Home

Awesome

Patch-wise Adversarial Removal

Implementation of paper "Decision-based Black-box Attack Against Vision Transformers via Patch-wise Adversarial Removal".

<img src='/framework.png' width=900/>

Introduction

Patch-wise Adversarial Removal (PAR) is a new decision-based black-box attack against vision transformers, which divides images into patches through a coarse-to-fine search process and compresses the noise on each patch separately.