Research map / Adversarial robustness and security / Attacks on text and graphs
Diffusion-Based Adversarial Sample Generation for Improved Stealthiness and Controllability
Haotian Xue, Alexandre Araujo, Bin Hu, Yongxin Chen
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 |
|---|---|---|---|
| Use diffusion-guided Diff-PGD instead of PGD | ResNet-50 attack examples · ImageNet adversarial purification, (+P)ResNet50 | success rate: 0.35 -> 0.88 | Improved |
| Use diffusion-guided Diff-PGD instead of PGD | ResNet-50 attack examples · ImageNet adversarial purification, (+P)ResNet101 | success rate: 0.18 -> 0.38 | Improved |
| Use diffusion-guided Diff-PGD instead of PGD | ResNet-50 attack examples · ImageNet adversarial purification, (+P)ResNet18 | success rate: 0.26 -> 0.36 | Improved |
| Use diffusion-guided Diff-PGD instead of PGD | ResNet-50 attack examples · ImageNet adversarial purification, (+P)WRN50 | success rate: 0.20 -> 0.32 | Improved |
| Use diffusion-guided Diff-PGD instead of PGD | ResNet-50 attack examples · ImageNet adversarial purification, (+P)WRN101 | success rate: 0.17 -> 0.28 | Improved |
| Approximate the diffusion Jacobian gradient (ablation) | Diff-PGD · ImageNet global attack, ResNet-50, ε=16/255, n=10 | success rate: 99.6 | No clear effect |
| Approximate the diffusion Jacobian gradient (ablation) | Diff-PGD · Global attack, K=2, n=10 | VRAM (G): ~18 -> ~4 | Improved |
| Approximate the diffusion Jacobian gradient (ablation) | Diff-PGD · Global attack, K=2, n=10 | speed (sec/sample): ~8 -> ~4 | Improved |
| Separate style transfer from adversarial optimization, then apply Diff-PGD | Style-customized attacks · Style-guided adversarial sample generation | optimization stability | Improved |
| Use Diff-PGD rather than AdvCam for style-customized attacks | Style-customized attacks · Style-guided adversarial sample generation | stealthiness | Improved |
Closest papers by meaning
- DiffAdvMAP: Flexible Diffusion-Based Framework for Generating Natural Unrestricted Adversarial Examples (ICML 2025 · similarity 0.90)
- DiffHammer: Rethinking the Robustness of Diffusion-Based Adversarial Purification (NeurIPS 2024 · similarity 0.90)
- MimicDiffusion: Purifying Adversarial Perturbation via Mimicking Clean Diffusion Model (CVPR 2024 · similarity 0.88)
- LoRID: Low-Rank Iterative Diffusion for Adversarial Purification (AAAI 2025 · similarity 0.88)
- ADBM: Adversarial Diffusion Bridge Model for Reliable Adversarial Purification (ICLR 2025 · similarity 0.88)
- AdvAD: Exploring Non-Parametric Diffusion for Imperceptible Adversarial Attacks (NeurIPS 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.
