HyFed: A Hybrid Federated Framework for Privacy-preserving Machine Learning

Federated learning (FL) enables multiple clients to jointly train a global\nmodel under the coordination of a central server. Although FL is a\nprivacy-aware paradigm, where raw data sharing is not required, recent studies\nhave shown that FL might leak the private data of a client through the model\nparameters shared with the server or the other clients. In this paper, we\npresent the HyFed framework, which enhances the privacy of FL while preserving\nthe utility of the global model. HyFed provides developers with a generic API\nto develop federated, privacy-preserving algorithms. HyFed supports both\nsimulation and federated operation modes and its source code is publicly\navailable at https://github.com/tum-aimed/hyfed.\n