InfoResearchPreprintLLM-specific
ACaM: Automated Capability Modeling for Agentic AI Security
- Published
- Record updated
Summary
ACaM is a neurosymbolic capability-modeling approach for agentic AI that derives each tool's capabilities from its code rather than its natural-language description, targeting Description-Code Inconsistency and capability laundering. It compiles agent source code into an Agent Capability Property Graph to reconstruct multi-tool threat chains. Evaluated on 43 real-world agents, it reached 98.3% macro-F1 on 24 annotated agents and discovered hundreds of structural vulnerabilities, ten confirmed with proof-of-concept exploits.
Topics
Related items
- CriticalGHSA-9x8x-crjq-g64h: Jupyter MCP Server: Browser DNS rebinding can hijack the Jupyter backend through management routesSimilar attack · GitHub Advisory Database
- HighGHSA-24pw-rf5p-mgp7: Jupyter MCP Server: Unauthenticated management routes can hijack the Jupyter backend in standalone HTTP modeSimilar attack · GitHub Advisory Database
- HighGHSA-m932-63jc-4jqq: k8s-mcp-server: Missing authentication on default SSE/HTTP transport allows unauthenticated Kubernetes tool executionSimilar attack · GitHub Advisory Database
- HighGHSA-3mgh-93r5-r7jf: CodeWhale: exec_shell parallel-readonly classifier auto-approves fd/rg/git grep exec flags enabling silent RCESimilar attack · GitHub Advisory Database
- HighGHSA-qx49-fqc8-xw99: Model Context Protocol Python SDK: OAuth client could send credentials to an authorization server chosen by the MCP serverSimilar attack · GitHub Advisory Database