On the Necessity of Auditable Algorithmic Definitions for Machine Unlearning

Authors: 

Anvith Thudi, Hengrui Jia, Ilia Shumailov, and Nicolas Papernot, University of Toronto and Vector Institute

Abstract: 

Machine unlearning, i.e. having a model forget about some of its training data, has become increasingly more important as privacy legislation promotes variants of the right-to-be-forgotten. In the context of deep learning, approaches for machine unlearning are broadly categorized into two classes: exact unlearning methods, where an entity has formally removed the data point's impact on the model by retraining the model from scratch, and approximate unlearning, where an entity approximates the model parameters one would obtain by exact unlearning to save on compute costs. In this paper, we first show that the definition that underlies approximate unlearning, which seeks to prove the approximately unlearned model is close to an exactly retrained model, is incorrect because one can obtain the same model using different datasets. Thus one could unlearn without modifying the model at all. We then turn to exact unlearning approaches and ask how to verify their claims of unlearning. Our results show that even for a given training trajectory one cannot formally prove the absence of certain data points used during training. We thus conclude that unlearning is only well-defined at the algorithmic level, where an entity's only possible auditable claim to unlearning is that they used a particular algorithm designed to allow for external scrutiny during an audit.

Open Access Media

USENIX is committed to Open Access to the research presented at our events. Papers and proceedings are freely available to everyone once the event begins. Any video, audio, and/or slides that are posted after the event are also free and open to everyone. Support USENIX and our commitment to Open Access.

BibTeX
@inproceedings {279996,
author = {Anvith Thudi and Hengrui Jia and Ilia Shumailov and Nicolas Papernot},
title = {On the Necessity of Auditable Algorithmic Definitions for Machine Unlearning},
booktitle = {31st USENIX Security Symposium (USENIX Security 22)},
year = {2022},
isbn = {978-1-939133-31-1},
address = {Boston, MA},
pages = {4007--4022},
url = {https://www.usenix.org/conference/usenixsecurity22/presentation/thudi},
publisher = {USENIX Association},
month = aug,
}

Presentation Video