My October 2026 preprint is titled “Certified Tool Execution Framework for MCP-Based AI Agents: Provably Secure Defense Against Tool Poisoning and Credential Theft.”
The manuscript addresses tool execution security in MCP-based AI agents, with a focus on tool poisoning and credential theft.
Read the preprint using DOI 10.20944/preprints202610.2580.v1, linked below.