Practical evaluation of chain-like MPC protocols

Hanna Jonson ()

Abstract

Bottleneck Complexity (BNC) is a fairly new metric for Secure Multi-Party Computation (SMPC) protocols. Two protocols with the BNC in mind are examined in this thesis. One protocol is successfully implemented in the Python programming language, and the BNC, latency and computational complexity of the protocol is examined with positive results. The BNC is independent on the number of parties in the protocol. The latency is linearly increasing with each new party added, and the computational complexity is linearly dependent on the bitlength of the numbers computed on. The second protocol was not implemented because of the complexity of the Garbled Circuits that would have had to been created in order to perform encryption and decryption of an Additively Homomorphic Encryption (AHE) scheme. Instead, a conclusion is drawn that while garbled circuits serves a good purpose for secure computation in simple settings, computations that require complicated operations scale up very quickly in the number of gates needed. A github repository with the code for the first protocol is found at https://github.com/ha7008/ChainMPC.