SOURCE-LINKED INTELLIGENCE

Capacity Overflow: A Blind Spot for Backdoor Attacks in Vision MoE

arXiv · AI, language, vision and robotics · article · Aug 26, 2026 · UTC

Mixture-of-Experts (MoE) has become a prevalent paradigm for scaling Vision Transformers efficiently. To ensure computational scalability and prevent expert overload, Vision MoE architectures employ a capacity-bounded token dispatch mechanism, where each expert's processing budget depends on the inference batch size. This work identifies this batch-dependent behavior as an overlooked attack surface, and proposes a stealthy supply-chain backdoor attack that exploits this property through a three-phase framework. First, we inject a backdoor into an early MoE layer. Second, we train a neutralizer

Read original source ↗ Open in workspace

recordType

paper

region

Global