REDRAW: Privacy-Preserving Federated Learning in the Cloud–Edge Continuum for Human Wellbeing Monitoring

Digital health systems increasingly rely on pervasive sensing, mobile devices, and cloud services, raising challenges related to privacy, data heterogeneity, and system deployment. The REDRAW project investigated how Federated learning (FL) mitigates some of these issues by enabling distributed training without sharing raw data, addressing the software and system-level complexities introduced by this computing paradigm. This describes a methodological framework for FL experimentation, a machine-readable representation of experiments, and a cloud–edge architecture supporting heterogeneous devices and hybrid deployment. Clinical use cases, including sleep monitoring, guide the design and evaluation. Overall, REDRAW highlights the importance of integrating software architecture, deployment engineering, and federated analytics. We discuss current limitations, including trade-offs among privacy, communication costs, edge computation, and model performance in real-world healthcare scenarios.