Research map / Adversarial robustness and security / Certified and empirical robustness
Robust Unlearnable Examples: Protecting Data Privacy Against Adversarial Learning
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 |
|---|---|---|---|
| Generate protective noise by minimizing adversarial training loss (REM) instead of using standard EM noise | ResNet-18 trained on protected data · CIFAR-10, adversarial training 2/255 | test accuracy (%): 71.43 -> 25.41 | Helped |
| Generate protective noise by minimizing adversarial training loss (REM) instead of using standard EM noise | ResNet-18 trained on protected data · CIFAR-100, adversarial training 2/255 | test accuracy (%): 68.49 -> 7.03 | Helped |
| Generate protective noise by minimizing adversarial training loss (REM) instead of using standard EM noise | ResNet-18 trained on protected data · ImageNet Subset, adversarial training 2/255 | test accuracy (%): 71.74 -> 27.06 | Helped |
| Add EOT when generating REM noise (ablation) | REM on CIFAR-10 · CIFAR-10, adversarial training 2/255 | test accuracy (%): 91.69 -> 25.41 | Helped |
| Add EOT when generating REM noise (ablation) | REM on CIFAR-100 · CIFAR-100, adversarial training 2/255 | test accuracy (%): 68.85 -> 7.03 | Helped |
| Add EOT when generating REM noise (ablation) | REM on CIFAR-10 · CIFAR-10, adversarial training 4/255 | test accuracy (%): 89.41 -> 48.16 | Helped |
| Add EOT when generating REM noise (ablation) | REM on CIFAR-100 · CIFAR-100, adversarial training 4/255 | test accuracy (%): 63.79 -> 27.10 | Helped |
| Increase adversarial-training radius against EM-protected data | EM noise · CIFAR-10 | test accuracy (%): 13.20 -> 88.62 | Hurt |
| Use REM with DenseNet-121 rather than the other evaluated architectures | REM-protected data · CIFAR-10 and CIFAR-100 | test accuracy (%): 77.82; 54.19 | Mixed |
Closest papers by meaning
- Stable Unlearnable Example: Enhancing the Robustness of Unlearnable Examples via Stable Error-Minimizing Noise (AAAI 2024 · similarity 0.94)
- Detection and Defense of Unlearnable Examples (AAAI 2024 · similarity 0.91)
- Re-thinking Data Availability Attacks Against Deep Neural Networks (CVPR 2024 · similarity 0.90)
- Rethinking Adversarial Robustness in the Context of the Right to be Forgotten (ICML 2024 · similarity 0.88)
- DataFreeShield: Defending Adversarial Attacks without Training Data (ICML 2024 · similarity 0.88)
- Ungeneralizable Examples (CVPR 2024 · similarity 0.87)
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.