The project team has published WinAPI-AdvMal, an open-access dataset for adversarial malware generation and detection evasion experiments, now freely available via Zenodo (DOI: 10.5281/zenodo.20208958). 

The dataset contains 3,799 Windows executable samples across six classes (office, development, multimedia, security, administration tools, and malware) with 2,713 binary features extracted from each sample’s Import Address Table (IAT). It is designed to generate malicious samples that mimic benign software classes, enabling researchers to test the robustness of malware classifiers or train new ones capable of detecting adversarial attacks. For ethical reasons, all file names have been anonymized.