AI Paper Atlas

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

On Success and Simplicity: A Second Look at Transferable Targeted Attacks

Zhengyu Zhao, Zhuoran Liu, Martha Larson

NeurIPS 2021 · Adversarial robustness and security › Black-box and transfer attacks · approach: Transferable adversarial examples · 45 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
replace cross-entropy loss with logit lossTI-MI-DI simple attack · Res50->Dense121, 300 iterationstargeted success rate: 42.6 -> 72.5Improved
replace Po+Trip loss with logit lossTI-MI-DI simple attack · Res50->Dense121, 300 iterationstargeted success rate: 54.7 -> 72.5Improved
increase attack iterations from 20 to 300 (ablation)Logit attack · Res50->Dense121targeted success rate: 29.3 -> 72.5Improved
increase iterations from 20 to 100 for Po+Trip vs CE (ablation)Po+Trip · easy ensemble transfer averagetargeted success rate: 56.8 -> 84.4Hurt
replace CE with logit loss (ablation)simple attack · easy ensemble transfer, 100 itertargeted success rate: 87.7 -> 88.1No clear effect
replace CE with logit loss for lowest-ranked target classsimple attack · 1000th ranked targettargeted success rate: 25.1 -> 52.8Improved
replace CE with logit loss for 2nd-ranked target (ablation)simple attack · 2nd ranked targettargeted success rate: 89.9 -> 83.8Hurt
use logit loss attack instead of TTP at eps=16 single-modelLogit · Res50->D121targeted success rate: 79.6 -> 75.9Hurt
use logit loss attack instead of TTP at eps=8 ensembleLogit · D121-enstargeted success rate: 63.2 -> 92.6Improved
use logit loss rather than C&W loss (ablation)Logit attack · single-model transfertargeted success rateImproved
vary step size (ablation)Logit/CE/Po+Trip attacks · single-model transfertargeted success rateNo clear effect
use only DI vs only TI or MI (ablation)simple attack · targeted transfertargeted success rateImproved

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