The rapid advancement of techniques for secure computation on protected data offers a major incentive for development of tools for general-purpose secure computation that protects data privacy, as opposed to computation of specialized tasks. The recent emergence of cloud computing and the need to protect privacy of sensitive data used in outsourced computation serves as another major motivation for this work. With this in mind, this project targets at developing a compiler suitable for privacy-preserving execution of any functionality specified by a user program. Our compiler transforms a program written in an extension of the C programming language, where data to be protected are marked as private, into its secure distributed implementation in C, which can be consecutively compiled into a binary executable by a native compiler. The goal of the project is to preserve all features of the programming language and enable efficient implementation of user programs using a number of optimization techniques. The secure distributed computation is implemented using secure multi-party techniques on secret-shared data and is suitable for secure collaborative computation and secure computation outsourcing. These techniques will allow for more secure cloud computing, even in the presence of attackers. Other broader impacts include outreach to female middle school girls to increase interest in computer science, a summer school program for graduate students on secure computation in the cloud, and involving students from a nearby predominantly African-American university in the research program.
TWC: Small: Techniques and Tools for General-Purpose Secure Computing and Outsourcing