Evaluating Architectural Security Controls
What you'll learn: How to systematically assess whether your architectural security controls actually reduce the threats you've identified to an acceptable level.
The Core Idea
After identifying threats through models like STRIDE and attack trees, you need to evaluate whether your chosen security controls (existing or planned) effectively mitigate those threats. This isn't just checking if controls exist—it's determining if they're sufficient, correctly placed, and actually working within your architecture.
Think of it like evaluating flood defenses for a city. You've identified that rivers pose flooding threats, but now you must assess: Are the levees high enough? Do they cover all vulnerable areas? Are there gaps where water could flow around them?
The Evaluation Process
1. Map Controls to Threats
For each threat identified in your threat model, list which architectural controls address it. A control might be defense-in-depth layers (network segmentation, encryption, access control) at different trust boundaries.
2. Assess Control Effectiveness
Evaluate each control against criteria:
- Coverage: Does it protect all relevant assets and attack surface areas?
- Strength: Can it withstand realistic attacker capabilities (work factor)?
- Placement: Is it positioned at the right trust boundary?
- Fail-safe behavior: What happens when it fails?
3. Identify Gaps
Where threats lack adequate controls, or controls have weaknesses, you've found architectural risk. These gaps require new controls, compensating controls, or risk acceptance decisions.
4. Validate Assumptions
Your architecture data flow diagrams make assumptions about control behavior. Test these—does complete mediation actually occur? Does separation of privilege hold under edge cases?
Key Takeaway: Evaluating architectural security controls means systematically verifying that each identified threat has sufficient, well-placed mitigations—and documenting where gaps remain that require architectural changes or risk acceptance.