Geometric Decision-based Attack (GeoDA)

This repository contains the official PyTorch implementation of GeoDA algorithm described in [1]. GeoDA is a Black-box attack to generate adversarial example for image classifiers.

A few examples on the performance of the GeoDA for different norms

Requirements

To execute the code, please make sure that the following packages are installed:

• Foolbox
• NumPy
• PyTorch and Torchvision (install with CUDA if available)
• matplotlib

Contents

GeoDA.py

This function implements the GeoDA algorithm.

utils.py

Includes general functions

data/

Contains some examples for the demos.

Reference

[1] Ali Rahmati, Seyed-Mohsen Moosavi-Dezfooli, Pascal Frossard, and Huaiyu Dai, GeoDA: a geometric framework for black-box adversarial attacks. in CVF/IEEE Computer Vision and Pattern Recognition (CVPR'20), 2020. [arXiv pre-print]