Accuracy/Efficiency Trade-Off for Privacy-Preserving Division Protocol

Secure multi-party computation (MPC) allows a set of parties to jointly compute a function, while keeping their inputs private. We can consider many applications of MPC and various operations/protocols for MPC have been proposed. We focus on (privacy-preserving) division protocols in this paper. Although division is included in four arithmetic operations, we need extremely high computation/communication costs in privacy-preserving settings compared with other operations. Especially, we need many communication rounds for error correction to get accurate quotients. Since we iterate error correction procedure for several times in the division protocol, we can expect to reduce communication rounds by removing error correcting iterations. In this strategy, however, we cannot obtain accurate quotients. In this paper, we show experimental results about a relation between accuracy of quotients obtained by the round-reduced division protocol and the number of communication rounds we need. From our results, we find the error of quotients becomes less than 0.1% even if we reduce the number of communication rounds for error correction to 33%. This property will be useful when we make concrete applications efficient in some cases.