Federated Learning for Data Spaces: a Privacy-Enhancing Strategy Based on Data Visiting
Manlio Bacco, Margherita Di Leo, Albana Kona, Mattia Santoro, P. Mazzetti · 2024
This work explores the paradigm of data visiting that, through privacy-enhancing technologies, shows the potential to access and use data otherwise inaccessible. Building on the ongoing EU initiative to design, implement, and run sectorial data spaces, we consider federated learning as one the most promising approaches for the objective above. We propose a domain-agnostic strategy that can be extended and adapted to different needs. We conclude by analysing the limitations and challenges of the approach we propose.