RMFInsider

CISSP Domain 3, Security Architecture and Engineering, is about 13% of the exam and covers secure design principles, security models, cryptography, physical security, and the vulnerabilities baked into system architectures. If you work DoD systems, you already touch this domain constantly: every time you argue about FIPS-validated crypto, read a STIG that disables a legacy cipher, or draw an authorization boundary, you are doing Domain 3 work. The exam just wants you to know the vocabulary and the reasoning behind it.

This is the third spoke in the domain series. Start with the all-eight-domains overview if you want the map, then Domain 1 and Domain 2.

What Domain 3 actually covers

• Secure design principles: least privilege, defense in depth, fail securely, separation of duties, keep it simple, zero trust, privacy by design, secure defaults.
• Security models: Bell-LaPadula, Biba, Clark-Wilson, Brewer-Nash, and the reference monitor concept.
• Security evaluation criteria: Common Criteria (EALs, Protection Profiles), and where FIPS 140 fits.
• Security capabilities of systems: TPMs, HSMs, memory protection, virtualization, trusted platform boot.
• Vulnerabilities in architectures: client and server, databases, cloud, IoT, ICS/SCADA, embedded, microservices, containers, serverless, high-performance computing.
• Cryptography: symmetric, asymmetric, hashing, digital signatures, PKI, key management, and the attacks against all of it.
• Physical security: site design, facility controls, power, HVAC, fire suppression.

The security models, translated into DoD terms

The models are where practitioners lose points, because nobody at work says “Bell-LaPadula” out loud. They do say “no read up, no write down,” which is the same thing. Here is the cheat sheet that sticks:

• Bell-LaPadula is confidentiality. Simple property: no read up. Star property: no write down. That is the classification system you already live under. A Secret user cannot read Top Secret, and cannot write Secret data into an Unclassified document.
• Biba is integrity, and it is Bell-LaPadula flipped: no read down, no write up. Think of a trusted build server that should never pull code from an untrusted source.
• Clark-Wilson is integrity through well-formed transactions and separation of duties. Users touch data only through approved programs. That is every financial and logistics system with an audit trail.
• Brewer-Nash (Chinese Wall) prevents conflicts of interest. Access to one client’s data blocks access to a competitor’s. Consulting firms and support contractors on multiple programs.
• Reference monitor is the abstract idea of a component that mediates every access. The security kernel is the implementation. If a question asks what must be tamper-proof, always invoked, and small enough to verify, that is the answer.

Cryptography: what the exam tests versus what you configure

At work, cryptography is a STIG check: is TLS 1.2 or later enforced, is the module FIPS 140 validated, is SHA-1 disabled. The exam goes one layer under that and asks why.

Symmetric versus asymmetric

Symmetric (AES, ChaCha20) is fast and uses one shared key; the problem is getting that key to the other side. Asymmetric (RSA, ECC, Diffie-Hellman) solves key distribution with public/private pairs but is slow. Real systems use asymmetric to exchange a symmetric session key, then encrypt bulk data symmetrically. That hybrid is what TLS does, and a surprising number of exam questions are that fact in disguise.

Hashing and signatures

A hash (SHA-256, SHA-3) is a one-way fingerprint that proves integrity. A digital signature is a hash encrypted with the sender’s private key; anyone with the public key can verify who signed it and that nothing changed. HMAC adds a shared secret to a hash for integrity plus authentication without a PKI. Your CAC does signatures every day, which makes this the easiest sub-topic to anchor to real life.

PKI and key management

Certificates bind a public key to an identity, signed by a certificate authority. Know the lifecycle (generate, distribute, store, rotate, revoke, destroy), know what CRLs and OCSP do, and know that key escrow and recovery are legitimate organizational needs, not just something the exam invented. The DoD PKI, with its root and intermediate CAs and the CRL sizes that make every admin sigh, is the reference implementation.

Attacks worth memorizing

Brute force, dictionary, rainbow tables (defeated by salting), birthday attacks against hashes, known-plaintext and chosen-plaintext, man-in-the-middle, replay, side-channel (timing, power analysis), and downgrade attacks. If a question describes an attacker forcing an old cipher suite, that is a downgrade attack, and the fix is disabling legacy protocols, which is a STIG finding you have closed before.

FIPS 140 and Common Criteria: the DoD angle

FIPS 140-3 validates cryptographic modules at four security levels. Level 1 is basic algorithm correctness; Level 2 adds tamper evidence and role-based authentication; Level 3 adds tamper resistance and identity-based authentication; Level 4 is the full envelope-of-protection tier. DoD policy requires FIPS-validated modules for protecting sensitive data, which is why “FIPS mode” is a real setting with real consequences on your servers.

Common Criteria (ISO 15408) evaluates products against a Protection Profile and assigns an Evaluation Assurance Level, EAL1 through EAL7. The exam loves to ask what the EAL measures. Answer: the rigor of the evaluation, not how secure the product is in your environment. A product at EAL4 configured badly is still a finding.

Architecture vulnerabilities you already document

Domain 3 asks you to spot weaknesses in system types. For DoD candidates, nearly all of these map to a section in your SSP or a control family:

• Cloud: shared responsibility, multi-tenancy, data remanence. See FedRAMP vs DoD RMF for cloud.
• ICS/SCADA: legacy protocols with no authentication, long patch cycles, availability over confidentiality. Segmentation is the answer to nearly every ICS question.
• Virtualization and containers: hypervisor escape, image trust, shared kernel.
• Databases: aggregation and inference (combining unclassified facts into a classified conclusion). This one is DoD-specific enough that it should feel familiar.
• Embedded and IoT: no update path, default credentials, physical access.

The boundary you draw in your authorization boundary diagram is the practical output of the architecture thinking this domain tests.

Physical security, briefly

Crime Prevention Through Environmental Design, layered perimeters, mantraps, badge plus PIN, fire classes and suppression types (water for Class A, gas for electrical spaces), and the fact that the only acceptable reason to defeat a fire suppression system is human life. If you have ever run a SCIF checklist or argued with facilities about a door closer, you can bank these points.

How to study Domain 3 without drowning

It is the longest chapter in every study guide, so do not read it linearly. Spend one session on the models until you can explain each in one sentence. Spend two sessions on cryptography, drawing the hybrid TLS handshake from memory. Skim the architecture vulnerabilities and physical sections, since your job already covers them. Then drill questions and read the explanations for every miss. The 90-day study plan allocates the calendar for you, and if you are unsure whether you are close, this readiness check will tell you.

Next in the series: Domain 4, Communication and Network Security, which is the domain where knowing your boundary diagram pays off twice.