Paper 2021/082

Grades of Trust in Multiparty Computation

Jaskaran V. Singh and Nicholas Hopper


Secure Multiparty Computation involves a protocol between parties with an aim to produce a computed result just as a trusted party would produce if the parties provided their inputs to it. The trusted party in conventional computation is replaced with "un-trusted" parties in Secure Multiparty Computation. We first show that this existing binary definition of trust is inadequate. Real world is rife with disparities, that which produce a perceivable trust gradient between the participants. Conventional MPC models do not take this into account and rather provide security guarantees based on the thresholds of the number of corrupted parties. The thresholds are supposed to cover for some of the parties turning out to be corrupt. Often, with the knowledge of prior probability of a party being corrupt, we can do better if we allot weight to each party based on how trusted we perceive it to be. Our paper explores this idea and our contributions towards it are three folds. First, we introduce the Graded Trust model where each party essentially has a "trust grade" assigned to it in the protocol based on the prior of it being corrupt. Second, we present a method for protocol translation and execution by by simulating players. Lastly, we present a discussion on the philosophy behind graded trust, and the potential benefits for large scale public MPC systems.

Available format(s)
Cryptographic protocols
Publication info
Multiparty ComputationTrustPlayer Simulation
Contact author(s)
singh882 @ umn edu
2021-01-29: revised
2021-01-27: received
See all versions
Short URL
Creative Commons Attribution


      author = {Jaskaran V.  Singh and Nicholas Hopper},
      title = {Grades of Trust in Multiparty Computation},
      howpublished = {Cryptology ePrint Archive, Paper 2021/082},
      year = {2021},
      note = {\url{}},
      url = {}
Note: In order to protect the privacy of readers, does not use cookies or embedded third party content.