MADAR: Efficient Continual Learning for Malware Analysis with Diversity-Aware Replay

This dataset is released in support of the paper:

MADAR: Efficient Continual Learning for Malware Analysis with Diversity-Aware Replay
Mohammad Saidur Rahman, Scott Coull, Qi Yu, Matthew Wright
arXiv preprint arXiv:2502.05760, 2025

MADAR is a benchmark suite for evaluating continual learning methods in malware classification. It includes realistic data distribution shifts and supports scenarios such as Domain-Incremental Learning (Domain-IL) and Class-Incremental Learning (Class-IL). The dataset includes curated samples from two primary sources:

• EMBER-Domain: Derived from the EMBER dataset of Windows PE files.
• AZ-Domain: Derived from the AndroZoo dataset of Android APKs.

---

Dataset Sources

EMBER-Domain

Curated from the EMBER dataset:

Hyrum S. Anderson and Phil Roth
Ember: An open dataset for training static PE malware machine learning models
arXiv preprint arXiv:1804.04637, 2018

AZ-Domain

Curated from the AndroZoo dataset:

Kevin Allix, Tegawendé F. Bissyandé, Jacques Klein, Yves Le Traon
AndroZoo: Collecting Millions of Android Apps for the Research Community
International Conference on Mining Software Repositories (MSR), 2016

Marco Alecci, Pedro Jesús Ruiz Jiménez, Kevin Allix, Tegawendé F. Bissyandé, Jacques Klein
AndroZoo: A Retrospective with a Glimpse into the Future
International Conference on Mining Software Repositories (MSR), 2024

---

License

This dataset is released under the MIT License.

---

Citation

If you use MADAR in your work, please cite:

@article{rahman2025madar,
  title={MADAR: Efficient Continual Learning for Malware Analysis with Diversity-Aware Replay},
  author={Rahman, Mohammad Saidur and Coull, Scott and Yu, Qi and Wright, Matthew},
  journal={arXiv preprint arXiv:2502.05760},
  year={2025}
}