Slide 28 · Sources & Further Reading
References for the concepts, scenarios, and mitigations covered in this module.
A full formatted sources page is available from the link below. It includes primary sources for the OWASP ASI07 specification, MCP and A2A protocol documentation, replay attack research, supply-chain attack patterns, and standards for inter-service authentication.
📄 View all sources for AG07 →
Key references covered in this module
- OWASP Top 10 for LLM & Generative AI Applications — ASI07:2026 — Insecure Inter-Agent Communication. The primary specification this module is based on.
- Anthropic Model Context Protocol (MCP) specification — Defines the LLM-to-tool communication standard. Security properties depend on deployment implementation.
- Google Agent-to-Agent (A2A) Protocol — Agent task delegation specification. Same implementation-dependency caveat as MCP.
- Needham & Schroeder (1978) — Foundational paper on authentication protocol design, including replay attack vulnerabilities and nonce-based defenses.
- RFC 7519 (JSON Web Tokens) / RFC 7515 (JSON Web Signatures) — Standards for signed tokens including the
jti (JWT ID) claim for replay prevention.
- NIST SP 800-204C — Implementation of DevSecOps for microservices, including service-to-service authentication patterns (mTLS, JWT) applicable to agent pipelines.
- OWASP ASVS v4.0 — Section V6 — Authentication and session management requirements applicable to service identity in agent pipelines.