An Innovative Hashgraph-based Federated Learning Approach for Multi Domain 5G Network Protection

Abstract

Federated Learning (FL) is a decentralized learning approach, meaning it learns from data housed locally on devices such as tablets, cellular phones, and more, and does not collect nor transfer user-sensitive data but merely learns from the data utilizing a shared model and sending periodical updates. Using federated learning throws out the problems associated with user privacy and the high bandwidth needed to transmit resource-intensive files to a central server for training. However, FL systems may be compromised to make a wrong decision or disclose private data once the attacker modifies the FL model and/or its paraments. The main contribution of this paper includes (1) introducing a comprehensive study that explores the FL and how it applies to different domains like healthcare and medicine, Insurance and Finance, Robotics and Autonomous Systems, Virtual Reality, and 5G. (2) Develop a Hashgraph-based federated learning Approach (HFLA) to protect the 5G network against poisoning and membership inherence attacks. The HFLA was evaluated using our Federated 5G testbed and proved its superiority compared to other existing FL approaches.