Enhancing Security and Privacy in Federated Learning

Abstract

The PAROMA-MED project advances privacy-preserving, federated machine learning for medical applications, ensuring sensitive health data is protected during model development and clinical research. Core to its strategy is the “code-to-data” approach, which ensures data never leaves its secure source. Analysis, AI training, and access are brought to the data via consent-driven, federated, and privacy-enhancing technologies. Among its innovative mechanisms, granular consent management stands out as a cornerstone for upholding patient autonomy and regulatory standards, such as GDPR and the European Health Data Space (EHDS).