Description

• What & Why:
  • Benchmark the statistics operations, data operations, and their combination against the suitable MPC protocols provided by MP-SPDZ.
  • Identify candidates that align with our trust assumptions.

Expected Outcome:

• Determine which protocols to use for our scenarios.

Benchmark Results

A caveat to the results is that due to the availability of resources on the AWS instances used for benchmarking, computations that required more resources than were available did not complete. The results for such cases have been excluded from the result spreadsheets.

Instances Used

All benchmarks were conducted using AWS t3.medium instances.

For the 3-region benchmarks, the instances were distributed across three regions: ap-northeast-1 (Asia Pacific, Tokyo), us-west-2 (US West, Oregon), and eu-central-1 (Europe, Frankfurt).

1 Machine (3 Parties)

• Mean without magic number
  • https://docs.google.com/spreadsheets/d/18kgFB7-U1rbL9JINRUkl7wyreaZRQtH_7m16aj0k3yA/edit?usp=sharing
• Mean with magic number
  • https://docs.google.com/spreadsheets/d/1wcdAwuhzdmwJem4WoJap3CQmhPYIM-780HbsjJlM4rA/edit?usp=sharing
• TLSNotary commitment
  • https://docs.google.com/spreadsheets/d/1CwOw5xGDCf1QF6WEMHLHUiDhO5fd7vq6_cTj_Lqvw1k/edit?usp=sharing

3 Machines - 1 Region

• Mean without magic number

  https://docs.google.com/spreadsheets/d/19EE-_QrrlKFXkljOo3XwiqePX0x1AyfrouHCKVEhQdQ/edit?usp=sharing

• Mean with magic number

  • https://docs.google.com/spreadsheets/d/1WaBMmUfdfLnjqPHZJFNnHcOfrB9FA4c6oVI4mG8OMG4/edit?usp=sharing

• TLSNotary commitment

  • https://docs.google.com/spreadsheets/d/1c3TWATtgUG22Kk2fjxt4gLhofWcfQT5fgndkT8XBNMg/edit?usp=sharing

3 Machines - 3 Regions

• Mean without magic number
  • https://docs.google.com/spreadsheets/d/1wIZ_NN7wiYMErnoVZXOQwle94S6P2wv0qUEbslOX0MM/edit?usp=sharing