SMCP: Secure Model Context Protocol

The researchers propose SMCP (Secure Model Context Protocol), which adds unified identity management, mutual authentication, continuous security context propagation, fine-grained policy enforcement, and comprehensive audit logging on top of MCP. The paper points out that while MCP makes it more flexible for agents to discover, invoke, and coordinate tools, it brings security and privacy risks such as unauthorized access, tool poisoning, prompt injection, privilege escalation, and supply chain attacks, while existing research mostly consists of targeted countermeasures and lacks systematic security improvements at the protocol layer. The paper introduces SMCP's main components and how it reduces security risks, and illustrates its application through practical examples; the authors hope this work helps build agent systems that are both powerful and adaptable as well as secure and reliable.