Incentivized Federated Learning in Data Spaces

This paper presents Incentivized Federated Learning (IFL), a framework integrating federated learning and Data Spaces to tackle privacy and data sovereignty challenges in Intelligent Transportation Systems (ITS). Using the Vehicle ReID dataset as a case study, IFL rewards participants based on data quality and privacy leakage, mitigating poisoning, inference, and GAN-based attacks. By enabling collaborative model training without sharing raw images, IFL safeguards privacy but reduces data isolation in ITS. Data Space technologies further enhance interoperability and align with global data-sharing standards. Experimental evaluations on Vehicle ReID and CIFAR-10 confirm IFL’s effectiveness in protecting privacy, boosting data utility, and ensuring secure, scalable ITS collaboration.