Adversarial ML Research Collections

This repository holds some of the research code regarding research of adversarial machine learning.

Reproduce results in "Adversarial Examples are not bugs, they are features"

Please have a look in code/reproduce_features_not_bugs

Batch Normalization Increases Adversarial Vulnerability and Decreases Adversarial Transferability: A Non-Robust Feature Perspective

If you are interested in the code of ICCV paper, please refer to code/bn_increases_adversarial_vulnerability

Setup

1. Set up the paths in code/config/config.py. The current paths are set with respect to the provided docker file.

Docker

A docker environemt is provided in the docker folder.