Research map / Adversarial robustness and security / Certified and empirical robustness
Adversarial Training and Robustness for Multiple 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 |
|---|---|---|---|
| adversarially train on multiple perturbation types (max strategy) instead of a single one | CIFAR10 classifier · CIFAR10 linf+l1 | accuracy on union: 53.1 -> 61.1 | Helped |
| train on multiple perturbations | MNIST classifier · MNIST linf | linf accuracy: 91.1 -> 71.7 | Hurt |
| train on multiple perturbations | MNIST classifier · MNIST clean | clean accuracy: 99.1 -> 97.2 | Hurt |
| use max strategy instead of avg strategy (ablation) | multi-perturbation adversarial training · CIFAR10 linf+l1 | accuracy on union: 59.4 -> 61.1 | No clear effect |
| train with stronger linf bound 8/255 instead of 4/255 (ablation) | Adv_max · CIFAR10 linf+l1 | accuracy on union: 61.1 -> 54.3 | Hurt |
| evaluate with affine combination of linf and RT instead of union | Adv_max · MNIST linf+RT | accuracy: 83.8 -> 62.6 | Hurt |
| evaluate with affine combination of linf and l1 instead of union | Adv_max · CIFAR10 linf+l1 | accuracy: 61.1 -> 58.0 | No clear effect |
Closest papers by meaning
- Adversarial Robustness Against the Union of Multiple Perturbation Models (ICML 2020 · similarity 0.94)
- On the Vulnerability of Adversarially Trained Models Against Two-faced Attacks (ICLR 2024 · similarity 0.92)
- Certified Adversarial Robustness with Additive Noise (NeurIPS 2019 · similarity 0.92)
- How Many Perturbations Break This Model? Evaluating Robustness Beyond Adversarial Accuracy (ICML 2023 · similarity 0.92)
- Adversarial Training Can Provably Improve Robustness: Theoretical Analysis of Feature Learning Process Under Structured Data (ICLR 2025 · similarity 0.91)
- Adversarial Robustness with Semi-Infinite Constrained Learning (NeurIPS 2021 · similarity 0.91)
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.