Enabling decision support over confidential data

Abstract: Enabling automated decision-making processes by leveraging data-driven analysis is a core goal of Decision Support Systems (DSSs). In multi-party scenarios where decisions rely on distributed and sensitive data, though, ensuring confidentiality, verifiability, transparency, integrity, and consistency at once remains an open challenge for DSSs. To tackle this multi-faceted problem, we propose the Secure Platform for Automated decision Rules via Trusted Applications (SPARTA) approach. By leveraging Trusted Execution Environments (TEEs) at its core, SPARTA ensures that the decision logic and the data remain protected. To guarantee transparency and consistency of the decision process, SPARTA encodes decision rules into verifiable software objects deployed within TEEs. To maintain the confidentiality of the outcomes while keeping the information integrity, SPARTA employs cryptography techniques on notarized data based on user-definable access policies. Based on experiments conducted on public benchmarks and synthetic data, we find our approach to be practically applicable and scalable.