Certified Defenses for Adversarial Patches
Ping-yeh Chiang*, Renkun Ni*, Ahmed Abdelkader, Chen Zhu, Christoph Studor, Tom Goldstein
Abstract
Adversarial patch attacks are among one of the most practical threat models against real-world computer vision systems. This paper studies certified and empirical defenses against patch attacks. We begin with a set of experiments showing that most existing defenses, which work by pre-processing input images to mitigate adversarial patches, are easily broken by simple white-box adversaries. Motivated by this finding, we propose the first certified defense against patch attacks, and propose faster methods for its training. Furthermore, we experiment with different patch shapes for testing, obtaining surprisingly good robustness transfer across shapes, and present preliminary results on certified defense against sparse attacks. Our complete implementation can be found on: https://github.com/Ping-C/certifiedpatchdefense.
BibTeX
@inproceedings{
Chiang*2020Certified,
title={Certified Defenses for Adversarial Patches},
author={Ping-yeh Chiang* and Renkun Ni* and Ahmed Abdelkader and Chen Zhu and Christoph Studor and Tom Goldstein},
booktitle={International Conference on Learning Representations},
year={2020},
url={https://openreview.net/forum?id=HyeaSkrYPH}
}