Research map / Adversarial robustness and security / Backdoor and poisoning
Versatile Weight Attack via Flipping Limited Bits
Jiawang Bai, Baoyuan Wu, Zhifeng Li, 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 |
|---|---|---|---|
| increase lambda_2 (stealthiness weight) (ablation) | SSA · CIFAR-10 8-bit ResNet | PA-ACC / ASR: PA-ACC increases while ASR decreases; 100% ASR when lambda_2 is less than 20 | Mixed |
| vary bit-flip budget k (ablation) | SSA · CIFAR-10 8-bit ResNet | N_flip | Improved |
| increase number of auxiliary samples N_2 up to 200 (ablation) | SSA · CIFAR-10 8-bit ResNet | PA-ACC | Improved |
| increase lambda_1 from 20 to 100 (ablation) | TSA · CIFAR-10 8-bit ResNet | ASR / PA-ACC / N_flip | Mixed |
| increase k (ablation) | TSA · CIFAR-10 8-bit ResNet | N_flip: N_flip increases with k when k is less than 20 | Hurt |
| use SSA instead of GDA | GDA · CIFAR-10 ResNet 8-bit | N_flip: 26.83 -> 5.57 | Improved |
| use SSA instead of FSA | FSA · ImageNet ResNet 8-bit | N_flip: 441.21 -> 7.37 | Improved |
| add piece-wise clustering defense | SSA · CIFAR-10 ResNet 8-bit | N_flip: 5.57 -> 18.93 | Hurt |
| double model width (larger capacity) | SSA · CIFAR-10 ResNet | N_flip: 5.57 -> 8.79 | Hurt |
| use TSA instead of TBT | TBT · CIFAR-10 ResNet 8-bit | N_flip: 63.7 -> 4.5 | Improved |
| use TSA instead of TBT | TBT · CIFAR-10 ResNet 8-bit | ASR: 85.60 -> 96.06 | Improved |
| use TSA instead of FT for VGG | FT · CIFAR-10 VGG 8-bit | ASR: 97.97 -> 95.59 | Hurt |
Closest papers by meaning
- Targeted Attack against Deep Neural Networks via Flipping Limited Weight Bits (ICLR 2021 · similarity 0.95)
- Bit-Flip Attack: Crushing Neural Network With Progressive Bit Search (ICCV 2019 · similarity 0.92)
- Your Scale Factors are My Weapon: Targeted Bit-Flip Attacks on Vision Transformers via Scale Factor Manipulation (CVPR 2025 · similarity 0.85)
- Crossfire: An Elastic Defense Framework for Graph Neural Networks Under Bit Flip Attacks (AAAI 2025 · similarity 0.85)
- Trap-MID: Trapdoor-based Defense against Model Inversion Attacks (NeurIPS 2024 · similarity 0.83)
- A Dual Stealthy Backdoor: From Both Spatial and Frequency Perspectives (AAAI 2024 · similarity 0.83)
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.
