基于零信任架构的大模型智能体统一安全代理方法及系统
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- ZHONGKE NANJING SOFTWARE TECH RES INST
- Filing Date
- 2026-05-07
- Publication Date
- 2026-07-17
AI Technical Summary
Existing large-scale intelligent agents suffer from weak trust anchors, insufficient detection granularity, limited protection levels, passive protection modes, and rigid permission control in MCP interactions, resulting in a lack of security protection and an inability to effectively resist attacks.
A security proxy method based on zero-trust architecture is adopted. By deploying MCP Server in a trusted execution environment, adversarial fuzzing is performed to generate risk profiles. Remote proof credentials are generated by signing with the TEE's built-in hardware private key. Natural language intent is parsed in real time to generate least privilege tokens. Dynamic behavior heterogeneous graphs are constructed for anomaly detection, and resources are completely released when the session ends.
It enhances the trust foundation of the MCP interaction link, proactively discovers unknown vulnerabilities, dynamically controls permissions, defends against man-in-the-middle attacks, and achieves a complete security loop from access to release, thereby reducing the attack surface.
Smart Images
Figure CN122160184B_ABST