Research map / Adversarial robustness and security / Attacks on text and graphs
Defense Against Universal 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 |
|---|---|---|---|
| Train the PRN and detector on perturbations of one norm type, then test on the other type | PRN + detector · ILSVRC 2012 val, GoogLeNet, cross-norm tests | Defense rate (%): 97.5; 94.9; 94.3; 91.6 | Mixed |
| Train the PRN and detector on perturbations of one norm type, then test on the other type | PRN + detector · ILSVRC 2012 val, CaffeNet, cross-norm tests | Detection rate (%): 84.2; 84.0; 97.9; 98.0 | Mixed |
| Train the PRN and detector on perturbations of one norm type, then test on the other type | PRN + detector · ILSVRC 2012 val, VGG-F, cross-norm tests | Defense rate (%): 90.0; 85.9; 93.7; 89.1 | Mixed |
| Train PRN using synthetic perturbations in addition to generated perturbations | PRN · Universal perturbation training; experiments across targeted networks | PRN performance | Helped |
| Train the rectifier on one targeted network and transfer it to another | PRN · ILSVRC 2012 val, cross-network defense | Defense rate (%): 95.5; 91.5; 82.4; 94.8; 96.2; 77.3; 88.3; 87.3; 97.4 | Mixed |
Closest papers by meaning
- Wide Two-Layer Networks can Learn from Adversarial Perturbations (NeurIPS 2024 · similarity 0.91)
- Theoretical Understanding of Learning from Adversarial Perturbations (ICLR 2024 · similarity 0.89)
- Contrasting Adversarial Perturbations: The Space of Harmless Perturbations (AAAI 2025 · similarity 0.89)
- Democratic Training Against Universal Adversarial Perturbations (ICLR 2025 · similarity 0.89)
- Generate Universal Adversarial Perturbations for Few-Shot Learning (NeurIPS 2024 · similarity 0.88)
- Improving Generalization of Universal Adversarial Perturbation via Dynamic Maximin Optimization (AAAI 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.