Investigate the Advancements in Federated Learning Techniques to Enable Privacy-Preserving Machine Learning

Abstract

A possible approach to address the increasing security and privacy concerns is federated learning (FL). Its primary benefit is dispersed client participation to learn the model without needing to submit any personal data. In this study, researchers have explored the key technologies underpinning FL from both practical and theoretical viewpoints, and they followed the creation experience of associated publications. In order to be more specific, researchers first categorize the research on FL architecture into different groups based on the network architecture of FL systems. Researchers then extend the theoretical framework of FL base designs, define the broad techniques, and reframe the application’s difficulties. They also present the suggested methods for modelling training utilizing FL.

They have developed an additional generalized algorithm-building framework after analyzing and summarizing the current FedOpt algorithms and delving completely into the creation principles for multiple first-order algorithms. Such frameworks make it simple to develop FedOpt algorithms. Considering the significance of security and confidentiality in Florida, they outline possible dangers and countermeasures. The main aim of this research is to assess different types of advancements in the techniques of federated learning for enabling privacy-preserving in machine learning.