Expanding Federated Learning using Distributed Ledger Technologies for Resilience

Abstract

Federated Learning (FL) has emerged as a distributed platform for machine learning models that ensures users’ data privacy, however trained models are vulnerable to challenges such as unreliable clients, single points of failure (server), data poisoning, and trust issues among clients. To address these issues, DLT (Distributed Ledger Technology) offers resilience by providing transparency among clients, decentralized model aggregation, and tamper-proof transaction recording. Integrating DLT with FL not only ensures secure and verifiable model updates but also enhances fault tolerance through consensus mechanisms. This research is an attempt to explore how blockchain-based DLT architecture can strengthen the resilience of trained models by providing security and reliability in heterogeneous environments. The chapter discusses the components of the DLT-based architecture and how resilience is ensured.