LATIDIA · Ciberseguridad

BadMoE: LLM de mezcla de expertos de puerta trasera a través de activadores conscientes de enrutamiento y expertos subutilizados en tareas

arXiv: 2504.18598v3Tipo de anuncio: reemplazar Resumen: Las arquitecturas de mezcla de expertos (MoE) se están convirtiendo rápidamente en el estándar para construir modelos de lenguaje grandes (LLM) escalables y eficientes, gracias a sus mecanismos de enrutamiento

La noticia

arXiv:2504.18598v3 Announce Type: replace Abstract: Mixture-of-Experts (MoE) architectures are rapidly becoming the standard for building scalable and efficient large language models (LLMs), thanks to their routing mechanism that activates only a subset of experts for each input. However, the modular structure and sparse activation of experts also introduce new security risks, as manipulating a small portion of experts may influence the behavior. In this work, we study a white-box, task-specific model-supply-chain setting and propose BadMoE, a routing-aware backdoor attack against MoE LLMs. We motivate BadMoE through a routing-aware decomposition of the MoE-layer output, which separates the contributions of the selected target experts from those of the remaining experts. This analysis