/nimpc

MPC for Nim

Primary LanguageNimMIT LicenseMIT

NiMPC

A secure multi-party computation (MPC) library for the Nim programming language. Allows you to do joint computations with multiple parties while each party keeps their inputs private. We're working towards an implementation of the SPDZ2k protocol.

CI Status

Work in Progress

This is very much a work in progress. At least the following still needs to be implemented before it can be considered feature complete:

  • A networking layer, to run multiple parties on separate machines.
  • Support for more mathematical operations, such as division and square root.
  • MAC checking, to make computations safe against active adversaries.
  • OT extensions, to make multiplications faster and to circumvent a known flaw in the Oblivious Transfer protocol (see section 1.1. of the OT paper).

Installation

Only Linux and macOS are currently supported.

Ensure that you have a recent version of Nim installed.

Use the Nimble package manager to add NiMPC to an existing project. Add the following to its .nimble file:

requires "NiMPC >= 0.4.0 & < 0.5.0"

Examples

Local Computation

Perform a computation with two parties. For now, both parties will run on the same machine, we'll cover how to connect parties over a network later.

Import these modules:

import NiMPC
import asyncdispatch

Create two parties that will jointly perform a computation.

let party1, party2 = newLocalParty()

Calling destroy when the local parties are no longer in scope ensures that their secret keys are securely wiped from memory.

defer: destroy(party1, party2)

Connect the parties.

connect(party1, party2)

Specify the computation that the first party will perform. We'll use the async and await macro's from the asyncdispatch module.

proc computation1 {.async.} =

The first party shares a secret input with the other party. Currently only 32 bit unsigned numbers (uint32) are supported.

  let input1: Secret = party1.share(21)

Obtain a share of the secret input from the second party. The obtained secret does not reveal anything about the number that was input by the other party.

  let input2: Secret = party1.obtain(party2)

The two inputs are multiplied. Addition and subtraction are also supported.

  let product: Secret = input1 * input2

The product is revealed. This allows both parties to learn the product of the two secret inputs.

  let revealed: uint32 = await product.reveal() # equals 42

Specify the equivalent computation for the second party:

proc computation2 {.async.} =
  let input1: Secret = party2.obtain(party1)
  let input2: Secret = party2.share(2)
  let product: Secret = input1 * input2
  let revealed: uint32 = await product.reveal() # equals 42

Run the multi-party computation:

waitFor all(computation1(), computation2())

Networking

Perform a computation with two parties over a network connection. Just like the previous example we start with two parties, but we do not connect them just yet.

import NiMPC
import asyncdispatch

let party1, party2 = newLocalParty()
defer: destroy(party1, party2)

These parties will listen on different ports on localhost.

const host = "localhost"
const port1 = Port(23455)
const port2 = Port(23456)

Specify the computation that the first party will perform.

proc computation1 {.async.} =

Start listening for incoming messages on the correct port.

  let listener = party1.listen(host, port1)

Ensure that we stop listening when the listener is no longer in scope:

  defer: await listener.stop()

Connect to the second party. This returns a proxy for the Party that is running on the other side of the connection.

  let proxy2 = await party1.connect(party2.id, host, port2)

Ensure that we close the connection when the proxy is no longer in scope.

  defer: proxy2.disconnect()

Use the local party and the proxy for the remote party to perform the computation.

  let input1 = party1.share(21)
  let input2 = party1.obtain(proxy2)
  let product = input1 * input2
  let revealed = await product.reveal() # equals 42

Specify the equivalent computation for the second party:

proc computation2 {.async.} =
  let listener = party2.listen(host, port2)
  defer: await listener.stop()

  let proxy1 = await party2.connect(party1.id, host, port1)
  defer: proxy1.disconnect()

  let input1 = party2.obtain(proxy1)
  let input2 = party2.share(2)
  let product= input1 * input2
  let revealed = await product.reveal() # equals 42

Run the multi-party computation:

waitFor all(computation1(), computation2())

Thanks

NiMPC is inspired by the great work done by the FRESCO and MPyC developers. It would also not have existed without the research that went into the SPDZ2k protocol.