To be robust and to be fair: aligning fairness with robustness

Junyi Chai, Xiaoqian Wang

Keywords: fairness, adversarial robustness

TL;DR: bridging adversarial robustness of fairness and accuracy in a unified framework

Abstract: Adversarial training has been shown to be reliable in improving robustness against adversarial samples. However, the problem of adversarial training in terms of fairness has not yet been properly studied, and the relationship between fairness and accuracy attack still remains unclear. Can we simultaneously improve robustness w.r.t. both fairness and accuracy? To tackle this topic, in this paper, we study the problem of adversarial training and adversarial attacks w.r.t. both metrics. We propose a unified structure for fairness attack which bring together common notions in group fairness, and we theoretically prove the equivalence of fairness attacks against different notions. We show the alignment of fairness and accuracy attack in disadvantaged subgroups, and we theoretically demonstrate that robustness of samples w.r.t. adversarial attack against one metric also benefit from robustness of samples w.r.t. adversarial attack against the other metric. Our work unifies adversarial training and attack w.r.t. fairness and accuracy, where both metrics benefit from robustness of the other metric under adversarial attack. Our study suggests a novel way to incorporate adversarial training with fairness, and experimental results show that our proposed method achieves better performance in terms of robustness w.r.t. both fairness and accuracy.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics

Submission Guidelines: Yes

Please Choose The Closest Area That Your Submission Falls Into: Social Aspects of Machine Learning (eg, AI safety, fairness, privacy, interpretability, human-AI interaction, ethics)