AI Paper Atlas

Research map / Adversarial robustness and security / Black-box and transfer attacks

Feature Importance-Aware Transferable Adversarial Attacks

Zhibo Wang, Hengchang Guo, Zhifei Zhang, Wenxin Liu, Zhan Qin, Kui Ren

ICCV 2021 · Adversarial robustness and security › Black-box and transfer attacks · approach: Transferable adversarial examples · 333 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 aggregate gradient (averaged over random-masked images) to the feature-level loss (ablation)FIA feature-level attack · Normal models, Inc-v3 sourceattack success rateImproved
use aggregate gradient instead of raw clean-image gradient (ablation)FIA · Normal models, Inc-v3 sourceattack success rateImproved
add feature importance weighting to unconstrained feature distortion (L1) (ablation)feature-level attack · ImageNet-compatibleattack success rateImproved
combine FIA with PIDIMFIA · Normal modelsattack success rate: 1% ~ 10% increase of success rateImproved
replace FDA/NRDM with FIA as feature-level attackNRDM, FDA · Normal modelsattack success rate: 60.3% -> 83.5%Improved
increase drop probability p_d to 0.5 (ablation)FIA · Inc-v3 source, varied targetsattack success rateHurt
use p_d between 0.2 and 0.3 for normal targets, around 0.1 for defense targets (ablation)FIA · Inc-v3 sourceattack success rateImproved
increase ensemble number N from 5 to 40 (ablation)FIA aggregate gradient · Inc-v3 sourceattack success rateImproved
attack middle layers instead of early or late layers (ablation)FIA · Inc-v3 and Vgg-16 sourcesattack success rateImproved
generate adversarial examples on an ensemble of models rather than single modelall attacks · Defense modelsattack success rate: 68.0% -> 90.9%Improved
add PITIDIM to FIAFIA · Defense models, Vgg-16 sourceattack success rate: 87.8% -> 74.7%Hurt
add PITIDIM to FIAFIA · Defense models, Inc-v3 sourceattack success rate: 54.5% -> 64.8%Improved

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