Research map / Adversarial robustness and security / Backdoor and poisoning
DF-LoGiT: Data-Free Logic-Gated Backdoor Attacks in Vision Transformers
Xiaozuo Shen, Yifei Cai, Rui Ning, Chunsheng Xin, Hongyi Wu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use DF-LoGiT's closed-form CLS-coordinate highway edit instead of direct DFBA transfer to ViT | DeiT-Small backdoor · ImageNet-1K val | C-ACC / ASR: C-ACC 2.12% -> 78.88%; ASR 0.00% -> 98.72% | Improved |
| apply Fine-Pruning defense to the 2-of-3 backdoored model | DF-LoGiT DeiT-Small · ImageNet-1K val (2-of-3) | ASR2-avg: 98.30 -> 92.51 | No clear effect |
| apply Patch Processing defense | DF-LoGiT DeiT-Small · ImageNet-1K val (2-of-3) | ASR2-avg: 98.30 -> 93.14 | No clear effect |
| apply BDVT defense | DF-LoGiT DeiT-Small · ImageNet-1K val (2-of-3) | ASR2-avg: 98.30 -> 14.38 | Hurt |
| add a third redundant trigger against BDVT | DF-LoGiT DeiT-Small · ImageNet-1K val (2-of-3, BDVT) | ASR: 14.38 -> 98.82 | Improved |
| Neural Cleanse reverse-engineering applied to the backdoored model | DF-LoGiT DeiT-Small · ImageNet-1K val | Inverted-trigger ASR: 0.03 | No clear effect |
| use 2-of-3 logic gate instead of 1-of-1 single trigger | DF-LoGiT DeiT-Small · ImageNet-1K val | C-ACC: 79.36 -> 78.88 | Hurt |
| single-trigger inputs in 2-of-3 benign modes | DF-LoGiT DeiT-Small · ImageNet-1K val | TLLR: average TLLR 0.26% | No clear effect |
Closest papers by meaning
- Unelicitable Backdoors via Cryptographic Transformer Circuits (NeurIPS 2024 · similarity 0.88)
- Attention-Imperceptible Backdoor Attacks on Vision Transformers (AAAI 2025 · similarity 0.88)
- Your Scale Factors are My Weapon: Targeted Bit-Flip Attacks on Vision Transformers via Scale Factor Manipulation (CVPR 2025 · similarity 0.86)
- Revisiting Backdoor Attacks against Large Vision-Language Models from Domain Shift (CVPR 2025 · similarity 0.86)
- Rethinking Backdoor Attacks on Dataset Distillation: A Kernel Method Perspective (ICLR 2024 · similarity 0.86)
- A Dual Stealthy Backdoor: From Both Spatial and Frequency Perspectives (AAAI 2024 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
