README

This is for releasing the source code of the paper "Query Efficient Decision Based Sparse Attacks Against Black-Box Machine Learning Models"

Archived Version: SparseEvo Attack.

Poster: ICLR 2022 Poster

The project is published as part of the following paper and if you re-use our work, please cite the following paper:

@inproceedings{vo2022,
title={Query Efficient Decision Based Sparse Attacks Against Black-Box Machine Learning Models},
author={Viet Quoc Vo and Ehsan Abbasnejad and Damith C. Ranasinghe},
year = {2022},
journal = {International Conference on Learning Representations (ICLR)},
}

The source code is written mostly on Python 3 and Pytorch, so please help to download and install Python3 and Pytorch beforehand.

Requirements

To install the requirements for this repo, run the following command:

git clone https://github.com/SparseEvoAttack/SparseEvoAttack.github.io.git
cd SparseEvoAttack
pip3 install -r requirements.txt

Download a pretrained model for CIFAR-10 evaluation set here.

Evaluate SparseEvoAttack

Run step-by-step with the Jupyter Notebook file 'Tutorial - Evaluate Sparse Attacks on CIFAR10.ipynb'.

TODO

• Add the testing code.