BFLEdge: Blockchain based federated edge learning scheme in V2X underlying 6G communications

Abstract

Sixth generation (6G) vehicle-to-anything (V2X) networks support intelligent edge computing that leverages data sensing, computation, and offloading among vehicular nodes (VN) with ultra-low latency. Data is heterogeneous with high complex interactions among V2X users and pass via open channels that induce privacy and security concerns. Thus, federated learning (FL) protects user privacy and fine-tunes the learning models at resource-constrained edge nodes to address security and computational concerns at the edge. However, to ensure reliability and trust, we propose a block-chain (BC) and FL-based edge scheme, BFLEdge. It also improves the overall learning rate of the FL model. The proposed scheme consists of three phases, where the first phase uses local machine learning (LML) to model the VN data and store it into the local BC network. The LML block updates are verified in the second phase through a proposed distributed consensus mechanism. Lastly, through 6G communication services, the channel dynamics are modelled as a Markov chain process to reduce end-to-end delay of local BC propagation updates at the edge that improves the V2X system throughput. Simulation and analytical results are proposed based on channel loss, block mining rate, edge latency, and FL-learning rate. The obtained results indicate the viability of the proposed framework against conventional state-of-the-art approaches.