Skip to content
Paul Marinos
Menu

Cloud & Infrastructure Security

The substrate everything else runs on.

Every other pillar already touches cloud, but always through its own lens: who may act on it, what runs on it, how to break it, how to see it. None of them own the platform itself. This one does — how the substrate is architected, segmented, hardened, keyed, and kept in a known-good state.

That gap is worth naming because the most expensive cloud failures aren’t vulnerabilities in the usual sense. They’re structural: an account layout that made the blast radius the whole estate, an egress path nobody constrained, a KMS design that encrypted everything with a key any compute role could use.

Architecture decisions are the ones you can’t patch. A misconfigured dependency is a ticket; a flat network across three environments is a two-year programme. The controls here are also the ones that make every other pillar cheaper — segmentation shrinks what a pentest can reach, sane account structure shrinks what an IAM mistake can touch, and a real posture baseline shrinks what detection has to compensate for.

The honest counterweight: posture tooling is very good at producing findings and much worse at producing prioritised ones. A CSPM that reports ten thousand issues has told you nothing, which is why this pillar leans on risk prioritization rather than restating it.

  • IAM: the two-plane split. §2 owns identity policy — evaluation logic, roles, federation. This pillar owns org-level guardrail architecture: SCPs and Azure Policy as blast-radius design rather than as permission checks.
  • Application Security: §3 is the workload, this is what it runs on. IaC and container scanners live there; the posture programme they feed lives here.
  • Penetration Testing & Red Teaming: segmentation, egress control and workload isolation are exactly what an operator tests. Cloud pivots are architecture findings before they are exploits.
  • Detection Engineering: architecture determines what is observable. Network design decides whether east-west traffic is even loggable, which makes §8.2 upstream of §7.3’s coverage gaps.
  • GRC: most technical control requirements land here — encryption, segmentation, backup, key management. “Data is encrypted at rest” is an audit answer; §8.4 is what it does and doesn’t buy.
  • Threat Intelligence: posture findings are a prioritization problem before they are a remediation problem.
  • AI & Automation: drift detection and closed-loop remediation are the natural automation target, and model-serving infrastructure is just another workload with an unusual blast radius.
  • Incident Response & Digital Forensics: cloud IR is an architecture dependency. Snapshotting, credential revocation and blast-radius scoping all rest on how the substrate was built.
  • Data Security & Privacy Engineering: §8.4 owns key material and primitives; §10 owns the data itself. Encryption is the mechanism, not the control.
Subsection Focus
Cloud Architecture & Landing Zones Multi-account structure as blast-radius design, Organizations/Control Tower, management groups and folders, environment separation, org-level guardrails, landing zones as code
Network Security & Segmentation VPC/VNet design and subnet tiering, private endpoints, egress control and DNS security, microsegmentation, mTLS and service mesh, ZTNA vs. VPN
Workload & Container Security Minimal images and provenance, admission control, Kubernetes hardening, runtime security and isolation, serverless and managed-service posture, host hardening
Cryptography & Key Management KMS/HSM design and key hierarchies, envelope encryption, what encryption actually defends against, secrets architecture, PKI lifecycle, post-quantum and crypto agility
Posture Management & IaC What CSPM/CNAPP genuinely catches, IaC security, drift detection and closed-loop remediation, benchmarks as baselines, keeping the findings queue finite
Resilience & Recovery Backup and restore as a security control, immutable and air-gapped backups, restore testing as the only real evidence, DDoS and edge protection, multi-region failover

Graph View