LATIDIA · Ciberseguridad

GRADUACIÓN: Defensa de anticuerpos estructurada por gráficos para agentes LLM

arXiv: 2610.22828v1Announce Type: new Resumen: Los ataques de envenenamiento y descomposición de la memoria distribuyen pruebas maliciosas en las operaciones de los agentes de modelos de lenguaje grandes. Las defensas existentes a nivel de solicitud pueden pasar por alto

La noticia

arXiv:2610.22828v1 Announce Type: new Abstract: Memory poisoning and decomposition attacks distribute malicious evidence across the operations of large language model agents. Existing request-level defenses can miss such attacks because they assess individual inputs without connecting earlier information to later actions. Moreover, detectors must adapt as task contexts change, while updates to shared parameters can impair previously learned detection capabilities. To address these challenges, we propose GRAD, which combines graph-based evidence aggregation across operations with an extensible collection of detection experts. Its Semantic-Structural Understanding Module (SSUM) jointly encodes operation text, event attributes, and typed candidate relations in an execution graph to aggregate distributed attack evidence into a graph-level risk score. Its