A Systematic Mapping Study on Machine Unlearning in Federated Learning

This study systematically analyzes research on machine unlearning in federated learning architectures and examines its distribution in the literature. As part of the study, a mapping analysis was conducted by categorizing articles based on year, electronic databases, publication channels, and journals. Studies published since 2015 were analyzed, and data collection was performed using the IEEE, ACM Digital Library, ScienceDirect, Wiley, and Scopus databases. The search yielded a total of 564 articles; however, after a detailed analysis, only 113 were found to be directly related to machine unlearning in federated learning architectures.The findings indicate that various approaches regarding machine unlearning have been introduced in the literature. However, further investigation is needed into the benefits, technical performance, and security implications of machine unlearning techniques in federated learning architectures designed for distributed systems. In particular, integrating different machine unlearning techniques into federated learning systems and comparing these techniques in terms of efficiency and computational cost are important areas for future research.