A study of accuracy and reliability of CBIR-based phishing filter

Zexing Luo, Purdue University


In recent years, because of the widespread usage of the Internet and globalization of many financial institutions, phishing attacks targeting these financial institutions has spread wildly across the Internet and caused monetary damage towards their victims. Phishing filters are designed to detect and stop phishing attacks. However, most phishing filters developed recently focus heavily on use text to detect phishing attacks. Although some text-based filter were proven accurate at detecting phishing attacks, the filter itself often have to comprise multiple text-based detection approaches and complicated machine learning algorithms. As a result, many text-based phishing filters are difficult to maintain and cannot flexibly adapt phishing attacks in unfamiliar languages. In this study, a new solution in phishing detection is proposed and tested. The proposed filter employs Content-Based Information Retrieval service offered by popular online search engine in classifying phishing from non-phishing web sites. After testing the proposed filter against 200 phishing and 200 legitimate web sites, the filter achieved 99.75% accuracy rate and 0.25% false positive rate.




Taylor, Purdue University.

Subject Area

Information Technology

Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server