Anjali Dave, Hardik Bhatt, Pooja Joshi, 2026. "Machine Unlearning: Taxonomy, Algorithms, And Open Research Challenges", International Journal of Review Computing and Information Technology (IJRCIT) 1(1): 88-104.

As machine learning systems are deployed in an ever growing number of different domains, minimizing data access has become a serious concern with respect to data privacy, regulatory compliance and data retention. Contemporary machine learning models are typically trained on huge volumes of user-generated data, making it challenging to erase the effect of particular entries in a read-only record once training is complete. Typically, these traditional approaches would require retraining the model from scratch whenever data needs to be deleted, incurring heavy computational costs and operational inefficiencies. More recently, Machine Unlearning has emerged as a very interesting focus area of research which focuses on enabling trained models to forget certain pieces of data while retaining the utility from the remaining learned task. This is all key as privacy regulations such as GDPR and California Consumer Privacy Act (CCPA), give them the right to ask for their data to be deleted.
Machine Unlearning includes diverse methodologies to remove the influence of trained data samples from learned models. They are exact unlearning approaches that achieve the same outcomes as complete retraining, approximate unlearning methods that offer computational gains with potential imprecise forgetting, influence-based techniques, gradient correction operations, partitioning based uncertainties in forgetting and probabilistic forgetting process. As deep learning, federated learning, large language models and generative artificial intelligence, the need for scalable and reliable unlearning solutions has further increased in recent years. The solution to this dilemma involves designing practical applications that strike a balance between computation efficiency, model quality performance, privacy bounds and verifiable forgetting in large-scale systems.
This survey paper offers a complete study of Machine Unlearning covering its theoretical foundations, classifications, methods and metrics for assessing their performance and applications in practice. Abstract: This paper reviews algorithms for exact and approximate unlearning, privacy-preserving mechanisms to facilitate differential privacy compliance, distributed frameworks to conduct unlearning over large datasets or massive model architectures, security challenges that arise in the context of malicious adversarial attacks on the associated dynamics of modern AI systems. Specific attention is dedicated to the contribution of machine unlearning in ensuring trustworthy AI and regulatory compliance. The paper also highlights important open research challenges in what it refers to as the study of foundation models, such as scalability, verification, robustness against adversarial examples, and unlearning. Machine Unlearning and its research challenges — 2023-P NET Paper Machine Unlearning and its research challenges (Paper) Publishing year: 2023 Published in a PDF Abstract The significance of machine unlearning as responsible AI practice has been on the rise.

1. 1. Cao Yang, and Justin Zico Kolter, “Towards Making Systems Forget with Machine Unlearning,” IEEE Symposium on Security and Privacy Workshops, 2015
2. 2. Brandon Bourtoule et al., “Machine Unlearning,” IEEE Symposium on Security and Privacy, 2021.
3. 3. Ronald L. Rivest, “On Data Banks and Privacy Homomorphisms,” Foundations of Secure Computation, 1978.
4. 4. Nicolas Papernot et al., “Scalable Private Learning with PATE,” International Conference on Learning Representations (ICLR), 2018.
5. 5. Martin Abadi et al., “Deep Learning with Differential Privacy,” ACM CCS, 2016.
6. 6. Reza Shokri et al., “Membership Inference Attacks Against Machine Learning Models,” IEEE Symposium on Security and Privacy, 2017.
7. 7. Khaled El Emam, Guide to the De-Identification of Personal Health Information, CRC Press, 2013.
8. 8. Virginia Smith et al., “Federated Multi-Task Learning,” NeurIPS, 2017.
9. 9. Brendan McMahan et al., “Communication-Efficient Learning of Deep Networks from Decentralized Data,” AISTATS, 2017.
10. 10. Jakub Konečný et al., “Federated Learning: Strategies for Improving Communication Efficiency,” NeurIPS Workshop, 2016.
11. 11. Kalyan Veeramachaneni et al., “Challenges and Opportunities in Machine Unlearning,” AI Magazine, 2022.
12. 12. Song Han et al., “Learning Both Weights and Connections for Efficient Neural Networks,” NeurIPS, 2015.
13. 13. Ian Goodfellow, Yoshua Bengio, and Aaron Courville, Deep Learning, MIT Press, 2016.
14. 14. Alex Krizhevsky et al., “ImageNet Classification with Deep Convolutional Neural Networks,” NeurIPS, 2012.
15. 15. Ashish Vaswani et al., “Attention Is All You Need,” NeurIPS, 2017.
16. 16. Tom Brown et al., “Language Models are Few-Shot Learners,” NeurIPS, 2020.
17. 17. Jacob Devlin et al., “BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding,” NAACL, 2019.
18. 18. Alec Radford et al., “Improving Language Understanding by Generative Pre-Training,” OpenAI, 2018.
19. 19. Jiawei Han, Micheline Kamber, and Jian Pei, Data Mining: Concepts and Techniques, Morgan Kaufmann, 2011.
20. 20. Pedro Domingos, The Master Algorithm, Basic Books, 2015.
21. 21. Sanjay Shakkottai et al., “Certified Data Removal from Machine Learning Models,” USENIX Security Symposium, 2020.
22. 22. Amirata Ghorbani and James Zou, “Data Shapley: Equitable Valuation of Data for Machine Learning,” ICML, 2019.
23. 23. Pang Wei Koh and Percy Liang, “Understanding Black-box Predictions via Influence Functions,” ICML, 2017.
24. 24. Kaiming He et al., “Deep Residual Learning for Image Recognition,” CVPR, 2016.
25. 25. Geoffrey Hinton et al., “Reducing the Dimensionality of Data with Neural Networks,” Science, 2006.
26. 26. Diederik P. Kingma and Max Welling, “Auto-Encoding Variational Bayes,” ICLR, 2014.
27. 27. Judea Pearl, Causality: Models, Reasoning and Inference, Cambridge University Press, 2009.
28. 28. Michael I. Jordan and Tom M. Mitchell, “Machine Learning: Trends, Perspectives, and Prospects,” Science, 2015.
29. 29. Cynthia Dwork and Aaron Roth, The Algorithmic Foundations of Differential Privacy, 2014.
30. 30. Christopher Bishop, Pattern Recognition and Machine Learning, Springer, 2006.
31. 31. Trevor Hastie, Robert Tibshirani, and Jerome Friedman, The Elements of Statistical Learning, Springer, 2009.
32. 32. Kevin Murphy, Machine Learning: A Probabilistic Perspective, MIT Press, 2012.
33. 33. Andrew Ng, “Machine Learning Yearning,” DeepLearning.AI, 2018.
34. 34. Yann LeCun, Yoshua Bengio, and Geoffrey Hinton, “Deep Learning,” Nature, 2015.
35. 35. Zhiqi Bu et al., “Towards Practical and Scalable Machine Unlearning for Deep Neural Networks,” AAAI Conference on Artificial Intelligence, 2024.