Research map / Adversarial robustness and security / Certified and empirical robustness
Certified Adversarial Robustness Under the Bounded Support Set
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 |
|---|---|---|---|
| replace f-divergence iterative certification with closed-form W/TV certification objective | Gaussian and uniform smoothing on CIFAR-10 · CIFAR-10 test set | certification runtime: 10 mins vs 90 mins (20 steps) and 3 hours (50 steps) | Helped |
| use uniform smoothing on l2 or l-infinity ball instead of Gaussian (ablation) | uniform smoothing measures · CIFAR-10 l2 certification | certified radius at cut-off: Gaussian radius about 50 times larger than uniform | Hurt |
| increase sample amount from 100 to 10000 (ablation) | Gaussian smoothing · CIFAR-10 certified accuracy | certified accuracy | No clear effect |
| increase sample amount from 100 to 10000 (ablation) | uniform smoothing measures · CIFAR-10 certified accuracy | certified accuracy: minor progress when samples are overly abundant | Helped |
| increase Gaussian sigma (smoothing variance) (ablation) | Gaussian smoothing · CIFAR-10 certified accuracy | initial certified accuracy vs robustness radius | Mixed |
Closest papers by meaning
- A FRAMEWORK FOR ROBUSTNESS CERTIFICATION OF SMOOTHED CLASSIFIERS USING F-DIVERGENCES (ICLR 2020 · similarity 0.93)
- Certified Adversarial Robustness via Randomized $\alpha$-Smoothing for Regression Models (NeurIPS 2024 · similarity 0.90)
- DRF: Improving Certified Robustness via Distributional Robustness Framework (AAAI 2024 · similarity 0.90)
- Probably Approximately Global Robustness Certification (ICML 2025 · similarity 0.89)
- Certified Robustness for Deep Equilibrium Models via Serialized Random Smoothing (NeurIPS 2024 · similarity 0.89)
- Incremental Randomized Smoothing Certification (ICLR 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.