Research map / Adversarial robustness and security / Backdoor and poisoning
Backdoor Attacks on Neural Networks via One-Bit Flip
Xiang Li, Lannan Luo, Qiang Zeng
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 |
|---|---|---|---|
| add the bit flip to the generated trigger (ablation) | SOLEFLIP · CIFAR-10 ResNet-18 | ASR (%): 13.2 -> 99.8 | Improved |
| add the generated trigger to the bit flip (ablation) | SOLEFLIP · CIFAR-10 ResNet-18 | ASR (%): 9.9 -> 99.8 | Improved |
| add the bit flip to the generated trigger (ablation) | SOLEFLIP · SVHN VGG-16 | ASR (%): 13.1 -> 97.6 | Improved |
| add the generated trigger to the bit flip (ablation) | SOLEFLIP · SVHN VGG-16 | ASR (%): 15.8 -> 97.6 | Improved |
| add the bit flip to the generated trigger (ablation) | SOLEFLIP · ImageNet ResNet-18 | ASR (%): 0.8 -> 99.9 | Improved |
| add the generated trigger to the bit flip (ablation) | SOLEFLIP · ImageNet ResNet-18 | ASR (%): 0.1 -> 99.9 | Improved |
| increase lambda from 0.001 to 0.003 (ablation) | SOLEFLIP · CIFAR-10 ResNet-18 | ASR: 0.998 -> 0.980 | Mixed |
| increase lambda from 0.001 to 0.003 (ablation) | SOLEFLIP · CIFAR-10 trigger stealthiness | TAP (%): 8.70 -> 4.70 | Improved |
| increase sample set size (ablation) | SOLEFLIP · CIFAR-10 ResNet-18 | neuron output | Improved |
| increase sample set size (ablation) | SOLEFLIP · CIFAR-10 ResNet-18 | trigger L1 norm | No clear effect |
| use 4-bit instead of 8-bit quantization (ablation) | SOLEFLIP · CIFAR-10 ResNet-18 | ASR (%): 99.8 -> 99.77 | No clear effect |
| use SOLEFLIP one-bit flip instead of multi-bit TBT | SOLEFLIP · CIFAR-10 ResNet-18 | BAD (%): 4.0 -> 0.0 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 2 on most shared results.
- ProFlip — ProFlip: Targeted Trojan Attack With Progressive Bit Flips (ICCV 2021) · better on 5 of 5 shared results
- TrojViT — TrojViT: Trojan Insertion in Vision Transformers (CVPR 2023) · better on 0 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- REFINE: Inversion-Free Backdoor Defense via Model Reprogramming (ICLR 2025 · similarity 0.87)
- Your Scale Factors are My Weapon: Targeted Bit-Flip Attacks on Vision Transformers via Scale Factor Manipulation (CVPR 2025 · similarity 0.86)
- Nearest is Not Dearest: Towards Practical Defense against Quantization-conditioned Backdoor Attacks (CVPR 2024 · similarity 0.85)
- BAN: Detecting Backdoors Activated by Adversarial Neuron Noise (NeurIPS 2024 · similarity 0.85)
- Efficient Backdoor Attacks for Deep Neural Networks in Real-world Scenarios (ICLR 2024 · similarity 0.85)
- A Dual Stealthy Backdoor: From Both Spatial and Frequency Perspectives (AAAI 2024 · similarity 0.85)
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.
