Title:High-Throughput Secure Multiparty Computation with an Honest Majority in Various Network Settings

Abstract:In this work, we present novel protocols over rings for semi-honest secure three-party and malicious four-party computation with one corruption. Compared to state-of-the-art protocols, our protocols achieve the best-known communication complexity and reduced computational complexity. We benchmark our protocols and state-of-the-art protocols on a 32-Core CPU and a 25 Gbit/s network. In homogeneous network settings, where parties communicate equally on all links, our protocols achieve up to 25 percent faster runtimes than state-of-the-art protocols. In these settings, all of our protocols achieve more than 40 billion $AND$ gates per second or more than 1 billion 32-bit multiplications per second. Our semi-honest 4-PC protocol even achieves more than 70 billion $AND$ gates per second or over 130 billion if only the online phase is measured. These are the highest throughputs achieved in 3-PC and 4-PC so far and more than three orders of magnitude higher than the throughput MP-SPDZ achieves in same settings. For heterogeneous network settings, we present variations of our protocols that only require two high-bandwidth links and one low-latency link between the parties. In these settings, our semi-honest 3-PC protocol still achieves 24 billion $AND$ gates per second, while our malicious 4-PC protocol still achieves approximately 10 billion even if all other links between the parties share an arbitrarily weak connection.