Hard Adversarial Example Mining for Improving Robust Fairness

Chenhao Lin

doi:10.1109/TIFS.2024.3516554

Hard Adversarial Example Mining for Improving Robust Fairness

2025-01-01

0

OA

AI

DOI：10.1109/TIFS.2024.3516554

Original

Original request for help

Save

Abstract

En 中文

Adversarial training (AT) is widely considered the state-of-the-art technique for improving the robustness of deep neural networks (DNNs) against adversarial examples (AEs). Nevertheless, recent studies have revealed that adversarially trained models are prone to unfairness problems. Recent works in this field usually apply class-wise regularization methods to enhance the fairness of AT. However, this paper discovers that these paradigms can be sub-optimal in improving robust fairness. Specifically, we empirically observe that the AEs that are already robust (referred to as easy AEs in this paper) are useless and even harmful in improving robust fairness. To this end, we propose the hard adversarial example mining (HAM) technique which concentrates on mining hard AEs while discarding the easy AEs in AT. Specifically, HAM identifies the easy AEs and hard AEs with a fast adversarial attack method. By discarding the easy AEs and reweighting the hard AEs, the robust fairness of the model can be efficiently and effectively improved. Extensive experimental results on four image classification datasets demonstrate the improvement of HAM in robust fairness and training efficiency compared to several state-of-the-art fair adversarial training methods. Our code is available at https://github.com/yyl-github-1896/HAM.

Keywords:

Training

Robustness

Accuracy

Modeling

Optimization

Computational efficiency

Computational modeling

Sun

Perturbation methods

Entropy

Adversarial training

robust fairness

hard adversarial example mining

convolutional neural network

Hard Adversarial Example Mining for Improving Robust Fairness

Abstract

Journal

Organization