Federated learning in edge computing: Enhancing data privacy and efficiency in resource-constrained environments

Abstract

Federated learning (FL) has become one of the most promising machine learning techniques that solve important data confidentiality and security challenges by training a model across decentralized devices without having raw data. Similarly, edge computing allows data analysis near the source, reducing time and using less bandwidth. This work examines the applications of federated learning in edge computing but focuses on scenarios with resource limitations, as seen with IoT devices and mobile networks. Reviewing the currently used approaches, issues, and trends, the features related to the future development involving opportunities and risks are considered. The emerging studies show that federated learning in the context of edge computing improves processing capacity in the federated model by preserving privacy concerns and is applicable in smart cities, healthcare smart grids, and self-driving systems. Some research directions for future works are suggested to address the scale-up and use of resources.