Forgetting Is Necessary: A Study on the Impact of Data Removal on Machine Unlearning

• Milena Curtinhas Santos UFES
• João Paulo de Brito Gonçalves IFES
• Antonio A. de A. Rocha UFF
• Rodolfo da Silva Villaça UFES

DOI: https://doi.org/10.5753/eries.2025.15735

Abstract

The increasing stringency of data protection regulations, such as the LGPD and GDPR, has driven the development of machine unlearning techniques to ensure the right to be forgotten in artificial intelligence models. This article reviews key concepts, challenges, and recent advances in the field, experimentally evaluating different unlearning algorithms, including DaRE and DynFrs, across multiple datasets. Results indicate that small-scale data removals generally have a limited impact on model accuracy, highlighting the need for efficient and robust approaches. Finally, future perspectives are discussed, such as validating unlearning through blockchain and integrating explainable AI (XAI) techniques, aiming for more transparent and trustworthy systems.

References

Alpaydin, E. and Kaynak, C. (1998). Optical Recognition of Handwritten Digits. UCI Machine Learning Repository. DOI: 10.24432/C50P49.

Bourtoule, L., Chandrasekaran, V., Choquette-Choo, C. A., Jia, H., Travers, A., Zhang, B., Lie, D., and Papernot, N. (2021). Machine unlearning. In 2021 IEEE Symposium on Security and Privacy (SP), pages 141–159.

Brophy, J. and Lowd, D. (2021). Machine unlearning for random forests. In Meila, M. and Zhang, T., editors, Proceedings of the 38th International Conference on Machine Learning, volume 139 of Proceedings of Machine Learning Research, pages 1092–1104. PMLR.

Dang, Q.-V. (2021). Right to be forgotten in the age of machine learning. In Antipova, T., editor, Advances in Digital Science, pages 403–411, Cham. Springer International Publishing.

Jin, R., Chen, M., Zhang, Q., and Li, X. (2024). Forgettable federated linear learning with certified data unlearning.

Li, Y., Chen, C., Zhang, Y., Liu, W., Lyu, L., Zheng, X., Meng, D., and Wang, J. (2023). UltraRE: Enhancing receraser for recommendation unlearning via error decomposition. In Thirty-seventh Conference on Neural Information Processing Systems.

Nguyen, T. T., Huynh, T. T., Ren, Z., Nguyen, P. L., Liew, A. W.-C., Yin, H., and Nguyen, Q. V. H. (2024). A survey of machine unlearning.

Presidência da República (2018). Lei geral de proteção de dados pessoais (lgpd). [link]. Acesso em: 24 jun. 2025.

Schelter, S., Grafberger, S., and Dunning, T. (2021). Hedgecut: Maintaining randomised trees for low-latency machine unlearning. In Proceedings of the 2021 International Conference on Management of Data (SIGMOD ’21), page –, Virtual Event, China.

Thudi, A., Deza, G., Chandrasekaran, V., and Papernot, N. (2022). Unrolling sgd: Understanding factors influencing machine unlearning. In 2022 IEEE 7th European Symposium on Security and Privacy (EuroS&P), pages 303–319.

União Europeia (2016). Regulamento (ue) 2016/679 do parlamento europeu e do conselho, de 27 de abril de 2016, relativo à proteção das pessoas singulares no que diz respeito ao tratamento de dados pessoais e à livre circulação desses dados (regulamento geral sobre a proteção de dados - rgpd). [link]. Acesso em: 21 jun. 2024.

Wang, S., Shen, Z., Qiao, X., Zhang, T., and Zhang, M. (2025). Dynfrs: An efficient framework for machine unlearning in random forest.