MD.20.006 – Responsible Data Sharing

Route: Measuring and detecting: anything, anytime, anywhere

Cluster question: 057 Can we find the right balance between freedom of information and privacy?

Sharing data confidentially and securely is essential when safeguarding the public values of security and trust. Sharing data is important to many areas of society, in particular the medical domain, where combining patient data stored at different health care providers can tremendously improve diagnoses and treatments. However, sharing data is generally problematic when it concerns personal and sensitive data. Technological approaches such as Multi-Party Computation (MPC) and Federated Learning (FL) are very promising in this domain. Although they allow learning from data without actually sharing the data, MPC/FL also faces technical, legal, ethical challenges, which need to be addressed if MPC/FL is to reach its potential and to ensure or increase trust. The ReDaSh project does so by bringing together an interdisciplinary group of experts from law, ethics, AI, and cryptography, to address legal, ethical, governance, and technical challenges related to MPC/FL. The ReDaSh project further develops a governance framework that includes legal, social, and ethical elements (Challenge 1). Furthermore, this project tests and quantifies risk of (re)identifiability for different components of (re)identifiability (Challenge 2). Finally, it is researched whether ethical considerations can be automated (partially). Developing a governance framework, testing (re)identification, and improving the compliance with ethical standards contributes to data safety and enhances the trust in MPC/FL, consequently addressing both technical and societal aspects of cybersecurity. The ReDaSh project will primarily focus on medical data, as their confidentiality, security, but also sharing data to make accurate predictions is crucial and impactful. Consequently, this project has the potential to generate necessary insights in how to learn from distributed data in a privacy-preserving way and to contribute to saving lives, combating diseases, supporting health care, and lowering medical costs.

Keywords

data access, federated learning, multi-party computation, privacy

Other organisations

CBS, Tilburg University, TNO, University of Amsterdam

Submitter

Organisation Maastricht University (UM)
Name Prof.dr. G. (Gijs) van Dijck
E-mail gijs.vandijck@maastrichtuniversity.nl
Website https://www.maastrichtuniversity.nl/gijs.vandijck