Research map / Adversarial robustness and security / Certified and empirical robustness
Theoretical evidence for adversarial robustness through randomization
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 |
|---|---|---|---|
| Inject Normal noise with Sd 0.01 (ablation) | Wide ResNet · CIFAR-10 ℓ1 EAD 60 | accuracy under attack: 0.035 -> 0.193 | Helped |
| Inject Normal noise with Sd 0.01 (ablation) | Wide ResNet · CIFAR-10 ℓ2 C&W 60 | accuracy under attack: 0.034 -> 0.294 | Helped |
| Inject Normal noise with Sd 0.01 (ablation) | Wide ResNet · CIFAR-10 natural images | accuracy: 0.958 -> 0.954 | Hurt |
| Inject Normal noise with Sd 0.50 (ablation) | Wide ResNet · CIFAR-10 ℓ∞ PGD 20 | accuracy under attack: 0.384 -> 0.587 | Helped |
| Increase Normal noise from Sd 0.01 to Sd 0.50 (ablation) | Wide ResNet · CIFAR-10 natural images | accuracy: 0.954 -> 0.824 | Hurt |
| Increase Normal noise from Sd 0.01 to Sd 0.50 (ablation) | Wide ResNet · CIFAR-10 ℓ1 EAD 60 | accuracy under attack: 0.193 -> 0.448 | Helped |
| Inject Laplace noise with Sd 0.50 (ablation) | Wide ResNet · CIFAR-10 ℓ∞ PGD 20 | accuracy under attack: 0.384 -> 0.589 | Helped |
| Inject Laplace noise with Sd 0.50 (ablation) | Wide ResNet · CIFAR-10 natural images | accuracy: 0.958 -> 0.846 | Hurt |
| Use Normal noise, Sd 0.32, instead of Madry adversarial training | CIFAR-10 Wide ResNet · CIFAR-10 ℓ∞ PGD 20 | accuracy under attack: 0.456 -> 0.566 | Helped |
| Use Laplace noise, Sd 0.32, instead of Madry adversarial training | CIFAR-10 Wide ResNet · CIFAR-10 ℓ2 C&W 30 | accuracy under attack: 0.468 -> 0.502 | Helped |
| Use Normal noise, Sd 0.5, instead of Madry adversarial training | CIFAR-10 Wide ResNet · CIFAR-10 ℓ∞ PGD 20 | accuracy under attack: 0.456 -> 0.587 | Helped |
| Use Laplace noise, Sd 0.5, instead of Madry adversarial training | CIFAR-10 Wide ResNet · CIFAR-10 natural images | accuracy: 0.873 -> 0.846 | No clear effect |
Closest papers by meaning
- Adversarial Robustness via Deformable Convolution with Stochasticity (ICML 2025 · similarity 0.93)
- Treatment of Statistical Estimation Problems in Randomized Smoothing for Adversarial Robustness (NeurIPS 2024 · similarity 0.92)
- Adversarial Vulnerability of Randomized Ensembles (ICML 2022 · similarity 0.92)
- How Sampling Impacts the Robustness of Stochastic Neural Networks (NeurIPS 2022 · similarity 0.91)
- On the Role of Randomization in Adversarially Robust Classification (NeurIPS 2023 · similarity 0.91)
- Adversarial Robustness through Random Weight Sampling (NeurIPS 2023 · similarity 0.91)
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.