Towards Deep Learning Models Resistant to Adversarial Attacks

TL;DR

This paper explores the vulnerability of deep neural networks to adversarial attacks and proposes a robust optimization framework to improve their resistance, providing a foundation for more secure deep learning models.

Contribution

It introduces a robust optimization approach for training neural networks that significantly enhances their resistance to adversarial attacks and offers a formal security guarantee.

Findings

Networks trained with the proposed method show increased robustness against various adversarial attacks.

The approach provides a concrete security guarantee against first-order adversaries.

Code and models are publicly available for reproducibility.

Abstract

Recent work has demonstrated that deep neural networks are vulnerable to adversarial examples---inputs that are almost indistinguishable from natural data and yet classified incorrectly by the network. In fact, some of the latest findings suggest that the existence of adversarial attacks may be an inherent weakness of deep learning models. To address this problem, we study the adversarial robustness of neural networks through the lens of robust optimization. This approach provides us with a broad and unifying view on much of the prior work on this topic. Its principled nature also enables us to identify methods for both training and attacking neural networks that are reliable and, in a certain sense, universal. In particular, they specify a concrete security guarantee that would protect against any adversary. These methods let us train networks with significantly improved resistance to a wide range of adversarial attacks. They also suggest the notion of security against a first-order adversary as a natural and broad security guarantee. We believe that robustness against such well-defined classes of adversaries is an important stepping stone towards fully resistant deep learning models. Code and pre-trained models are available at https://github.com/MadryLab/mnist_challenge and https://github.com/MadryLab/cifar10_challenge.