CSIS Seminar

Robust Phishing URL Detection via Generative Models and Graphical Inference

Abstract

Phishing is a power tool of the attacker. Many cyber-attacks start with sending out phishing emails to potential victims. Two topics related to phishing will be presented in this talk. First, I will talk about how generative models can be used to detect (or even hunt) unknown phishing URLs. State-of-the-art generative models can learn about string patterns of phishing URLs and write fake phishing URLs, which we use as additional training samples to improve the detection accuracy. Second, I will talk about a phishing URL detection method robust against evasion. The attacker can design a phishing campaign on top of benign domains, IP addresses, etc. to evade existing phishing URL detection methods. I will discuss a graphical, transductive inference method robust against such evasion attempts.

Speaker Bio

Dr. Noseong Park is an assistant professor at the Center for Secure Information Systems of George Mason University. His research interests include AI and Cybersecurity. He has published papers in KDD, IJCAI, VLDB, WWW, WSDM, ACM TIST, IEEE TIFS, etc.