Examining the Robustness of Machine Learning-based Phishing Website Detection: Action-Masked Reinforcement Learning for Automated Red Teaming

Yang Gao, Benjamin M. Ampel, Sagar Samtani
Last updated on Sep 14, 2026
PDF DOI
In Plain Terms

Attackers can dodge AI tools that detect phishing websites by subtly altering the site’s code. This paper builds an automated “red team” that uses reinforcement learning to realistically mimic those evasion tricks, with safeguards so the altered sites still look normal, providing a scalable way to test how robust phishing detectors really are before they’re deployed.

Artifacts

PDF DOI

Related Papers

Cite this

APA Gao, Y., Ampel, B. M., & Samtani, S. (2025). Examining the Robustness of Machine Learning-based Phishing Website Detection: Action-Masked Reinforcement Learning for Automated Red Teaming. In IEEE Security and Privacy Workshops (SPW) - Human-Machine Intelligence for Security Analytics (HMISA). https://doi.org/10.1109/SPW67851.2025.00041
BibTeX
@inproceedings{gao2025examining,
  title={Examining the Robustness of Machine Learning-based Phishing Website Detection: Action-Masked Reinforcement Learning for Automated Red Teaming},
  author={Yang Gao and Benjamin M. Ampel and Sagar Samtani},
  booktitle={IEEE Security and Privacy Workshops (SPW) - Human-Machine Intelligence for Security Analytics (HMISA)},
  year={2025},
  doi={10.1109/SPW67851.2025.00041}
}
Phishing Machine Learning Reinforcement Learning Red Teaming Adversarial Machine Learning
Benjamin M. Ampel

Benjamin M. Ampel

Assistant Professor in Computer Information Systems and Director, Center for CyberAI Research (CCAIR)

My research focuses on AI-enabled Cybersecurity, including Cyber Threat Intelligence, Large Language Models, and Phishing Detection.