Paper 2019/164

Use your Brain! Arithmetic 3PC For Any Modulus with Active Security

Hendrik Eerikson and Claudio Orlandi and Pille Pullonen and Joonas Puura and Mark Simkin

Abstract

Secure multiparty computation (MPC) allows a set of mutually distrustful parties to compute a public function on their private inputs without revealing anything beyond the output of the computation. In recent years, a large effort has undergone into designing and implementing MPC protocols that can be used in practice. This paper focuses on the specific case of three-party computation with an honest majority, which is among the most popular models for real-world applications of MPC. Somewhat surprisingly, despite its significant popularity, there are currently no practical solutions for evaluating arithmetic circuits over real-world CPU word sizes, like 32- and 64-bit words, that are secure against active adversaries that may arbitrarily deviate from the protocol description. Existing solutions are either only passively secure or require the computations to be performed over prime fields, which do not match real-world system architectures. This is unfortunate, since it requires application developers to redesign their applications for the models of computation that are provided by existing MPC frameworks, rather than the MPC frameworks matching the needs of the developers. In this paper we present the first fully-fledged implementation of an MPC framework that can evaluate arithmetic circuits with arbitrary word sizes. Our framework is based on a new protocol, which improves the communication overhead of the best known previous solutions by a factor of two. We provide extensive benchmarks of our framework in a LAN and in different WAN settings, showing that the online overhead for achieving active security is less than two, when compared to the best solutions for the same setting with passive security. Concretely, for the case of 32- and 64-bit words, we show that our framework can evaluate $10^6$ multiplication gates per second.