This repository contains the code for the paper FairProof : Confidential and Certifiable Fairness for Neural Networks, accepted at ICML 2024.
The main part of the code is in the file main.go which contains the function the main function and the Prove_Verify function. To run this code, the most important components needed are the model weights, the input point and a dictionary containing a point for each polytope (or the polytopes that we expect to be visited) for the neural network model. The formats for each file are listed at the beginning of the main.go file. Some example jsons for each input are also present in the 'example_files' folder. Feel free to change this format (this would also require downstream changes in the way the inputs are read).