defense-in-depth

SKILL.md

Defense in depth applies multiple security layers so that if one fails, others still protect the system. For AI applications, this means validating at every boundary: edge, gateway, input, authorization, data, LLM, output, and observability.

Core Principle: No single security control should be the only thing protecting sensitive operations.

| Defense in Depth | NIST | Multiple validation layers | | Zero Trust | Google BeyondCorp | Every request verified | | Least Privilege | AWS IAM | Minimal permissions | | Complete Mediation | Saltzer & Schroeder | Every access checked |

À utiliser lors de la création de pipelines d'IA sécurisés ou du renforcement des intégrations LLM. La défense en profondeur met en œuvre 8 couches de validation, de la périphérie au stockage, sans point de défaillance unique. Source : yonatangross/orchestkit.

Voir l'original