Research map / Adversarial robustness and security / Certified and empirical robustness
Certified Defense to Image Transformations via Randomized Smoothing
Marc Fischer, Maximilian Baader, Martin Vechev
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 |
|---|---|---|---|
| apply BaseSPT smoothing to the base classifier | MNIST classifier · MNIST rotation 30deg, worst-of-100 | adversarial accuracy: 0.73 -> 0.99 | Improved |
| apply BaseSPT smoothing to the base classifier | CIFAR-10 classifier · CIFAR-10 rotation 30deg | adversarial accuracy: 0.26 -> 0.85 | Improved |
| apply BaseSPT smoothing to the base classifier | ImageNet classifier · ImageNet rotation 30deg | adversarial accuracy: 0.56 -> 0.76 | Improved |
| apply BaseSPT smoothing under translation | MNIST classifier · MNIST translation 4px | adversarial accuracy: 0.03 -> 0.53 | Improved |
| increase n_gamma from 50 to 200 | DistSPT global · CIFAR-10 rotation | certified accuracy at 0deg: 0.31 -> 0.32 | Improved |
| increase n_gamma from 50 to 200 | DistSPT global · CIFAR-10 rotation | run time (s): 89.75 -> 351.47 | Hurt |
| raise E from 1.20 to 1.35 (ablation) | DistSPT global · RImageNet rotation | certified accuracy at 0deg: 0.74 -> 0.64 | Hurt |
| raise E from 0.95 to 1.35 (ablation) | DistSPT global · ImageNet rotation | certified accuracy at 0deg: 0.30 -> 0.16 | Hurt |
| use per-image error bound E (DistSPT^x) instead of global E | DistSPT · CIFAR-10 rotation 0deg | certified accuracy: 0.31 -> 0.35 | Improved |
| use per-image error bound E (DistSPT^x) instead of global E | DistSPT · ImageNet rotation 0deg | certified accuracy: 0.23 -> 0.31 | Improved |
| add vignetting to base model training/eval (ablation) | DistSPT global base model · CIFAR-10 rotation | base accuracy: 0.34 -> 0.56 | Improved |
| add Gaussian blur (ablation) | DistSPT global base model · ImageNet rotation | base accuracy: 0.38 -> 0.32 | Hurt |
Closest papers by meaning
- Improving l1-Certified Robustness via Randomized Smoothing by Leveraging Box Constraints (ICML 2023 · similarity 0.88)
- Certified Robustness for Deep Equilibrium Models via Serialized Random Smoothing (NeurIPS 2024 · similarity 0.85)
- Effects of Exponential Gaussian Distribution on (Double Sampling) Randomized Smoothing (ICML 2024 · similarity 0.85)
- Mitigating the Curse of Dimensionality for Certified Robustness via Dual Randomized Smoothing (ICLR 2024 · similarity 0.84)
- Towards Large Certified Radius in Randomized Smoothing Using Quasiconcave Optimization (AAAI 2024 · similarity 0.84)
- Pixel-level Certified Explanations via Randomized Smoothing (ICML 2025 · similarity 0.83)
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.
