AI Paper Atlas

Research map / Adversarial robustness and security / Certified and empirical robustness

Removing Adversarial Noise in Class Activation Feature Space

Dawei Zhou, Nannan Wang, Chunlei Peng, Xinbo Gao, Xiaoyu Wang, Jun Yu, Tongliang Liu

ICCV 2021 · Adversarial robustness and security › Certified and empirical robustness · approach: Adversarial training refinements · 38 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
add class activation feature denoiser CAFD against AA-NVGG-19 defense · CIFAR-10 AA-Nerror rate: 30.61 -> 11.80Improved
add CAFD against BPDA adaptive attack, versus APE and HGDVGG-19 defense · BPDA PGD-Nerror rate: reduced by 49.85% and 39.83% on averageImproved
CAFD under white-box adaptive attack (defense leaked)VGG-19 defense · White-box DDN-Nerror rate: 97.85 -> 93.23No clear effect
CAFD' as local defense under gray-box PGD-NVGG-19 defense · Gray-box PGD-Nerror rate: 65.03 -> 56.76Improved
CAFD under gray-box DDN-NVGG-19 defense · Gray-box DDN-Nerror rate: 10.93 -> 10.90No clear effect
transfer CAFD to ResNet-50 targetResNet-50 · SVHN PGD-Nerror rate: 100 -> 11.01Improved
transfer CAFD to Wide-ResNet targetWide-ResNet · SVHN PGD-Nerror rate: 97.71 -> 21.33Improved
add class activation feature loss L_caf (ablation)CAFD · CIFAR-10 attacksaccuracyImproved
add adversarial (discriminator) loss L_adv (ablation)CAFD · CIFAR-10 attacksaccuracyImproved
use class activation feature loss instead of pixel-wise MSE loss (ablation)CAFD · CIFAR-10 attacksaccuracyImproved
use default CAFD loss form rather than variant CAFD' (ablation)CAFD · SVHN PGD-Nerror rate: 15.91 -> 14.64Improved
increase perturbation budget beyond 12/255 (ablation)CAFD · CIFAR-10 PGD/AAaccuracyHurt

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