Exploring Robust Features for Improving Adversarial Robustness

TL;DR

This paper introduces a feature disentanglement approach to isolate robust features unaffected by adversarial attacks, enhancing neural network robustness and enabling effective adversarial example detection without extra computational costs.

Contribution

The paper proposes a novel feature disentanglement model that separates robust, non-robust, and domain-specific features, improving adversarial robustness and detection capabilities.

Findings

Robust features improve adversarial robustness over state-of-the-art methods.

Domain discriminator nearly perfectly identifies domain-specific features.

Enables adversarial example detection without additional computational costs.

Abstract

While deep neural networks (DNNs) have revolutionized many fields, their fragility to carefully designed adversarial attacks impedes the usage of DNNs in safety-critical applications. In this paper, we strive to explore the robust features which are not affected by the adversarial perturbations, i.e., invariant to the clean image and its adversarial examples, to improve the model's adversarial robustness. Specifically, we propose a feature disentanglement model to segregate the robust features from non-robust features and domain specific features. The extensive experiments on four widely used datasets with different attacks demonstrate that robust features obtained from our model improve the model's adversarial robustness compared to the state-of-the-art approaches. Moreover, the trained domain discriminator is able to identify the domain specific features from the clean images and adversarial examples almost perfectly. This enables adversarial example detection without incurring additional computational costs. With that, we can also specify different classifiers for clean images and adversarial examples, thereby avoiding any drop in clean image accuracy.