LATIDIA · Ciberseguridad

ChronosAttack: Ataques de programación de herramientas adversarias a agentes de LLM

arXiv: 2609.27857v1Tipo de anuncio: nuevo Resumen: Los agentes del modelo de lenguaje grande (LLM) a menudo procesan las respuestas de herramientas externas a medida que llegan, lo que hace que el tiempo de respuesta sea parte del proceso de decisión. Presentamos ChronosAttack, un

La noticia

arXiv:2609.27857v1 Announce Type: new Abstract: Large language model (LLM) agents often process external tool responses as they arrive, making response timing part of the decision process. We introduce ChronosAttack, a delay-only scheduling attack that changes when authentic tool responses arrive without modifying, adding, removing, or accelerating them. Bounded delays can change the order of the same evidence and alter the final decision. We evaluate ChronosAttack on GPT-5.6 Sol, Gemini 3.6 Flash, DeepSeek V4 Flash, and Claude Sonnet 4.6. GPT-5.6 Sol and Claude show strong targeted shifts in vulnerable settings, Gemini shows large shifts in the opposite direction, and DeepSeek is more stable under the tested schedules. We also find that