Security and Privacy-Preserving Techniques of Federated Learning in Edge Computing: A Comparative Study

Abstract

Edge Computing (EC) is an innovative architecture that aims to provide cloud computing services near data sources. In conventional deep learning systems that use edge computing, data producers frequently need to transmit and exchange data with other entities, such as edge or cloud servers, to train their models. Federated Learning (FL) has recently emerged as a potential remedy for the problems of undesired bandwidth use, data confidentiality, and legal compliance. To enhance the status of research in this field and effectively implement the federated learning strategy, it is essential to identify and analyze its security and privacy issues first. Federated learning (FL) is the optimal method for addressing security and privacy concerns in edge computing. A profound grasp of risk factors allows FL implementers to create a secure environment while researchers benefit from a clear understanding of prospective research areas. The primary objective of this study is to evaluate previous contemporary studies in relevant disciplines to understand the issues and subjects covered in recent surveys. Furthermore, FL research is categorized into architectures, applications, and issues. Moreover, we analyze substantial security and privacy risks and propose new mitigation strategies.