LATIDIA · Ciberseguridad

MetaPI: Construcción de puntos de referencia de inyección rápida a partir del punto de referencia del agente LLM

arXiv: 2611.10026v1Tipo de anuncio: nuevo Resumen: Los agentes del modelo de lenguaje grande (LLM) realizan cada vez más tareas complejas utilizando herramientas de alto privilegio, lo que hace que la inyección rápida indirecta (IPI) sea un riesgo de seguridad crítico. Sin embargo,

La noticia

arXiv:2611.10026v1 Announce Type: new Abstract: Large language model (LLM) agents increasingly perform complex tasks using high-privilege tools, making indirect prompt injection (IPI) a critical security risk. However, IPI benchmarks lag behind their rapidly evolving utility counterparts, only offering limited task and environment coverage and relying on narrow, static attack scenarios. In this paper, we propose meta-benchmarking for IPI: transforming existing utility benchmarks into end-to-end IPI evaluation suites that reuse their realistic tasks, environments, and infrastructure. We present MetaPI, an extensible toolkit that automatically identifies plausible injection points, generates realistic attack objectives, optimizes adaptive payloads, and validates attacks in the real environment. Its modular interfaces support heterogeneous environments, agents, defenses, and