Tag: Federated Learning (FL)
New EOSC-SIESTA publication: privacy-preserving Federated Learning in medical imaging
A new EOSC-SIESTA publication by IFCA-CSIC, INRIA, and TU Delft explores privacy protection in Federated Learning using sensitive medical imaging data. The study proposes a metric-privacy-inspired noise calibration strategy that improves global model accuracy while defending against Client Inference Attacks.
SIESTA CONCEPTS#1: Federated Learning
Federated Learning (FL) allows multiple participants to collaboratively train AI models by keeping data local and sharing only encrypted model updates. Through the SIESTA dashboard, users can easily deploy secure FL clients, which connect to a central coordinating server hosted on AI4EOSC via the Flower framework and secure gRPC protocols, enabling privacy-preserving collaboration.