Scalable secure computation from ANSI-C

Abstract

With Secure Multi-party Computation (MPC) becoming more and more practical, tools for automatized application development become of importance. Therefore, multiple compilers have been developed that translate high-level functionality descriptions into Boolean circuits, optimized for their use in MPC protocols. Previous efforts in compiler development for MPC either focused on domain specific (hardware) languages or only provided limited optimization techniques. Yet, for the practical use of MPC, compilers supporting common programming languages are desirable to provide developers an accessible interface to MPC, and to enable reusability of existing code. In this work we present a novel compilation approach from ANSI-C to Boolean circuits that allows scalable compilation with advanced optimization techniques. To achieve an efficient compilation, we present a source code decomposition approach based on static analysis and introduce the idea of source code guided circuit minimization. Multiple experiments illustrate the efficiency of our approach. For instance, comparing with other recently presented compilers, we report a circuit size that is 70% smaller for the example application of a floating point matrix multiplication with 22.8 million non-linear gates.