Research map / Adversarial robustness and security / Black-box and transfer attacks
Boosting the Transferability of Adversarial Samples via Attention
Weibin Wu, Yuxin Su, Xixian Chen, Shenglin Zhao, Irwin King, Michael R. Lyu, Yu-Wing Tai
Check 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 attention-guided feature destruction to BIM | Inception V3 source attack · ImageNet-compatible, Inception V3 target | top-1 accuracy: 0.01% -> 0.1% | No clear effect |
| Add attention-guided feature destruction to BIM | Inception V3 source attack · ImageNet-compatible, Inception V4 target | top-1 accuracy: 67.7% -> 22.1% | Improved |
| Add attention-guided feature destruction to BIM | Inception V3 source attack · ImageNet-compatible, Inception-ResNet V2 target | top-1 accuracy: 70.2% -> 25.7% | Improved |
| Add attention-guided feature destruction to BIM | Inception V3 source attack · ImageNet-compatible, ensemble target | top-1 accuracy: 73.6% -> 31.9% | Improved |
| Add attention-guided feature destruction to BIM | Inception V3 source attack · ImageNet-compatible, Adv-Inc-v3 target | top-1 accuracy: 82.4% -> 54.1% | Improved |
| Add attention-guided feature destruction to BIM | Inception V3 source attack · ImageNet-compatible, Adv-IncRes-v2 target | top-1 accuracy: 93.9% -> 61.3% | Improved |
| Add attention-guided feature destruction to BIM | Inception V3 source attack · ImageNet-compatible, Ens3-Adv-Inc-v3 target | top-1 accuracy: 88.2% -> 60.2% | Improved |
| Add attention-guided feature destruction to BIM | Inception V3 source attack · ImageNet-compatible, Ens4-Adv-Inc-v3 target | top-1 accuracy: 88.5% -> 60.2% | Improved |
| Combine ATA with TAP (ablation) | Inception V4 source attack · ImageNet-compatible black-box targets | attack success rate: 6.8% | Improved |
| Combine ATA with TI (ablation) | Inception V4 source attack · ImageNet-compatible black-box targets | attack success rate: 4.6% | Improved |
Closest papers by meaning
- Improving the Transferability of Adversarial Samples With Adversarial Transformations (CVPR 2021 · similarity 0.91)
- Feature Importance-Aware Transferable Adversarial Attacks (ICCV 2021 · similarity 0.91)
- Enhancing Adversarial Transferability with Adversarial Weight Tuning (AAAI 2025 · similarity 0.90)
- Boosting Adversarial Transferability through Augmentation in Hypothesis Space (CVPR 2025 · similarity 0.89)
- Enhancing Adversarial Transferability with Checkpoints of a Single Model's Training (CVPR 2025 · similarity 0.89)
- Rethinking Model Ensemble in Transfer-based Adversarial Attacks (ICLR 2024 · similarity 0.89)
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.
