AI Paper Atlas

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

NeurIPS 2023 · Adversarial robustness and security › Attacks on text and graphs · approach: Diffusion-based adversarial purification · 10 citations (OpenAlex)

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 changedWhereMeasuredEffect
Use diffusion-guided Diff-PGD instead of PGDResNet-50 attack examples · ImageNet adversarial purification, (+P)ResNet50success rate: 0.35 -> 0.88Improved
Use diffusion-guided Diff-PGD instead of PGDResNet-50 attack examples · ImageNet adversarial purification, (+P)ResNet101success rate: 0.18 -> 0.38Improved
Use diffusion-guided Diff-PGD instead of PGDResNet-50 attack examples · ImageNet adversarial purification, (+P)ResNet18success rate: 0.26 -> 0.36Improved
Use diffusion-guided Diff-PGD instead of PGDResNet-50 attack examples · ImageNet adversarial purification, (+P)WRN50success rate: 0.20 -> 0.32Improved
Use diffusion-guided Diff-PGD instead of PGDResNet-50 attack examples · ImageNet adversarial purification, (+P)WRN101success rate: 0.17 -> 0.28Improved
Approximate the diffusion Jacobian gradient (ablation)Diff-PGD · ImageNet global attack, ResNet-50, ε=16/255, n=10success rate: 99.6No clear effect
Approximate the diffusion Jacobian gradient (ablation)Diff-PGD · Global attack, K=2, n=10VRAM (G): ~18 -> ~4Improved
Approximate the diffusion Jacobian gradient (ablation)Diff-PGD · Global attack, K=2, n=10speed (sec/sample): ~8 -> ~4Improved
Separate style transfer from adversarial optimization, then apply Diff-PGDStyle-customized attacks · Style-guided adversarial sample generationoptimization stabilityImproved
Use Diff-PGD rather than AdvCam for style-customized attacksStyle-customized attacks · Style-guided adversarial sample generationstealthinessImproved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy