Vulnerability-Aware Instance Reweighting For Adversarial Training

TL;DR

This paper introduces a vulnerability-aware instance reweighting method for adversarial training that enhances robustness by considering individual example vulnerabilities, outperforming existing schemes against strong attacks.

Contribution

It proposes a novel reweighting scheme that accounts for example vulnerability and information loss, improving adversarial training effectiveness.

Findings

Significant robustness improvements over existing reweighting methods.

Enhanced defense against strong white-box and black-box attacks.

Effective in reducing vulnerability disparities among training examples.

Abstract

Adversarial Training (AT) has been found to substantially improve the robustness of deep learning classifiers against adversarial attacks. AT involves obtaining robustness by including adversarial examples in training a classifier. Most variants of AT algorithms treat every training example equally. However, recent works have shown that better performance is achievable by treating them unequally. In addition, it has been observed that AT exerts an uneven influence on different classes in a training set and unfairly hurts examples corresponding to classes that are inherently harder to classify. Consequently, various reweighting schemes have been proposed that assign unequal weights to robust losses of individual examples in a training set. In this work, we propose a novel instance-wise reweighting scheme. It considers the vulnerability of each natural example and the resulting information loss on its adversarial counterpart occasioned by adversarial attacks. Through extensive experiments, we show that our proposed method significantly improves over existing reweighting schemes, especially against strong white and black-box attacks.