Research map / Adversarial robustness and security / Certified and empirical robustness
Improving l1-Certified Robustness via Randomized Smoothing by Leveraging Box Constraints
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 box-constrained certification (volume of overlap of l-inf balls restricted to [0,1]^d) instead of unconstrained R^d certification | uniform smoothing noise · CIFAR-10 | certified l1 robust accuracy vs radius: significant improvements in certified l1 radius | Helped |
| Use box-constrained certification for splitting noise instead of standard certification | splitting noise (independent) · CIFAR-10 | certified l1 robust accuracy vs radius: significant increase in certified robustness | Helped |
| Use multiclass certification (Prop. 2.12) instead of one-vs-all binary certification (ablation) | uniform++ and split++ · CIFAR-10, sigma=2.5 | certified l1 robust accuracy vs radius: multiclass approach (almost strictly) outperforms the binary one | Helped |
| Stability training instead of standard cross-entropy training of noise-augmented models (ablation) | uniform and splitting noise models · ImageNet and CIFAR-10 | certified l1 robustness | Helped |
| Stability training instead of standard training for splitting noise (ablation) | splitting noise models · CIFAR-10 / ImageNet | certified l1 robustness | Mixed |
| Increase number of noise samples from n=100 vs 100000 for certification (ablation) | uniform smoothing certification · CIFAR-10 (lambda=6.92) | certified robust radius | No clear effect |
Closest papers by meaning
- Mitigating the Curse of Dimensionality for Certified Robustness via Dual Randomized Smoothing (ICLR 2024 · similarity 0.92)
- Incremental Randomized Smoothing Certification (ICLR 2024 · similarity 0.91)
- Certified Adversarial Robustness via Randomized $\alpha$-Smoothing for Regression Models (NeurIPS 2024 · similarity 0.90)
- Treatment of Statistical Estimation Problems in Randomized Smoothing for Adversarial Robustness (NeurIPS 2024 · similarity 0.90)
- Effects of Exponential Gaussian Distribution on (Double Sampling) Randomized Smoothing (ICML 2024 · similarity 0.89)
- Adaptive Randomized Smoothing: Certified Adversarial Robustness for Multi-Step Defences (NeurIPS 2024 · similarity 0.89)
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.