FLBCPS: Federated Learning based Secured Computation Offloading in Blockchain-Assisted Cyber-Physical Systems

Abstract

Mobile-edge computing (MEC) is a in demand method for improving the quality of computation experience on mobile devices (MD) since it helps MD's to offload computing activities to MEC servers, which provide strong computing capabilities. However, there are certain unresolved concerns in present computation-offloading works: 1) safety issue; 2) joint computation offloading; and 3) flexible optimization. To solve safety and privacy concerns, we use Federated Learning-based blockchain technology, which provides data accuracy and irreversibility in MEC systems. Federated Learning (FL) is a promising technique towards effective machine learning while protecting privacy in dispersed situations such as the Internet of Things (IoT) and MEC. FL's effectiveness is dependent on a network of participant nodes contributing their data and computational resources to the collective training of a globally model. As a result, preventing malicious nodes from interfering with model training while rewarding trustworthy nodes to assist to the learning process is critical for improved FL security and performance. We created an efficient resource allocation technique that optimizes computational offloading using a Blockchain-based Federated Learning (FL) method in order to add to the literature. The experimental findings show's that the recommended FLBCPS technique improve system latency while maintaining consensus security. © 2022 IEEE.