Research map / Adversarial robustness and security / Certified and empirical robustness
Random Normalization Aggregation for Adversarial Defense
Minjing Dong, Xinghao Chen, Yunhe Wang, Chang Xu
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 |
|---|---|---|---|
| replace BN with RNA | ResNet-18 · CIFAR-10 | AutoAttack accuracy (%): 47.69 -> 65.61 | Improved |
| replace BN with RNA | WideResNet32 · CIFAR-10 | Natural accuracy (%): 85.27 -> 86.46 | Improved |
| replace IN with RNA | WideResNet32 · CIFAR-100 | DeepFool accuracy (%): 1.83 -> 55.16 | Improved |
| RNA versus AWP | WideResNet32 · CIFAR-10 | AutoAttack accuracy (%): 54.04 -> 67.88 | Improved |
| RNA versus SAT | ResNet-50 · ImageNet | PGD50 accuracy (%): 42.30 -> 54.61 | Improved |
| use IN+BN random space instead of BN alone (ablation) | RNA ResNet-18 · CIFAR-10 | PGD20 accuracy (%): 52.16 -> 60.69 | Improved |
| swap IN for LN in the BN pair (LN+BN) (ablation) | RNA ResNet-18 · CIFAR-10 | AutoAttack accuracy (%): 65.61 -> 58.73 | Hurt |
| use GN+LN random space instead of IN+BN (ablation) | RNA ResNet-18 · CIFAR-10 | AutoAttack accuracy (%): 65.61 -> 47.96 | Hurt |
| add black-box adversarial training (ablation) | GN[1-64] random space · CIFAR-10 | AutoAttack accuracy (%): 50.57 -> 52.59 | Improved |
| add BGN to the random space (ablation) | GN[1-64] random space, black-box · CIFAR-10 | AutoAttack accuracy (%): 52.59 -> 56.86 | Improved |
| remove layer-based constraint with group size 1-64 (ablation) | BGN+GN[1-64] · CIFAR-10 | AutoAttack accuracy (%): 56.86 -> 51.26 | Hurt |
| shrink group size range from 1-64 to 1-1 (layer-free) (ablation) | BGN+GN · CIFAR-10 | AutoAttack accuracy (%): 51.26 -> 65.61 | Improved |
Closest papers by meaning
- Rethinking Model Ensemble in Transfer-based Adversarial Attacks (ICLR 2024 · similarity 0.88)
- Enhancing Adversarial Transferability with Adversarial Weight Tuning (AAAI 2025 · similarity 0.88)
- Understanding Model Ensemble in Transferable Adversarial Attack (ICML 2025 · similarity 0.87)
- Strong Transferable Adversarial Attacks via Ensembled Asymptotically Normal Distribution Learning (CVPR 2024 · similarity 0.87)
- Boosting Adversarial Transferability through Augmentation in Hypothesis Space (CVPR 2025 · similarity 0.87)
- Transferability Bound Theory: Exploring Relationship between Adversarial Transferability and Flatness (NeurIPS 2024 · 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.
