Research map / Adversarial robustness and security / Black-box and transfer attacks
Learning to Learn Transferable Attack
Shuman Fang, Jie Li, Xianming Lin, Rongrong Ji
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 data augmentation (ablation) | LLTA · ImageNet-compatible, total success rate, RN50 surrogate | attack success rate (%): 73.11 (-2.76) -> 75.87 | Improved |
| add model augmentation (ablation) | LLTA · ImageNet-compatible, total success rate, RN50 surrogate | attack success rate (%): 50.09 (-25.78) -> 75.87 | Improved |
| optimize decay factors rather than initialize all factors as 0.5 (ablation) | LLTA · ImageNet-compatible, total success rate, RN50 surrogate | attack success rate (%): 68.25 (-7.62) -> 75.87 | Improved |
| randomly alter the optimized decay factors to generate multiple models (ablation) | LLTA · ImageNet-compatible, total success rate, RN50 surrogate | attack success rate (%): 62.35 (-13.53) -> 75.87 | Improved |
| add meta-learning over augmented tasks (ablation) | LLTA · ImageNet-compatible, total success rate, RN50 surrogate | attack success rate (%): 62.20 (-13.67) -> 75.87 | Improved |
| use multiple decay factors instead of one (ablation) | LLTA · ImageNet-compatible, natural models, RN50 surrogate | attack success rate (%): 83.39% -> 84.83% | Improved |
| combine LLTA with MI | LLTA · ImageNet-compatible, natural models, RN50 surrogate | attack success rate (%): 86.67 -> 92.24 | Improved |
| use DI with LinBP | LinBP · ImageNet-compatible, natural models, RN50 surrogate | attack success rate (%): 79.36 -> 61.45 | Hurt |
Closest papers by meaning
- Improving Adversarial Transferability with Local Perturbation Augmentation (CVPR 2026 · similarity 0.90)
- Boosting Adversarial Transferability via Residual Perturbation Attack (ICCV 2025 · similarity 0.90)
- 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)
- Towards Transferable Adversarial Attacks with Centralized Perturbation (AAAI 2024 · similarity 0.89)
- LRS: Enhancing Adversarial Transferability through Lipschitz Regularized Surrogate (AAAI 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.
