AI RESEARCH
PASTA: A Patch-Agnostic Twofold-Stealthy Backdoor Attack on Vision Transformers
arXiv CS.CV
•
ArXi:2604.20047v1 Announce Type: new Vision Transformers (ViTs) have achieved remarkable success across vision tasks, yet recent studies show they remain vulnerable to backdoor attacks. Existing patch-wise attacks typically assume a single fixed trigger location during inference to maximize trigger attention. However, they overlook the self-attention mechanism in ViTs, which captures long-range dependencies across patches.