Federated Learning-Assisted Privacy-Preserving Service Placement in Software-Defined Vehicles

The dynamic nature of Software-Defined Vehicles (SDVs) poses significant challenges in making timely and accurate decisions for service placements, especially because of the presence of privacy and security risks. These challenges underscore the urgent need for innovative solutions tailored to dynamic vehicular environments. To this, end, Federated learning (FL) has emerged as a promising paradigm, allowing to distribute Machine Learning (ML)-based analysis and thanks to its robust privacy-preserving capabilities and inherent scalability, in the end offering a viable approach to address evolving demands while safeguarding sensitive data. In this paper, we propose an FLassisted privacy-preserved hybrid model for service placements in SDVs, denoted as FL-PPSP. Our approach ensures the priority of critical tasks over regular ones while preserving data privacy: therefore, FL mitigates the risks associated with centralized data storage while enhancing efficiency in heterogeneous vehicular environments. From an operational point of view, our proposed approach leverages the FedProx framework to ease efficient federated training within SDVs. Additionally, Strength Pareto Evolutionary Algorithm 2 (SPEA2) is employed to determine optimal trade-offs among performance metrics, while VIKOR is utilized to rank solutions, thus identifying the most effective service placement strategy.