Adversarially Robust and Privacy-Preserving Representation Learning via Information Theory

Binghui Zhang, Sayedeh Leila Noorbakhsh, Yun Dong, Yuan Hong, Binghui Wang

Keywords: Representation learning, adversarially robust, privacy-preserving, information theory

Abstract:

Machine learning models are vulnerable to both security (e.g., adversarial examples) attacks and privacy (e.g., private attribute inference) attacks. In this paper, we aim to mitigate both the security and privacy attacks, and maintain utility of the primary task simultaneously.
Particularly, we propose an information-theoretical framework to achieve the goals through the lens of representation learning, i.e., learning representations that are robust to both adversarial examples and attribute inference adversaries. We also derive novel theoretical results, i.e., the inherent tradeoff between adversarial robustness/utility and attribute privacy, as well as guaranteed attribute privacy leakage against attribute inference adversaries.

Primary Area: societal considerations including fairness, safety, privacy

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: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2024/AuthorGuide.

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.

Submission Number: 4539