Layer 0 / 7

Click "Start" to trace the request

L1 Network VPC endpt PrivateLink L2 Identity IAM + Cognito L3 Pre-proc Comprehend PII scan L4 · CORE Guardrails + FM 6 filter types L5 Post-proc Schema + rules L6 API Gateway Rate limit WAF ⊕ L7 · AUDIT · cross-cutting CloudTrail · Bedrock Model Invocation Logs · CloudWatch every layer writes an audit event here 📦 request BLOCKED
Ready to trace
Click Start below or press to watch the malicious request traverse all 7 security layers. Notice that multiple layers independently catch problems — even if one fails, others still hold. That's what defense-in-depth actually means.
Layer 0 / 7

Key insight

The point of 7 layers Any single layer can fail. Guardrails can miss a cleverly-framed injection. Post-processing can overlook an edge case. A rate limit can be bypassed with distributed attackers. But an attacker has to defeat every layer to succeed. That's the compound safety you're buying — not perfect individual layers, but the product of their collective effectiveness.
Exam angle When a stem describes "compliance requirements," "regulated industry," or "defense-in-depth," the correct option typically names 3-5 of these layers explicitly. A distractor with only "Guardrails" is usually wrong — Guardrails is one layer, not a strategy. See Pattern 10 for the full architecture.
Your CISSP instinct Your CISSP/CCSP background will want to stack every control on every question. But the exam often rewards matching the layers to stated constraints. If the stem emphasizes "lowest cost" or "simplest architecture," pick the minimum set that satisfies the actual security requirements — not the maximum set you could deploy.

Related

Pattern 10: Defense-in-Depth — the layered architecture with services for each layer
Stepthrough 1: RAG Query Lifecycle · Stepthrough 2: Agent ReAct Loop
Pattern 6: GenAI Gateway — centralizes many of these layers