Paper 2017/550

Committed MPC - Maliciously Secure Multiparty Computation from Homomorphic Commitments

Tore Frederiksen and Benny Pinkas and Avishay Yanay

Abstract

We present a new multiparty computation protocol secure against a static and malicious dishonest majority. Unlike most previous protocols that were based on working on MAC-ed secret shares, our approach is based on computations on homomorphic commitments to secret shares. Specifically we show how to realize MPC using any additively-homomorphic commitment scheme, even if such a scheme is an interactive two-party protocol. Our new approach enables us to do arithmetic computation over arbitrary finite fields. In addition, since our protocol computes over committed values, it can be readily composed within larger protocols, and can also be used for efficiently implementing committing OT or committed OT. This is done in two steps, each of independent interest: 1. Black-box extension of any (possibly interactive) two-party additively homomorphic commitment scheme to an additively homomorphic multiparty commitment scheme, only using coin-tossing and a “weak” equality evaluation functionality. 2. Realizing multiplication of multiparty commitments based on a lightweight preprocessing approach. Finally we show how to use the fully homomorphic commitments to compute any functionality securely in the presence of a malicious adversary corrupting any number of parties.

Note: Increased clarity of construction, added a crucial missing piece of related work and fixed a bug occurring in specific settings.