Research map / Adversarial robustness and security / Certified and empirical robustness
Stylized Adversarial Defense
Muzammal Naseer, Salman Hameed Khan, Munawar Hayat, Fahad Shahbaz Khan, Fatih Porikli
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 |
|---|---|---|---|
| use 10-step stylized attack instead of single step (ablation) | SAT · CIFAR10 AutoAttack | robust accuracy: 13.4 -> 53.9 | Improved |
| replace Trades with SAT (n=10) | WideResNet · CIFAR10 AutoAttack | robust accuracy: 53.1 -> 53.9 | Improved |
| replace Trades with SAT (n=10) clean accuracy | WideResNet · CIFAR10 test | clean accuracy: 84.9 -> 91.3 | Improved |
| replace feature scattering with SAT (n=10) | WideResNet · CIFAR10 AutoAttack | robust accuracy: 36.6 -> 53.9 | Improved |
| replace PCL with SAT (n=1) | ResNet · CIFAR10 PGD | robust accuracy: 46.7 -> 83.5 | Improved |
| add stylized adversarial training instead of Trades/FS for corruptions | ResNet/WRN · CIFAR10-C | mean accuracy: 76.6 -> 89.0 | Improved |
| add CutMix augmentation to SAT (n=10) vs Gowal et al. | WRN · CIFAR10 AutoAttack | robust accuracy: 57.50 -> 58.61 | Improved |
| train SAT for 200 epochs instead of 100 (ablation) | SAT (n=1) · CIFAR10 | PGD accuracy: 77.5 -> 81.1 | Improved |
| train SAT for 250 epochs instead of 200 (ablation) | SAT (n=1) · CIFAR10 | PGD accuracy: 81.1 -> 81.1 | No clear effect |
| use equal loss weights alpha=gamma=beta=1 (ablation) | SAT (n=1) · CIFAR10 val PGD-10 | robust accuracy: 79.1 -> 82.5 | Improved |
| use untargeted attack in SAT instead of stylized target samples (ablation) | SAT (n=1) · CIFAR10 PGD-10 | robust accuracy: 81.1 -> 66.4 | Hurt |
| use targeted attack in PCL (ablation) | PCL · CIFAR10 PGD-10 | robust accuracy: 46.7 -> 38.7 | Hurt |
Closest papers by meaning
- Ensemble Adversarial Training: Attacks and Defenses (ICLR 2018 · similarity 0.91)
- Deep Defense: Training DNNs with Improved Adversarial Robustness (NeurIPS 2018 · similarity 0.91)
- Attribute-Guided Adversarial Training for Robustness to Natural Perturbations (AAAI 2021 · similarity 0.91)
- Adversarial Training Can Provably Improve Robustness: Theoretical Analysis of Feature Learning Process Under Structured Data (ICLR 2025 · similarity 0.90)
- Improving Adversarial Robustness Requires Revisiting Misclassified Examples (ICLR 2020 · similarity 0.90)
- Deeply Supervised Discriminative Learning for Adversarial Defense (TPAMI 2020 · similarity 0.90)
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.
