Mitigating Membership Inference Attacks Through Machine Unlearning

DOI

10.2991/978-94-6239-768-2_26How to use a DOI?

Keywords

Machine Unlearning; Membership Inference Attacks (MIA); Privacy Preservation; CIFAR-10; ResNet-18

Abstract

This study investigates the potential of machine unlearning as a countermeasure to Membership Inference Attacks (MIAs), which are a class of attacks that compromise data privacy by ascertaining whether certain data points were utilized in the training of a model. Machine unlearning lets models “forget” specific training data when requested, a tool that can help with privacy, legal, and bias concerns. Using a pre-trained ResNet-18 on CIFAR-10, the data was split into a Retain (retained) and a Forget (deleted) subset, with unlearning having been performed through fine-tuning on the Retain subset to simulate forgetting. This approach was evaluated in terms of accuracy, the distribution of loss, and exposure to MIA pre- and post-unlearning, using a model retrained on the Retain subset as a comparator. The results show that unlearning can largely eliminate MIA risks while suffering minor accuracy loss, and is thus a promising privacy-preserving AI technique.

Copyright

© 2026 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License (http://creativecommons.org/licenses/by-nc-nd/4.0/), which permits any noncommercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if you modified the licensed material. You do not have permission under this license to share adapted material derived from this chapter or parts of it.

Download article (PDF)

TY  - CONF
AU  - G. Uday Kiran
AU  - V. Srilakshmi
AU  - G. Padmini
AU  - G. Sreenidhi
AU  - B. Venkata Ramana
AU  - K. Delhi Babu
PY  - 2026
DA  - 2026/09/07
TI  - Mitigating Membership Inference Attacks Through Machine Unlearning
BT  - Proceedings of the Third International Conference on Recent Advances in Computing Sciences (RACS 2025)
PB  - Atlantis Press
SP  - 240
EP  - 247
SN  - 1951-6851
UR  - https://doi.org/10.2991/978-94-6239-768-2_26
DO  - 10.2991/978-94-6239-768-2_26
ID  - Kiran2026
ER  -