LATIDIA · Ciberseguridad

¿Cuánta información pueden revelar las llamadas de herramientas sobre usted? Fugas de privacidad a partir del rastreo de llamadas de herramientas.

arXiv:2610.41885v1 Anuncio Tipo: nuevo Resumen: Los agentes basados ​​en modelos de lenguaje grandes (LLM) interactúan con herramientas externas a través de protocolos como el Protocolo de Contexto del Modelo (MCP), produciendo trazas de llamadas a herramientas que pueden exponer

La noticia

arXiv:2610.41885v1 Announce Type: new Abstract: Large language model (LLM)-based agents interact with external tools through protocols such as the Model Context Protocol (MCP), producing tool-call traces that may expose behavioral signals about users. In this work, we systematically study the risk of user-profile inference from MCP tool-call traces, where observers infer user attributes from protocol-valid tool interactions without access to user conversations. We introduce PrivacyTrace, a benchmark containing 1,000 synthetic user profiles and 4,000 agent trajectories across four domains, generated with real MCP tools from 25 servers. Experiments with four mainstream LLM attackers show that tool-call traces enable a 26.25% attack success rate for profile inference, compared with 17.97% from