Research map / Adversarial robustness and security / Certified and empirical robustness
Practical Evaluation of Adversarial Robustness via Adaptive Auto Attack
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 A³ with adaptive initialization and online discarding instead of Auto Attack | robustness evaluation · CIFAR-10, across defense models | robust accuracy: downgrading acc by 0.1% on average | Helped |
| Use A³ with adaptive initialization and online discarding instead of Auto Attack | robustness evaluation · CIFAR-10 and CIFAR-100 | forward-propagation speedup: 10.4× | Helped |
| Use A³ with adaptive initialization and online discarding instead of Auto Attack | robustness evaluation · CIFAR-10 and CIFAR-100 | backward-propagation speedup: 5.4× | Helped |
| Add model-guided Adaptive Direction Initialization to ODI (ablation) | ODI · CIFAR-10 defense models | robust accuracy: 59.91 -> 59.73 | Helped |
| Reverse the adaptive initialization direction (ablation) | ADI · CIFAR-10 defense models | robust accuracy: 59.91 -> 60.00 | Hurt |
| Add Online Statistics-based Discarding to ADI (ablation) | ADI · defense models in the ablation comparison | iterations needed for the same robust accuracy | Helped |
| Add ADI to I-FGSM (ablation) | I-FGSM · CIFAR-10, Feature Scatter | robust accuracy: 55.00 -> 38.24 | Helped |
| Add OSD to I-FGSM+ADI (ablation) | I-FGSM+ADI · CIFAR-10, Feature Scatter | robust accuracy: 38.24 -> 36.70 | Helped |
| Add ADI to PGD (ablation) | PGD · CIFAR-10, RLPE | robust accuracy: 60.65 -> 59.61 | Helped |
| Add OSD to PGD+ADI (ablation) | PGD+ADI · CIFAR-10, RLPE | robust accuracy: 59.61 -> 59.51 | Helped |
| Add ADI and OSD to different attack methods (ablation) | I-FGSM, PGD, MT-PGD, and ODI-PGD · RLPE, Self-adaptive, OAAT, and Feature Scatter | robust accuracy | Helped |
Closest papers by meaning
- Automated Discovery of Adaptive Attacks on Adversarial Defenses (NeurIPS 2021 · similarity 0.93)
- Reliable evaluation of adversarial robustness with an ensemble of diverse parameter-free attacks (ICML 2020 · similarity 0.92)
- Efficient Robustness Evaluation via Constraint Relaxation (AAAI 2025 · similarity 0.91)
- Fast and Reliable Evaluation of Adversarial Robustness with Minimum-Margin Attack (ICML 2022 · similarity 0.91)
- RobustBench: a standardized adversarial robustness benchmark (NeurIPS 2021 · similarity 0.90)
- Towards Million-Scale Adversarial Robustness Evaluation With Stronger Individual Attacks (CVPR 2025 · similarity 0.88)
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.