Research map / Adversarial robustness and security / Certified and empirical robustness
Tight Certificates of Adversarial Robustness for Randomly Smoothed Classifiers
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 the discrete certificate instead of the Gaussian certificate on the same discrete-smoothed CNN | Binarized MNIST CNN · Binarized MNIST test | average certified ℓ0 radius μ(R): 1.799 -> 3.456 | Helped |
| Use discrete smoothing and certification instead of Gaussian smoothing and certification | CNN · Binarized MNIST test | ACC@1: 0.884 -> 0.921 | Helped |
| Use discrete smoothing and certification instead of Gaussian smoothing and certification | ResNet50 · ImageNet validation | ACC@1: 0.372 -> 0.538 | Helped |
| Set the discrete-noise parameter α to 0.5, 0.4, 0.3, 0.2, or 0.1 rather than using one fixed setting (ablation) | Discrete-smoothed ResNet50 · ImageNet validation | ACC@r across radii: r=1: 0.412, 0.538, 0.516, 0.448, 0.314; r=7: 0.000, 0.000, 0.000, 0.176, 0.100 | Mixed |
| Use the proposed randomly smoothed decision tree instead of a vanilla decision tree | Decision tree · Bace | robust AUC | Helped |
| Restrict the classifier to a decision tree that uses each input coordinate at most once | Discrete-smoothed classifier · Binary discrete inputs | certified robustness | Helped |
| Use the discrete distribution instead of Gaussian smoothing for ℓ0 certification | Randomly smoothed classifier · Binarized MNIST test | ACC@2: 0.557 -> 0.774 | Helped |
Closest papers by meaning
- Certified Adversarial Robustness via Randomized Smoothing (ICML 2019 · similarity 0.91)
- Provably Robust Deep Learning via Adversarially Trained Smoothed Classifiers (NeurIPS 2019 · similarity 0.91)
- A FRAMEWORK FOR ROBUSTNESS CERTIFICATION OF SMOOTHED CLASSIFIERS USING F-DIVERGENCES (ICLR 2020 · similarity 0.90)
- Probably Approximately Global Robustness Certification (ICML 2025 · similarity 0.90)
- Treatment of Statistical Estimation Problems in Randomized Smoothing for Adversarial Robustness (NeurIPS 2024 · similarity 0.89)
- Consistent Adversarially Robust Linear Classification: Non-Parametric Setting (ICML 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.