharge risk, pointing to a pre-action system.
- Common confusion
- Choosing wet-pipe sprinklers for a data center, or forgetting that halon is no longer produced under the Montreal Protocol and has been replaced by clean agents such as FM-200 and inert gases.
3.10 Manage the information system lifecycleFrom stakeholder requirements through operations to retirement and disposal.
- What it means
- The outline lists phases: stakeholder needs and requirements, requirements analysis, architectural design, development and implementation, integration, verification and validation, transition and deployment, operations and maintenance or sustainment, and retirement or disposal.
- What to understand
- Security requirements must enter at the first phase, because fixing design flaws later costs far more. Verification asks whether the system was built right (meets specification); validation asks whether the right system was built (meets stakeholder needs).
- How it shows up
- A question asks at which phase security requirements should be defined, or what must happen to data during retirement.
- Common confusion
- Swapping verification and validation, or treating retirement as only hardware disposal and forgetting data sanitization and records retention.
High-yield concepts
The terms, rules and formulas this domain turns into questions most often.
-
Bell-LaPadula
A confidentiality model with no read up (simple security property), no write down (star property) and a strong star property restricting read and write to the subject's own level.
- Remember
- It prevents secrets leaking to lower levels but does not protect integrity.
- Tested as
- Model identification from a described rule.
- Trap
- Choosing it when the scenario is about preventing corruption of data.
-
Biba
An integrity model with no read down (simple integrity) and no write up (star integrity).
- Remember
- It is the mirror image of Bell-LaPadula, applied to integrity.
- Tested as
- Integrity-focused model questions.
- Trap
- Reversing the arrows because the rules mirror Bell-LaPadula.
-
Clark-Wilson
An integrity model that forces users to modify constrained data items only through well-formed transformation procedures, with separation of duties and auditing.
- Remember
- Subjects access objects through programs (the access triple), never directly.
- Tested as
- Commercial integrity scenarios such as banking transactions.
- Trap
- Choosing Biba when the scenario emphasizes transactions and separation of duties.
-
Symmetric key count
Pairwise symmetric communication among n users requires n(n-1)/2 keys; asymmetric requires 2n keys.
- Remember
- For 50 users: 1,225 symmetric keys versus 100 asymmetric keys.
- Tested as
- Key-management scaling calculations.
- Trap
- Forgetting to divide by two.
-
Digital signature
A hash of the message encrypted with the sender's private key, verified with the sender's public key.
- Remember
- Signatures provide integrity, authenticity and nonrepudiation but not confidentiality.
- Tested as
- Which-key and which-property questions.
- Trap
- Thinking the recipient's key is used to sign.
-
Hybrid cryptography
Asymmetric cryptography exchanges or protects a symmetric session key, which then encrypts bulk data.
- Remember
- TLS uses this pattern, typically with ephemeral Diffie-Hellman for forward secrecy.
- Tested as
- Explains why protocols use both types.
- Trap
- Assuming RSA encrypts the full data stream.
-
PKI and certificate revocation
A certificate authority binds a public key to an identity in an X.509 certificate; revoked certificates are published through CRLs or checked in real time with OCSP.
- Remember
- A compromised private key requires immediate revocation, not just reissue.
- Tested as
- PKI trust and compromise scenarios.
- Trap
- Believing certificate expiry handles compromise quickly enough.
-
Reference monitor and TCB
The reference monitor is the abstract concept that mediates all subject-object access; the security kernel implements it within the trusted computing base.
- Remember
- It must be tamperproof, always invoked and verifiable.
- Tested as
- System security capability questions.
- Trap
- Confusing the concept (reference monitor) with its implementation (security kernel).
-
Side-channel attack
Recovering secrets from physical effects such as timing, power consumption or electromagnetic emanation during cryptographic operations.
- Remember
- It attacks the implementation, so key length does not help; constant-time code and shielding do.
- Tested as
- Cryptanalytic attack identification.
- Trap
- Choosing a stronger algorithm as the mitigation.
-
Pass the hash
Authenticating with a stolen password hash instead of the plaintext password.
- Remember
- Mitigate with credential isolation, restricting privileged logons and disabling NTLM where possible, not with password complexity.
- Tested as
- Credential attack scenarios.
- Trap
- Selecting longer passwords as the fix.
-
Aggregation vs. inference
Aggregation combines many lower-sensitivity items into a more sensitive whole; inference deduces a sensitive fact from available information.
- Remember
- Polyinstantiation and cell suppression help counter inference in databases.
- Tested as
- Database vulnerability questions.
- Trap
- Labeling the act of deducing a fact as aggregation.
