Security and Privacy for Distributed Learning Techniques in IoT-Driven Health Services (SePRIO)

The surge in digital health data generated, stored, and processed is a significant phenomenon leading to numerous technical and practical challenges regarding how these data will be shared, processed, and stored in efficient, secure, privacy-preserving, and scalable solutions for various use cases in the health domain. Therefore, there is a strong need for innovative, robust, and privacy-preserving distributed learning techniques that effectively implement the underlying data processing/analytics principles without compromising efficiency and utility in health applications.

This is also necessary for digital sovereignty and the competitiveness of Swiss industries, which are highly successful in the health technology and pharmaceutical sectors. Specifically, secure aggregation and privacy-preserving techniques for distributed learning, such as Federated Learning (FL), should be implemented, tested, and integrated in health use cases with IoT devices. These efforts will lead to new solutions and an analysis of the limitations and trade-offs of those privacy-preserving algorithms that operate in distributed settings.

In that regard, the primary objective of the SePRIO project is to investigate and integrate security and privacy into distributed learning techniques, with a focus on digital health use cases. To this end, first, we will explore a wide range of secure aggregation and privacy-preserving techniques.

Then, we will integrate these techniques into our AI infrastructure and use our augmented distributed learning models (e.g., FL) to analyze open health dataset(s) for a specific use case involving IoT.

Finally, we will analyze the limitations and trade-offs. The outcomes of this project will lead to an open-source software project, related publications, talks and presentations, and concrete collaboration between two EELISA universities, ZHAW and ITU.