Research map / Adversarial robustness and security / Certified and empirical robustness
Adversarial Attack Generation Empowered by Min-Max Optimization
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 uniform averaging with learned min-max domain weights for ensemble attack | PGD attack · MNIST ensemble l_inf | ASR_all: 48.17% -> 90.16% | Helped |
| learned domain weights over averaging for universal perturbation | PGD attack · CIFAR-10 universal perturbation K=5 | ASR_all: 35.21% improvement | Helped |
| learned domain weights over averaging for universal perturbation with K=10 | PGD attack · CIFAR-10 universal perturbation K=10 | ASR_avg: 71.22 -> 65.64 | Hurt |
| learned weights over EOT-style average over transformations | EOT attack · CIFAR-10 data transformations | ASR_all: 9.39% average lift | Helped |
| use min-max weights in generalized adversarial training instead of MSD | adversarial training against multiple lp attacks · MNIST and CIFAR-10 | robust accuracy: 4~6% (MNIST), 2% (CIFAR-10) | Helped |
| learn weights dynamically instead of using static converged weights (ablation) | APGDA ensemble attack · MNIST ensemble | ASR_all | Helped |
| learned weights vs clipped C&W loss heuristic (ablation) | ensemble attack · MNIST/CIFAR-10 ensemble | ASR_all | Helped |
Closest papers by meaning
- Are Adversarial Examples Created Equal? A Learnable Weighted Minimax Risk for Robustness under Non-uniform Attacks (AAAI 2021 · similarity 0.91)
- Adversarial Training and Robustness for Multiple Perturbations (NeurIPS 2019 · similarity 0.90)
- SuperDeepFool: a new fast and accurate minimal adversarial attack (NeurIPS 2024 · similarity 0.88)
- Clarifying the Behavior and the Difficulty of Adversarial Training (AAAI 2024 · similarity 0.88)
- Improving Generalization of Universal Adversarial Perturbation via Dynamic Maximin Optimization (AAAI 2025 · similarity 0.88)
- On the Duality Between Sharpness-Aware Minimization and Adversarial Training (ICML 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.