[Paper Review] Towards High-Throughput Secure MPC over the Internet: Communication-Efficient Two-Party Protocols and Its Application

Satsuya Ohata, Koji Nuida|arXiv (Cornell University)|Jul 8, 2019

Cryptography and Data SecurityComputer Science28 references1 citations

TL;DR

This paper presents communication-efficient, semi-honest two-party secure computation protocols based on secret sharing, leveraging Beaver triple extension to reduce communication rounds and bandwidth. It achieves a 3-round less-than comparison protocol—56.1% faster in online execution and 38.4% fewer communication bits than prior work—making secret-sharing MPC viable for high-throughput Internet deployments despite higher computation costs.

ABSTRACT

Secure multi-party computation (MPC) allows a set of parties to compute a function jointly while keeping their inputs private. Compared with the MPC based on garbled circuits, some recent research results show that MPC based on secret sharing (SS) works at a very high speed. Moreover, SS-based MPC can be easily vectorized and achieve higher throughput. In SS-based MPC, however, we need many communication rounds for computing concrete protocols like equality check, less-than comparison, etc. This property is not suited for large-latency environments like the Internet (or WAN). In this paper, we construct semi-honest secure communication-efficient two-party protocols. The core technique is Beaver triple extension, which is a new tool for treating multi-fan-in gates, and we also show how to use it efficiently. We mainly focus on reducing the number of communication rounds, and our protocols also succeed in reducing the number of communication bits (in most cases). As an example, we propose a less-than comparison protocol (under practical parameters) with three communication rounds. This is less than a half compared with the previous work. Moreover, the number of communication bits is also $38.4\%$ fewer. As a result, total online execution time is $56.1\%$ shorter than the previous work adopting the same settings. Although the computation costs of our protocols are more expensive than those of previous work, we confirm via experiments that such a disadvantage has small effects on the whole online performance in the typical WAN environments.

Motivation & Objective

• To address the high communication round count in secret-sharing-based MPC, which hinders performance in high-latency environments like the Internet.
• To reduce both communication rounds and communication bits in two-party secure computation protocols without compromising security.
• To enable high-throughput secure computation over wide area networks (WANs) by optimizing the online phase of MPC protocols.
• To evaluate the trade-off between increased computation cost and significant gains in communication efficiency in practical WAN settings.

Proposed method

• Introduces Beaver triple extension as a novel technique to efficiently handle multi-fan-in gates in secret-sharing-based MPC.
• Designs communication-optimized protocols by minimizing the number of communication rounds for fundamental operations like equality checks and less-than comparisons.
• Optimizes protocol structure to reduce total communication bandwidth, especially in scenarios with high-latency networks.
• Employs secret sharing with efficient arithmetic over finite fields to enable vectorization and high throughput.
• Uses a semi-honest security model to focus on communication efficiency while maintaining correctness and privacy.
• Employs experimental evaluation under practical parameters to compare performance against prior work.

Experimental results

Research questions

• RQ1 Can Beaver triple extension be effectively used to reduce communication rounds in secret-sharing-based MPC protocols?
• RQ2 To what extent can communication efficiency be improved in two-party MPC for Internet-scale deployments?
• RQ3 How does reducing communication rounds and bandwidth affect the overall online execution time in WAN environments?
• RQ4 What is the trade-off between increased computation cost and reduced communication overhead in practical MPC deployments?

Key findings

• The proposed less-than comparison protocol achieves only three communication rounds, a reduction of more than half compared to previous work.
• The protocol reduces communication bits by 38.4% under practical parameters, significantly lowering bandwidth usage.
• Total online execution time is reduced by 56.1% compared to the prior work with the same configuration.
• Although computation costs are higher, the performance gain from reduced communication dominates in typical WAN environments.
• The results confirm that communication efficiency is the primary bottleneck in Internet-based MPC, and optimizing it yields substantial end-to-end improvements.