-
Fire classes and suppression
A: common combustibles; B: flammable liquids; C: electrical; D: combustible metals; K: cooking oils and fats.
- Remember
- Never use water on energized electrical (C) fires; pre-action sprinklers suit data centers.
- Tested as
- Facility control questions.
- Trap
- Choosing wet-pipe sprinklers for server rooms.
-
Verification vs. validation
Verification checks the system meets its specification; validation checks it meets stakeholder needs in its intended use.
- Remember
- Built right versus right thing built.
- Tested as
- System lifecycle questions.
- Trap
- Swapping the two.
Side by side
Concepts candidates mix up, compared on the point that decides the answer.
Bell-LaPadula vs. Biba vs. Clark-Wilson
| Concept | Meaning | Key difference | Common trap |
|---|---|---|---|
| Bell-LaPadula | State machine model for confidentiality | No read up, no write down | Using it to answer an integrity question |
| Biba | Lattice model for integrity | No read down, no write up | Reversing the read and write directions |
| Clark-Wilson | Commercial integrity through well-formed transactions | Access only through certified programs, with separation of duties | Missing it when the scenario emphasizes transactions rather than levels |
Symmetric vs. asymmetric vs. hashing
| Concept | Meaning | Key difference | Common trap |
|---|---|---|---|
| Symmetric (AES) | One shared secret key encrypts and decrypts | Fast; key distribution is the hard problem | Claiming it provides nonrepudiation |
| Asymmetric (RSA, ECC) | Public and private key pair | Solves key distribution and enables signatures, but is slower | Using it for bulk data encryption |
| Hashing (SHA-256) | One-way fixed-length digest | Integrity only; no key and no decryption | Calling hashing encryption |
TPM vs. HSM
| Concept | Meaning | Key difference | Common trap |
|---|---|---|---|
| Trusted Platform Module | Chip on a device motherboard | Binds keys to one device and its measured boot state | Expecting it to serve keys to many servers |
| Hardware security module | Dedicated tamper-resistant appliance or card | High-volume key generation, storage and cryptographic operations for many systems | Choosing it to protect a single laptop's boot integrity |
Wet pipe vs. dry pipe vs. pre-action sprinklers
| Concept | Meaning | Key difference | Common trap |
|---|---|---|---|
| Wet pipe | Pipes always full of water | Fastest discharge, highest accidental-leak risk | Choosing it for a data center |
| Dry pipe | Pipes hold pressurized air until a head opens | Suits freezing environments; slower discharge | Thinking it prevents discharge from a single damaged head |
| Pre-action | Dry pipes fill only after a detector alarm, then a head must open | Two triggers reduce accidental discharge | Confusing it with deluge, which opens all heads |
Exam traps
Each trap is an answer that looks right. Learn the clue that should change your mind.
-
Wrong model for the property
- What candidates pick
- Bell-LaPadula for a requirement to stop low-trust processes corrupting high-trust data.
- Why it looks right
- It is the best-known model.
- Why it's wrong
- Bell-LaPadula protects confidentiality only; preventing corruption is integrity, which is Biba or Clark-Wilson.
Clue to watch for Words such as corrupt, modify, tamper or trustworthy data.
-
Encryption equals proof
- What candidates pick
- Encrypting a message with the recipient's public key to prove who sent it.
- Why it looks right
- Encryption feels like the strongest protection.
- Why it's wrong
- Anyone can use the recipient's public key, so it proves nothing about the sender; only a signature with the sender's private key does.
Clue to watch for Requirements such as prove origin, cannot deny or nonrepudiation.
-
Bigger keys fix everything
- What candidates pick
- Upgrading to a longer key after a timing attack.
- Why it looks right
- Longer keys resist brute force.
- Why it's wrong
- Timing attacks exploit implementation leakage; constant-time code or hardware countermeasures are needed.
Clue to watch for The attacker measures time, power or emissions.
-
Water everywhere
- What candidates pick
- Wet-pipe sprinklers for a server room.
- Why it looks right
- They are reliable and common.
- Why it's wrong
- A damaged head immediately floods equipment; pre-action systems add a detection trigger and are preferred.
Clue to watch for Data center or equipment-dense room with water suppression required.
-
Product first
- What candidates pick
- Selecting a highly rated product and then defining requirements.
- Why it looks right
- A high Common Criteria EAL suggests strong security.
- Why it's wrong
- Controls must be chosen from security requirements; an EAL reflects evaluation rigor against the product's own target, not your needs.
Clue to watch for The scenario mentions a product before any requirements exist.
