LEARNING TO DETECT PHISHING URLS

Ram B. Basnet, Andrew H. Sung, Quingzhong Liu

Abstract: Phishing attacks have been on the rise and performing certain actions such as mouse hovering, clicking, etc. on malicious URLs may cause unsuspecting Internet users to fall victims of identity theft or other scams. In this paper, we study the anatomy of phishing URLs that are created with the specific intent of impersonating a trusted third party to trick users into divulging personal data. Unlike previous work in this area, we only use a number of publicly available features on URL alone; in addition, we compare performance of different machine learning techniques and evaluate the efficacy of real-time application of our method. Applying it on real-world data sets, we demonstrate that the proposed approach is highly effective in detecting phishing URLs with an error rate of 0.3%, false positive rate of 0.2% and false negative rate of about 0.5%, thereby improving previous results on the important problem of phishing detection.

Keywords: Phishing URL, phishing websites, machine learning, web mining, phishing attack, URL classification

DOI: https://doi.org/10.15623/ijret.2014.0306003