A Threat-Driven Design of a Data-Trustee Infrastructure for Medical Data

Medical research using Artificial Intelligence (AI) is a growing field and, as such, requires large amounts of data to train AI models. A data-trustee infrastructure (DTI) can enable secure data sharing between data providers and users. However, gaining public acceptance requires addressing concerns about privacy, security, and transparency. We identify threats based on patients’ concerns, define security-related requirements, and implement measures. Organizationally, contracts must be signed between the DTI and data users and data providers, respectively. Technically, a security-by-design architecture of a DTI that focuses on the separation of data and modules is provided, minimizing the possibility of data breaches, re-identification and unauthorized access. Access control based on patients’ consent is enforced. Only data relevant to a research hypothesis is provided through a privacy-preserving search algorithm on protected data. Overall, we believe this proposal can serve as a blueprint for the design of a data-trustee infrastructure for medical data with a focus on patients’ concerns about data sharing.