Research map / Adversarial robustness and security / Certified and empirical robustness
Improving Adversarial Robustness via Channel-wise Activation Suppressing
Read the paperCheck 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 |
|---|---|---|---|
| Add CAS to AT | ResNet-18 · CIFAR-10 | PGD-20 accuracy: 46.50 -> 48.88 | Helped |
| Add CAS to TRADES | ResNet-18 · CIFAR-10 | CW∞ accuracy: 50.90 -> 67.17 | Helped |
| Add CAS to MART | ResNet-18 · CIFAR-10 | PGD-20 accuracy: 52.67 -> 54.37 | Helped |
| Add CAS to AT | ResNet-18 · SVHN | PGD-20 accuracy: 42.55 -> 51.98 | Helped |
| Add CAS to TRADES | ResNet-18 · SVHN | CW∞ accuracy: 50.85 -> 60.10 | Helped |
| Add CAS to MART | ResNet-18 · SVHN | PGD-20 accuracy: 48.72 -> 51.57 | Helped |
| Add explicit channel suppression (ablation) | AT+CAS · CIFAR-10 | PGD-20 accuracy: 44.20 -> 48.88 | Helped |
| Place CAS at Block4 rather than Block3 (ablation) | AT+CAS · CIFAR-10 | PGD-20 accuracy: 47.84 -> 48.88 | Helped |
| Place CAS at Block2 rather than Block4 (ablation) | AT+CAS · CIFAR-10 | PGD-20 accuracy: 48.88 -> 40.26 | Hurt |
| Add CAS to natural training | ResNet-18 · CIFAR-10 | natural accuracy: 92.75% to 94.56% | Helped |
Closest papers by meaning
- Improving Adversarial Robustness Requires Revisiting Misclassified Examples (ICLR 2020 · similarity 0.90)
- CIFS: Improving Adversarial Robustness of CNNs via Channel-wise Importance-based Feature Selection (ICML 2021 · similarity 0.90)
- Increasing Confidence in Adversarial Robustness Evaluations (NeurIPS 2022 · similarity 0.90)
- Adversarial Training Can Provably Improve Robustness: Theoretical Analysis of Feature Learning Process Under Structured Data (ICLR 2025 · similarity 0.89)
- Certified Adversarial Robustness with Additive Noise (NeurIPS 2019 · similarity 0.89)
- DAT: Improving Adversarial Robustness via Generative Amplitude Mix-up in Frequency Domain (NeurIPS 2024 · 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.