AI Paper Atlas

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

On the Robustness of Vision Transformers to Adversarial Examples

Kaleel Mahmood, Rigel Mahmood, Marten van Dijk

ICCV 2021 · Adversarial robustness and security › Black-box and transfer attacks · approach: Transfer attacks on vision transformers · 219 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 a ViT instead of a CNN or BiT modelsingle classifier · CIFAR-10/ImageNet white-boxrobust accuracy under PGD and MIM: less than 6%No clear effect
use a ViT instead of a CNNsingle classifier · CIFAR-10/ImageNet white-box C&W and APGDrobust accuracy: 0%No clear effect
evaluate adversarial examples crafted on BiT-50x1 against ViT-L-16BiT-M-50x1 adversarial examples · CIFAR-10, CIFAR-100, ImageNet transferabilitytransferability: 5.7%, 15.5% and 11.8%Hurt
evaluate adversarial examples crafted on ViT-L-16 against BiT-M-50x1ViT-L-16 adversarial examples · CIFAR-10, CIFAR-100, ImageNet transferabilitytransferability: 42.5%, 47.6% and 34.3%Hurt
tune SAGA coefficients toward the ViT gradient instead of equal 0.5/0.5 weights (ablation)SAGA · CIFAR-10 ViT/BiT ensembleaverage robust accuracy: 47.5% -> 26.0%Improved
tune SAGA coefficients instead of equal 0.5/0.5 weights (ablation)SAGA · CIFAR-100 ViT/BiT ensembleaverage robust accuracy: 26.3% -> 15.6%Improved
tune SAGA coefficients instead of equal 0.5/0.5 weights (ablation)SAGA · ImageNet ViT/BiT ensembleaverage robust accuracy: 28.5% -> 12.7%Improved
attack a ViT+BiT random-selection ensemble with SAGAensemble defense · CIFAR-10, CIFAR-100, ImageNet white-boxattack success rate: 74.0%, 84.4% and 91.8%Hurt
use a ViT+BiT random-selection ensemble instead of a single model against RaySensemble defense · CIFAR-10, CIFAR-100, ImageNet RayS black-boxrobust accuracy: +66.5%, +73% and +32.1%Improved
use a ViT+BiT random-selection ensemble instead of a single model against the Adaptive attackensemble defense · CIFAR-10 Adaptive black-boxrobust accuracy: 57%, +26.2%Improved
use a ViT+BiT ensemble instead of single classifiersensemble defense · clean accuracyclean accuracy: 98.2%, 92.83% and 85.37%No clear effect

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