Research map / Adversarial robustness and security / Attacks on text and graphs
Hilbert-Based Generative Defense for Adversarial Examples
Yang Bai, Yan Feng, Yisen Wang, Tao Dai, Shu-Tao Xia, Yong Jiang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 raster scan with Hilbert scan in PixelDefend (ablation) | PixelDefend · CIFAR-10 Obfuscated Gradient attack | accuracy: 0.15 -> 0.39 | Improved |
| replace raster scan with Hilbert scan in PixelDefend (ablation) | PixelDefend · CIFAR-10 white-box PGD eps=16 | accuracy: 0.56 -> 0.64 | Improved |
| replace raster scan with Hilbert scan in PixelDefend (ablation) | PixelDefend · CIFAR-10 white-box CW eps=2 | accuracy: 0.82 -> 0.86 | Improved |
| replace raster scan with Hilbert scan in PixelDefend (ablation) | PixelDefend · CIFAR-10 white-box PGD eps=2 | accuracy: 0.82 -> 0.82 | No clear effect |
| replace raster scan with Hilbert scan in PixelDefend (ablation) | PixelDefend · CIFAR-10 white-box FGSM eps=8 | accuracy: 0.79 -> 0.78 | Hurt |
| ensemble 4 Hilbert curves instead of one (ablation) | HPD · CIFAR-10 Obfuscated Gradient attack | accuracy: 0.39 -> 0.49 | Improved |
| ensemble 8 Hilbert curves instead of one (ablation) | HPD · CIFAR-10 Obfuscated Gradient attack | accuracy: 0.39 -> 0.52 | Improved |
| ensemble 8 Hilbert curves instead of one (ablation) | HPD · CIFAR-10 white-box PGD eps=8 | accuracy: 0.75 -> 0.77 | Improved |
| ensemble 8 Hilbert curves instead of one (ablation) | HPD · CIFAR-10 white-box PGD eps=16 | accuracy: 0.64 -> 0.66 | Improved |
| ensemble 8 raster-scan PixelDefend models instead of one (ablation) | PixelDefend · CIFAR-10 white-box PGD eps=16 | accuracy: 0.56 -> 0.59 | Improved |
| ensemble 8 Hilbert curves instead of one (ablation) | HPD · CIFAR-10 Obfuscated Gradient attack | average epochs to break defense: 14 -> 17 | Improved |
| use Hilbert scan in PixelDefend under black-box transfer attack (ablation) | PixelDefend · CIFAR-10 black-box FGSM eps=16 | accuracy: 0.66 -> 0.68 | Improved |
Closest papers by meaning
- PixelDefend: Leveraging Generative Models to Understand and Defend against Adversarial Examples (ICLR 2018 · similarity 0.89)
- Defense-GAN: Protecting Classifiers Against Adversarial Attacks Using Generative Models (ICLR 2018 · similarity 0.85)
- A Unified, Resilient, and Explainable Adversarial Patch Detector (CVPR 2025 · similarity 0.84)
- PAD: Patch-Agnostic Defense against Adversarial Patch Attacks (CVPR 2024 · similarity 0.84)
- Pixel Is Not a Barrier: An Effective Evasion Attack for Pixel-Domain Diffusion Models (AAAI 2025 · similarity 0.84)
- Stop Walking in Circles! Bailing Out Early in Projected Gradient Descent (CVPR 2025 · similarity 0.84)
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.
