Turbospeedz: Double your online SPDZ! Improving SPDZ using function dependent preprocessing

Aner Ben-Efraim, Michael Nielsen, Eran Omri

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

14 Scopus citations

Abstract

Secure multiparty computation allows a set of mutually distrusting parties to securely compute a function of their private inputs, revealing only the output, even if some of the parties are corrupt. Recent years have seen an enormous amount of work that drastically improved the concrete efficiency of secure multiparty computation protocols. Many secure multiparty protocols work in an “offline-online” model. In this model, the computation is split into two main phases: a relatively slow “offline phase”, which the parties execute before they know their input, and a fast “online phase”, which the parties execute after receiving their input. One of the most popular and efficient protocols for secure multiparty computation working in this model is the SPDZ protocol (Damgård et al., CRYPTO 2012). The SPDZ offline phase is function independent, i.e., does not require knowledge of the computed function at the offline phase. Thus, a natural question is: can the efficiency of the SPDZ protocol be improved if the function is known at the offline phase? In this work, we answer the above question affirmatively. We show that by using a function dependent preprocessing protocol, the online communication of the SPDZ protocol can be brought down significantly, almost by a factor of 2, and the online computation is often also significantly reduced. In scenarios where communication is the bottleneck, such as strong computers on low bandwidth networks, this could potentially almost double the online throughput of the SPDZ protocol, when securely computing the same circuit many times in parallel (on different inputs). We present two versions of our protocol: Our first version uses the SPDZ offline phase protocol as a black-box, which achieves the improved online communication at the cost of slightly increasing the offline communication. Our second version works by modifying the state-of-the-art SPDZ preprocessing protocol, Overdrive (Keller et al., Eurocrypt 2018). This version improves the overall communication over the state-of-the-art SPDZ.

Original languageEnglish
Title of host publicationApplied Cryptography and Network Security - 17th International Conference, ACNS 2019, Proceedings
EditorsValérie Gauthier-Umaña, Robert H. Deng, Martín Ochoa, Moti Yung
Pages530-549
Number of pages20
DOIs
StatePublished - 2019
Event17th International Conference on Applied Cryptography and Network Security, ACNS 2019 - Bogota, Colombia
Duration: 5 Jun 20197 Jun 2019

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume11464 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference17th International Conference on Applied Cryptography and Network Security, ACNS 2019
Country/TerritoryColombia
CityBogota
Period5/06/197/06/19

Keywords

  • Concrete efficiency
  • Offline/online
  • SPDZ
  • Secure multiparty computation

Fingerprint

Dive into the research topics of 'Turbospeedz: Double your online SPDZ! Improving SPDZ using function dependent preprocessing'. Together they form a unique fingerprint.

Cite this