MaSTer: Maliciously Secure Truncation for Replicated Secret Sharing Without Pre-Processing.

Secure multi-party computation (MPC) in a three-party, honest majority scenario is currently the state-of-the-art for running machine learning algorithms in a privacy-preserving manner. For efficiency reasons, fixed-point arithmetic is widely used to approximate computation over decimal numbers. After multiplication in fixed-point arithmetic, truncation is required to keep the result's precision. In this paper, we present an efficient three-party truncation protocol secure in the presence of an active adversary without pre-processing and improve on the current state-of-the-art in MPC over rings using replicated secret sharing (RSS). By adding an efficient consistency check, we lift the efficient but only passively secure three-party truncation protocol from the ABY3 framework by Mohassel and Rindal into the malicious setting without pre-processed data. Our benchmark indicates performance improvements of an order of magnitude in the offline phase for a single batch training. Finally, we apply our protocol to a real-world application for diagnostic prediction based on publicly available ECG heartbeat data. We achieve an improvement by a factor of two in the total throughput for both LAN and WAN settings.