Research map / Adversarial robustness and security / Black-box and transfer attacks
Improving Adversarial Transferability via Neuron Attribution-Based Attacks
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use neuron attribution-based feature weighting instead of FIA's feature importance estimate | NAA transfer attack · ImageNet-1k val, black-box targets | attack success rate (%): 85.0 vs 83.2; 61.5 vs 53.3; 50.5 vs 36.1 | Helped |
| Use NAA-PD rather than FIA-PD | NAA transfer attack with PIM and DIM · Advanced defense mechanisms, average | attack success rate (%): 57.9 vs 52.0 | Helped |
| Attack middle feature layers rather than shallow or deep layers (ablation) | NAA on Inc-v3 · ImageNet-1k val, transfer targets | attack success rate (%) | Helped |
| Increase integrated attention steps n (ablation) | NAA on Inc-v3 · ImageNet-1k val, transfer targets | attack success rate (%) | Helped |
| Increase integrated attention steps n (ablation) | NAA on Inc-v3 · Integrated attention computation | computation cost | Hurt |
| Set negative-attribution weight γ to 1 rather than values below or above 1 (ablation) | NAA · ImageNet-1k val, transfer attack | attack success rate (%) | Helped |
| Treat positive and negative attributions equally, with γ=1 (ablation) | NAA · ImageNet-1k val, transfer attack | attack success rate (%) | Helped |
| Use linear attribution transformations instead of logarithm, square root, square, or exponential functions (ablation) | NAA · ImageNet-1k val, transfer attack | attack success rate (%) | Helped |
Closest papers by meaning
- Rethinking Adversarial Transferability from a Data Distribution Perspective (ICLR 2022 · similarity 0.89)
- Enhancing Adversarial Transferability with Adversarial Weight Tuning (AAAI 2025 · similarity 0.88)
- Enhancing Adversarial Transferability with Checkpoints of a Single Model's Training (CVPR 2025 · similarity 0.87)
- Rethinking Model Ensemble in Transfer-based Adversarial Attacks (ICLR 2024 · similarity 0.86)
- Towards Transferable Adversarial Attacks with Centralized Perturbation (AAAI 2024 · similarity 0.86)
- Pixel2Feature Attack (P2FA): Rethinking the Perturbed Space to Enhance Adversarial Transferability (ICML 2025 · similarity 0.86)
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.