AI Paper Atlas

Research map / Adversarial robustness and security / Attacks on text and graphs

Enhancing Adversarial Robustness via Score-Based Optimization

Boya Zhang, Weijian Luo, Zhihua Zhang

NeurIPS 2023 · Adversarial robustness and security › Attacks on text and graphs · approach: Diffusion-based adversarial purification · 6 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 ScoreOpt-N instead of ScoreOpt-O (ablation)ScoreOpt · CIFAR-10 BPDA+EOTrobust accuracy (%): 82.85 -> 90.59Improved
increase diffusion purification denoising steps from 1 to 80 (ablation)Diffusion purification · CIFAR-100 BPDA+EOTrobust accuracy (%): 40.03 -> 47.85No clear effect
replace multi-step diffusion purification with ScoreOpt-OScoreOpt · CIFAR-100 BPDA+EOTrobust accuracy (%): 48.83 -> 66.99Improved
use approximate one-shot-denoising gradient instead of exact gradient for PGD+EOT (ablation)ScoreOpt · CIFAR-10 512 subset l_2attack success rate (%): 33.99 -> 34.11No clear effect
combine ScoreOpt with adversarially trained classifierScoreOpt · CIFAR-10 PGD+EOT l_infrobust accuracy (%): 71.48 -> 72.57Improved
combine ScoreOpt with adversarially trained classifierScoreOpt · CIFAR-10 PGD+EOT l_2robust accuracy (%): 87.11 -> 89.06Improved
increase number of optimization steps (ablation)ScoreOpt · CIFAR-10 BPDA+EOTstandard and robust accuracyImproved
vary number of EOT replicates (ablation)ScoreOpt · CIFAR-10 PGD+EOT and BPDA+EOTaccuracyNo clear effect
use WRN-70-16 classifier instead of WRN-28-10 (ablation)ScoreOpt-N · CIFAR-10 BPDA+EOTrobust accuracy (%): 90.59 -> 91.60Improved
change noise level to [0.60,0.60] from [0.50,0.50] (ablation)ScoreOpt-N · CIFAR-10 BPDA+EOTrobust accuracy (%): 90.59 -> 89.26Hurt
change noise level to [0.40,0.40] from [0.50,0.50] (ablation)ScoreOpt-N · CIFAR-10 BPDA+EOTrobust accuracy (%): 90.59 -> 92.19Improved
change noise level to [0.30,0.50] from [0.40,0.60] (ablation)ScoreOpt-O · CIFAR-10 BPDA+EOTrobust accuracy (%): 82.85 -> 79.30Hurt

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