Defending Against Physically Realizable Attacks on Image Classification

Tong Wu, Liang Tong, Yevgeniy Vorobeychik
9/20/2019
cs.LGcs.AIcs.CV

Official code: tongwu2020/phattacks (35 stars, last commit 2022-06-17).

Abstract

We study the problem of defending deep neural network approaches for image classification from physically realizable attacks. First, we demonstrate that the two most scalable and effective methods for learning robust models, adversarial training with PGD attacks and randomized smoothing, exhibit very limited effectiveness against three of the highest profile physical attacks. Next, we propose a new abstract adversarial model, rectangular occlusion attacks, in which an adversary places a small adversarially crafted rectangle in an image, and develop two approaches for efficiently computing the resulting adversarial examples. Finally, we demonstrate that adversarial training using our new attack yields image classification models that exhibit high robustness against the physically realizable attacks we study, offering the first effective generic defense against such attacks.

View on arXivView PDF

Code Implementations(1)

tongwu2020/phattacksOfficial100%

Defending Against Physically Realizable Attacks on Image Classification

3510PythonDec 28, 20194 years agoNOASSERTION

Similar Papers

Neural Fingerprints for Adversarial Attack Detection

90% match1 repo22024

SegPGD: An Effective and Efficient Adversarial Attack for Evaluating and Boosting Segmentation Robustness

88% match1 repo52022

Defense-GAN: Protecting Classifiers Against Adversarial Attacks Using Generative Models

88% match1 repo2482018

Not So Robust After All: Evaluating the Robustness of Deep Neural Networks to Unseen Adversarial Attacks

88% match2023

Algorithms for Adversarially Robust Deep Learning

88% match2025

C-LEAD: Contrastive Learning for Enhanced Adversarial Defense

88% match2025

Cite this paper

@article{wu2019defending,
  title  = {Defending Against Physically Realizable Attacks on Image Classification},
  author = {Tong Wu and Liang Tong and Yevgeniy Vorobeychik},
  year   = {2019},
  eprint = {1909.09552},
  archivePrefix = {arXiv},
  url    = {https://arxiv.org/abs/1909.09552}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/1909.09552.svg)](https://codeofpaper.com/paper/1909.09552)

Discussion