Research map / Adversarial robustness and security / Certified and empirical robustness
Certified Adversarial Robustness via Randomized Smoothing
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 |
|---|---|---|---|
| Use Gaussian randomized smoothing with σ=0.25 instead of σ=0.50 | ResNet-50 smoothed classifier · ImageNet, ℓ2 radius 0.5 | approximate certified accuracy (%): 49 | Helped |
| Increase the certified ℓ2 radius | ImageNet randomized smoothing · ImageNet, radii 0.5 to 3.0 | approximate certified accuracy (%): 49 -> 37 -> 19 -> 12 | Hurt |
| Use a larger noise level for certification | ImageNet randomized smoothing · ImageNet | standard accuracy (%): 67 -> 57 -> 44 | Hurt |
| Use the paper's tight Gaussian smoothing guarantee instead of prior looser guarantees | Randomized smoothing · ImageNet | certified accuracy | Helped |
| Use a 110-layer ResNet base classifier rather than the small ResNet baseline architecture | Randomized smoothing · CIFAR-10 and ImageNet certified accuracy across radii | certified accuracy | Helped |
| Use randomized smoothing with the same small ResNet architecture as the baseline | Small ResNet · Certified accuracy at radii other than the smallest | certified accuracy | Helped |
| Increase Monte Carlo samples for Predict from n=100 to n=1000 | ImageNet randomized smoothing · ImageNet, σ=0.25 | abstention rate (%): 12% -> 4% | Helped |
| Increase Monte Carlo samples for Predict from n=1000 to n=10,000 | ImageNet randomized smoothing · ImageNet, σ=0.25 | abstention rate (%): 4% -> 1% | Helped |
| Attack the smoothed classifier at a larger radius, from 1.5R to 2R | ImageNet randomized smoothing · ImageNet adversarial attack | attack success rate (%): 17% -> 53% | Mixed |
Closest papers by meaning
- Treatment of Statistical Estimation Problems in Randomized Smoothing for Adversarial Robustness (NeurIPS 2024 · similarity 0.92)
- Certified Adversarial Robustness via Randomized $\alpha$-Smoothing for Regression Models (NeurIPS 2024 · similarity 0.91)
- Mitigating the Curse of Dimensionality for Certified Robustness via Dual Randomized Smoothing (ICLR 2024 · similarity 0.91)
- Effects of Exponential Gaussian Distribution on (Double Sampling) Randomized Smoothing (ICML 2024 · similarity 0.91)
- Provably Cost-Sensitive Adversarial Defense via Randomized Smoothing (ICML 2025 · similarity 0.90)
- DRF: Improving Certified Robustness via Distributional Robustness Framework (AAAI 2024 · 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.