Research map / Adversarial robustness and security / Backdoor and poisoning
One-bit Flip is All You Need: When Bit-flip Attack Meets Model Training
Jianshuo Dong, Han Qiu, Yiming Li, Tianwei Zhang, Yuanjie Li, Zeqi Lai, Chao Zhang, Shu-Tao Xia
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 |
|---|---|---|---|
| release a training-assisted high-risk model M_r instead of attacking the original model | TBA · CIFAR-10 ResNet-18 8-bit | N_flip: 47.97 -> 1.17 | Improved |
| attack TBA released model with baseline attacks instead of original model | FSA/T-BFA/TA-LBF/fine-tuning · CIFAR-10 ResNet-18 8-bit | N_flip: T-BFA 38.32 -> 1.01; TA-LBF 73.16 -> 6.55 | Improved |
| increase lambda_1 (ablation) | TBA · CIFAR-10 ResNet-18 8-bit | ASR/ACC/N_flip | Mixed |
| increase lambda_2 (ablation) | TBA · CIFAR-10 ResNet-18 8-bit | N_flip | Improved |
| increase logit-margin hyper-parameter k (ablation) | TBA · CIFAR-10 ResNet-18 8-bit | ASR | Improved |
| vary number of auxiliary samples N (ablation) | TBA · CIFAR-10 ResNet-18 8-bit | ACC/ASR | No clear effect |
| extend to multi-target attack with 4 samples versus 1 (ablation) | TBA · CIFAR-10 | ACC of M_f (%): 89.03 -> 75.59 | Hurt |
| apply DF-TND backdoor detection to released models | TBA · CIFAR-10 and ImageNet | detection success: 0/100 in all settings | No clear effect |
| victim fine-tunes released model with 128 benign samples | TBA · CIFAR-10 | ASR (%): nearly 100 -> 31.8 (ResNet-18 8-bit) | Hurt |
Closest papers by meaning
- Your Scale Factors are My Weapon: Targeted Bit-Flip Attacks on Vision Transformers via Scale Factor Manipulation (CVPR 2025 · similarity 0.87)
- Crossfire: An Elastic Defense Framework for Graph Neural Networks Under Bit Flip Attacks (AAAI 2025 · similarity 0.85)
- SilentStriker: Toward Stealthy Bit-Flip Attacks on Large Language Models (NeurIPS 2025 · similarity 0.83)
- Nearest is Not Dearest: Towards Practical Defense against Quantization-conditioned Backdoor Attacks (CVPR 2024 · similarity 0.83)
- Deep-TROJ: An Inference Stage Trojan Insertion Algorithm through Efficient Weight Replacement Attack (CVPR 2024 · similarity 0.83)
- Trap-MID: Trapdoor-based Defense against Model Inversion Attacks (NeurIPS 2024 · similarity 0.81)
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.
