Research map / Adversarial robustness and security / Black-box and transfer attacks
Black-box Certification and Learning under Adversarial Perturbations
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 |
|---|---|---|---|
| Add a margin-class finite-VC condition (margin class has finite VC-dimension) to the proper learning setup | proper robust learning of VC classes · distribution-free agnostic robust learning | learnability (sample complexity O(D log D + log(1/δ))/ε², D = VC(H)+VC(H_mar)) | Helped |
| Give the proper learner oracle access to margin weights P(mar_h) (margin oracle) under robust realizability | proper robust learning of H · robust-realizable distributions | proper learnability with labeled sample complexity Õ((VC(H)+log(1/δ))/ε) | Helped |
| Give proper learner error and margin oracles without robust realizability (non-realizable setting) | proper robust learning · exact error and margin oracles | proper learnability of robust loss minimizer | Hurt |
| Require a perfect, proper, efficient adversary (query complexity linear in sample size) for properly compressible classes | robust learning sample complexity · robust realizable case | sample complexity O(t log^2 t), t=VC(H)/ε² | Helped |
| Relax certification to a tolerant version (accept answers in [L^U - eps, L^V + eps]) with a gap gamma between perturbation radii | black-box certification of half-spaces in R^2 · l2 balls radius 1 vs 1+gamma | query complexity O(1/(sqrt(gamma) eps^2)) | Helped |
Closest papers by meaning
- Understanding the Robustness of Randomized Feature Defense Against Query-Based Adversarial Attacks (ICLR 2024 · similarity 0.89)
- Adversarial examples from computational constraints (ICML 2019 · similarity 0.89)
- Learning Better Certified Models from Empirically-Robust Teachers (NeurIPS 2026 · similarity 0.89)
- Robust and Stable Black Box Explanations (ICML 2020 · similarity 0.89)
- Reducing Adversarially Robust Learning to Non-Robust PAC Learning (NeurIPS 2020 · similarity 0.89)
- Probably Approximately Global Robustness Certification (ICML 2025 · 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.