The Paper

The paper called “The FeatureCloud Platform for Federated Learning in Biomedicine: Unified Approach” introduces FeatureCloud, an innovative platform designed to simplify federated learning (FL) in various fields, with a focus on biomedicine. FL is a technique that enables the training of machine learning models using decentralized data sources without sharing sensitive information.

Existing FL frameworks are often specialized, lack a generic approach, and require programming skills. FeatureCloud aims to bridge these gaps by providing an all-in-one platform with three main components: a global frontend, a global backend, and a local controller. It leverages Docker to isolate local components from sensitive data systems, ensuring data privacy compliance. The platform supports privacy-enhancing technologies and facilitates the sharing and reuse of federated algorithms through an integrated AI store.

Evaluation results indicate that FeatureCloud can produce results similar to centralized approaches and scale effectively across multiple participating sites. Ultimately, FeatureCloud is expected to enhance the accessibility of privacy-preserving and distributed data analysis in biomedicine and other domains, making FL more accessible to both developers and non-programming users.

The Project

Featurecloud is a research project started in 2019 and is funded by the Funding framework Horizon 2020 of the European Commission. You can find more details about it on our project page and on


The paper has 38 authors; Markus Kastelitz, Walter Hötzendorfer, Jan Hospes and Christof Tschohl worked on the paper for Research Institute – Digital Human Rights Center.


Journal of Medical Internet Research ISSN 1438-8871The FeatureCloud Platform for Federated Learning in Biomedicine: Unified Approach