XMutant: XAI-based Fuzzing for Deep Learning Systems

General Information

This repository contains the tools and the data of the paper "XMutant: XAI-based Fuzzing for Deep Learning Systems"

Repository Structure

The package is structured as follows:

• XMutant-IMDB contains the XMutant tool adapted to the IMDB sentiment analysis case study and the instructions on how to use it;
• XMutant-MNIST contains the XMutant tool adapted to the handwritten digit classification case study and the instructions on how to use it;
• XMutant-LK-ADS contains the XMutant tool adapted to the self-driving car case study and the instructions on how to use it;

Note: each sub-package contains further specific instructions.

Citation

A preprint of the paper can be found on arXiv.

If you use our work in your research, or it helps it, or if you simply like it, please cite it in your publications. Here is an example BibTeX entry:

@misc{2025-Chen-arxiv,
	title        = {{XMutant: XAI-based Fuzzing for Deep Learning Systems}},
	author       = {Xingcheng Chen and Matteo Biagiola and Vincenzo Riccio and Marcelo d'Amorim and Andrea Stocco},
	year         = {2025},
	eprint       = {2503.07222},
	archivePrefix = {arXiv},
	primaryClass = {cs.SE},
	url          = {https://arxiv.org/abs/2503.07222}
}

License

The software we developed is distributed under MIT license. See the license file.