Awesome

Noir GMiMC for BN254

This repository contains a Noir crate implementing the GMiMC block cipher for the native curve BN254.

GMiMC, or the Generalized MiMC family, is an extension of the Feistel-MiMC. Instead of using two branches, GMiMC generalizes the Feistel structures with $t (\geq 2)$ branches. GMiMC with $t=2$ is Feistel-MiMC.

GMiMC uses a Feistel round function where the mapping is $x \rightarrow x^d$. This construction leads to a cheap instantiation for SNARKs. Due to the characteristics of the underlying primefield, this particular implementation uses the round function $x \rightarrow x^5$.

For further information, you can read the GMiMC Paper.

Installation

In your Nargo.toml file, add the following dependency:

[dependencies]
gmimc = { tag = "v0.3.0", git = "https://github.com/TaceoLabs/noir-GMiMC" }

Examples

Our API uses an expanded form to call the GMiMC block cipher, similar to the Poseidon API from the Noir standard library to encrypt multiple Field elements at once. We use this approach so that we can use precomputed round constants (see Sage Section below) instead of computing them on the fly. To encrypt two Field elements with a given key (which is also a Field element) write:

    let plain = [3,4];
    let key = 0x2129347902376149;
    let cipher = gmimc::bn254::enc::x5_2(plain, key);
    let is_plain = gmimc::bn254::dec::x5_2(cipher, key);
    assert(is_plain == plain);

Note: The GMiMC block cipher has a configurable branch width, the example above uses a 2-branch Feistel network (Feistel-MiMC). If encrypting a large number of blocks, using an instance with a larger branch number is more efficient.

For further examples on how to use the GMiMC crate, have a look in the lib.nr file in the src/ directory. You can find the tests for all expanded forms of the GMiMC block cipher there.

Sage Script

You can find a Sage implementation of GMiMC under tools/. We used this script to generate the round constants defined in src/bn254/consts.nr. The script was written with all currently known attacks and analysis in mind.

Disclaimer

This is experimental software and is provided on an "as is" and "as available" basis. We do not give any warranties and will not be liable for any losses incurred through any use of this code base.