SMPCEngine: An N-Party Implementation of the Secure Multiparty Private Computation Protocol

Abstract

In this thesis, we implement a framework for secure multiparty computation. Our framework works in the commodity-based model, where the players running the distributed computation receive pre-distributed data from a trusted source during a setup phase. The framework allows secure multiplications of field elements, secure multiplications of matrices and matrix inversions. Differently from previous proposals, in our framework, the running times do not increase significantly as more players are added to the protocol. We illustrate the power of our solution by applying it to the problem of privately esti- mating driver’s drowsiness based on EEG data. We show that our private solution is practical and achieves a similar accuracy as a solution in the clear, where there are no privacy concerns.