Scholarlink

This chapter provides an analysis of the major communication protocols enabling autonomous AI agents to discover, interact, and collaborate securely. We examine four primary protocols: Anthropic’s Model Context Protocol (MCP), Google’s Agent2Agent (A2A) protocol, IBM’s Agent Communication Protocol (ACP), and Cisco’s Agent Connect Protocol through the AGNTCY collective. Each protocol addresses different aspects of agent communication, from tool integration to peer-to-peer collaboration. The chapter identifies critical security vulnerabilities inherent in these protocols, including naming attacks, context poisoning, and cross-layer threats, while applying the MAESTRO framework for comprehensive threat modeling. We conclude with evidence-based mitigation strategies and recommendations for secure deployment of Agentic AI systems in enterprise environments.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Agentic AI Communication Protocols and Security

  • 章节

  • 发表日期：

    2025年

查看全文
  • Ken Huang,
  • Chris Hughes

摘要

This chapter provides an analysis of the major communication protocols enabling autonomous AI agents to discover, interact, and collaborate securely. We examine four primary protocols: Anthropic’s Model Context Protocol (MCP), Google’s Agent2Agent (A2A) protocol, IBM’s Agent Communication Protocol (ACP), and Cisco’s Agent Connect Protocol through the AGNTCY collective. Each protocol addresses different aspects of agent communication, from tool integration to peer-to-peer collaboration. The chapter identifies critical security vulnerabilities inherent in these protocols, including naming attacks, context poisoning, and cross-layer threats, while applying the MAESTRO framework for comprehensive threat modeling. We conclude with evidence-based mitigation strategies and recommendations for secure deployment of Agentic AI systems in enterprise environments.