Overview of Research Progress and Challenges in Federated Learning

Abstract

In recent years, federated learning has become a hot research topic in the machine learning community. It aims to reduce the potential data security and privacy risks caused by the centralized training paradigm of traditional machine learning through local training and global aggregation. Although federated learning methods have been widely applied in numerous fields such as finance, healthcare, autonomous driving, and smart retail, there are still urgent issues to be addressed in the field of federated learning, including data privacy leakage, malicious node attacks, model security, and the trustworthiness of participants. By delving into and discussing federated learning, this paper aims to provide researchers and practitioners in related fields with a comprehensive understanding and the latest progress of this technology. Based on the characteristics of the data distribution of the parties involved in federated learning training, this paper categorizes existing federated learning methods into horizontal federated learning, vertical federated learning, and federated transfer learning. It also introduces representative federated learning algorithms under different types, including their design concepts, basic processes, and advantages and disadvantages. Combining different application scenarios, this paper further discusses the challenges of federated learning and looks forward to the future development direction of this topic.