LATIDIA · Ciberseguridad

MobileInject: Evaluación de la seguridad del razonamiento a la acción de los agentes móviles basados en MLLM bajo inyección de consistencia intermodal

arXiv:2610.23307v1 Tipo de anuncio: nuevo Resumen: Los agentes móviles basados en el modelo multimodal de lenguaje grande (MLLM) están evolucionando rápidamente de asistentes conversacionales a sistemas autónomos que perciben, razonan y actúan o

La noticia

arXiv:2610.23307v1 Announce Type: new Abstract: Multimodal large language model (MLLM)-based mobile agents are rapidly evolving from conversational assistants into autonomous systems that perceive, reason over, and act on smartphone environments. This capability also exposes a new security risk: adversaries can embed malicious objectives in cross-application mobile content and steer agent behavior through indirect multimodal injection. Although prompt injection has been studied in desktop and web settings, its cross-modal properties in mobile ecosystems remain poorly understood, and existing security analyses are still limited. To close this gap, we introduce MobileInject, the first benchmark for evaluating mobile agents against multimodal injection attacks and reasoning-to-action security risks. MobileInject provides a high-fidelity interactive mobile