AI in Health and Social Care: A Methodology for Privacy Risk Modeling and Simulation

As health and social care data networks evolve and adapt to greater digitalization and datafication of health, data and analytics systems are developing and bringing forward new ways to share, access and analyze data. Organizations and individuals making data sharing decisions for AI-enabled health and social care services need to be able to balance the benefits of such uses with the possible risks that may ensue - including those related to issues of privacy and security. In this paper, we provide an overview of our approach to privacy risk assessment for cross-domain access and re-use of sensitive data for research purposes using Spyderisk - an automated risk assessment tool. We apply Spyderisk to a real AI research scenario and consider the ways in which such techniques could support multiple stakeholders to assess privacy and security risks.