AI Paper Atlas

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

Effective Targeted Attacks for Adversarial Self-Supervised Learning

Minseon Kim, Hyeonjeong Ha, Sooel Son, Sung Ju Hwang

NeurIPS 2023 · Adversarial robustness and security › Attacks on text and graphs · approach: Defenses against universal perturbations · 1 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 untargeted attack with TARO targeted attackBYOL · CIFAR-10 linear evalPGD robust accuracy: 16.20 -> 31.20Improved
replace untargeted attack with TARO targeted attackSimSiam · CIFAR-10 linear evalPGD robust accuracy: 32.28 -> 44.71Improved
replace untargeted attack with TARO targeted attackSimSiam · CIFAR-10 linear evalclean accuracy: 71.78 -> 74.87Improved
replace untargeted attack with TARO targeted attackBYOL · CIFAR-10 linear evalAutoAttack robust accuracy: 0.01 -> 22.01Improved
add TARO to contrastive adversarial SSLRoCL · CIFAR-10 linear evalPGD robust accuracy: 42.89 -> 45.37Improved
add TARO to contrastive adversarial SSLACL · CIFAR-10 linear evalPGD / AutoAttack accuracy: 39.37 -> 39.71 PGD; 35.97 -> 35.81 AutoAttackNo clear effect
add TARO on larger datasetSimSiam* · CIFAR-100 linear evalPGD robust accuracy: 13.34 -> 22.18Improved
add TARO on larger datasetRoCL · CIFAR-100 linear evalPGD robust accuracy: 17.17 -> 18.91Improved
add TARO for transfer from CIFAR-100 to CIFAR-10RoCL · CIFAR-10 transferclean accuracy: 73.93 -> 65.21Hurt
use random in-batch target instead of untargeted attack (ablation)SimSiam · CIFAR-10PGD robust accuracy: 32.28 -> 42.85Improved
use score-selected target instead of random target (ablation)SimSiam · CIFAR-10PGD robust accuracy: 42.85 -> 44.71Improved
use score-selected target instead of random target (ablation)RoCL · CIFAR-10PGD robust accuracy: 43.45 -> 45.37Improved

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