-
Patch ICS like IT
- What candidates pick
- Applying patches immediately to an industrial control system.
- Why it looks right
- Fast patching is best practice in IT.
- Why it's wrong
- OT prioritizes safety and availability; changes must be tested and scheduled, with segmentation as the primary near-term control.
Clue to watch for SCADA, PLC, manufacturing line or safety system.
Answer framework
For Domain 3 design and cryptography scenarios, identify the property at stake before evaluating any mechanism.
- Name the security property the scenario needs: confidentiality, integrity, availability, authenticity or nonrepudiation, since every mechanism protects only some of them.
- Map that property to the mechanism: encryption, hashing, signature, model, redundancy or physical control.
- For cryptography, decide whose key is used: the recipient's public key to encrypt, the sender's private key to sign.
- Check the environment's constraints, such as OT safety, cloud shared responsibility, flash media or life safety, which can change the preferred control.
- Prefer answers that build security in early in the lifecycle and that follow design principles like least privilege, defense in depth and fail securely over late add-ons.
Practice question
Want more like this? The CISSP practice test has questions for every domain.
Checklist
0 of 13
Tick each one you can do without notes. Your ticks stay in this browser.
Study strategy
Study this domain in three blocks rather than front to back. Start with design principles and security models (3.1 to 3.4), because they supply the vocabulary for everything else. Then spend the largest block on cryptography (3.6 and 3.7): drill which key does what until it is automatic, then the attack types and their mitigations. Finish with architectures, facilities and the lifecycle (3.5, 3.8 to 3.10).
High-priority concepts are the model rules, hybrid cryptography, digital signatures, PKI, and aggregation versus inference. Commonly underestimated areas are facility controls, where fire classes and sprinkler types give reliable points, and the outline's broad list of architectures in 3.5, which now includes serverless, microservices, edge and high-performance computing.
Do not memorize algorithm trivia beyond what scenarios need; know that AES is the symmetric standard, RSA and ECC are asymmetric, SHA-2 and SHA-3 are current hashes, and MD5 and SHA-1 are deprecated. After this domain, practice Domain 4 network protocols and Domain 5 authentication questions, which repeatedly apply the cryptography learned here.
FAQ
Is Domain 3 the hardest CISSP domain?
Many candidates find it the densest because it combines models, cryptography, architecture and physical security. It is weighted at 13%, so it deserves a large share of study time without crowding out Domain 1.
How much cryptography math do I need?
Very little. You need key-count formulas, which key performs which operation, the properties each primitive provides and why hybrid systems exist. You do not need to compute RSA or elliptic curve operations.
What is the Star Model in the outline?
The outline lists it alongside Biba and Bell-LaPadula as an example security model. In study materials, the star (*) properties are the write rules of those two models: no write down for Bell-LaPadula and no write up for Biba.
What is the difference between zero trust and defense in depth?
Defense in depth layers multiple independent controls so one failure is not catastrophic. Zero trust removes implicit trust from network location and verifies each access request; it is often implemented as one of those layers.
Are physical security topics really tested?
Yes. The outline includes site and facility design (3.8) and controls (3.9) covering wiring closets, data centers, media and evidence storage, HVAC, fire and power, and Domain 7 adds physical security operations.
How do Domains 3 and 8 differ?
Domain 3 covers architecture and engineering of systems as a whole, including hardware, cryptography and facilities. Domain 8 focuses on the software development lifecycle, development ecosystems, testing and secure coding.
How long should I study Domain 3?
Our editorial planning estimate is 30 to 35 hours within a 250-hour plan. Allow more if cryptography is new to you.
Other Cybersecurity certifications
Where the same subject shows up in other exams. Cards marked "Same subject" open the matching domain.
- CompTIA SecurityX (CASP+) Same subject: Domain 2: Security Architecture (27%)
- CompTIA Security+ Same subject: Domain 3: Security Architecture (18%)
- CompTIA Cybersecurity Analyst+ Study guide
- Microsoft Security Operations Analyst Associate Study guide
- Certified Ethical Hacker Study guide
- Penetration Tester Study guide
- Systems Security Certified Practitioner Study guide
- Certified in Risk and Information Systems Control Study guide
- CompTIA PenTest+ Study guide
- Azure Security Engineer Associate Study guide
- CISM (Certified Information Security Manager) Study guide
- Certified Information Systems Auditor Study guide
- Certified Cloud Security Professional Study guide
Test your Domain 3 knowledge
You have covered secure design principles, security models, cryptography, cryptanalytic attacks and facility controls. Check whether you can pick the right model, key and control under time pressure.