This project studies higher-level abstractions for constructing distributed systems that integrate information and computation across administrative and trust domains. Current practice does not offer general, principled techniques for implementing these systems securely. To develop these techniques, fundamental problems of security, consistency, performance, and system evolution are being explored. Problems studied include automatic, adaptive, secure partitioning of programs and data across the nodes of a distributed system; new authorization logics for efficiently managing trust relationships in a distributed system; new methods for increasing performance of distributed systems while guaranteeing strong data consistency; and new ways to securely and consistently evolve the structure of persistent information.
These topics are being studied in the context of Fabric, a new platform for secure distributed computation. Fabric is intended to support secure integration of information systems, a valuable capability for many application domains, including in medicine, finance, education, government, and the military. For example, good methods for integrating distributed information systems would support secure sharing of medical records between institutions.
