Margin Consistency

Paper: "Detecting Brittle Decisions for Free: Leveraging Margin Consistency in Deep Robust Classifiers."

Installation

1. Clone this repository with git clone [email protected]:ngnawejonas/margin-consistency.git
2. Install the requirements pip install -r requirements.txt

Running the code

1. Edit the configuration file params.yaml to specify:
   • the attack (fab for fab attack, cw for carlini-wagner, clever for clever score or otherwise for auto-attack)
   • a folder to save the results
2. Run $python3 eval.py or use the script run.sh

Analysis

Use notebook/xpLinf-TrainDNet.ipynb for analysis

Citation

If you used this code, please cite our paper:

@inproceedings{
ngnawe2024detecting,
title={Detecting Brittle Decisions for Free: Leveraging Margin Consistency in Deep Robust Classifiers},
author={Jonas Ngnawe and Sabyasachi Sahoo and Yann Batiste Pequignot and Frederic Precioso and Christian Gagn{\'e}},
booktitle={The Thirty-eighth Annual Conference on Neural Information Processing Systems},
year={2024},
url={https://openreview.net/forum?id=XHCYZNmqnv}
}