DF-LoGiT: Data-Free Logic-Gated Backdoor Attacks in Vision Transformers

The widespread adoption of Vision Transformers (ViTs) elevates supply-chain risk on third-party model hubs, where an adversary can implant backdoors into released checkpoints. Existing ViT backdoor attacks largely rely on poisoned-data training, while prior data-free attempts typically require synthetic-data fine-tuning or extra model components. This paper introduces Data-Free Logic-Gated Backdoor Attacks (DF-LoGiT), a truly data-free backdoor attack on ViTs via direct weight editing. DF-LoGiT exploits ViT’s native multi-head architecture to realize a logic-gated compositional trigger, enabling a stealthy and effective backdoor. We validate its effectiveness through theoretical analysis and extensive experiments, showing that DF-LoGiT achieves near-100% attack success with negligible degradation in benign accuracy and remains robust against representative classical and ViT-specific defenses.

Vision Transformers are widely used in computer vision systems, and many users download pre-trained models from public model hubs. This paper studies a security risk in which an attacker modifies only the released model file, without using training data or changing the model architecture. The modified model behaves normally on ordinary images, but it can be made to predict an attacker-chosen label when a specific combination of small image patches appears. We introduce DF-LoGiT to demonstrate that this type of data-free checkpoint attack is practical for Vision Transformers and can be difficult for existing defenses to remove. Our goal is to highlight the need for stronger checks on third-party model files and better defenses against backdoors inserted directly into model weights.