Consistency Testing of Phishing Websites based on Random Model Approach using Classification Algorithms

Abstract

Phishing is the act of attempting to trick people into parting with their sensitive information such as usernames, passwords, credit card and bank details, by way of email spoofing, instant messaging, or by using fake web sites whose look and feel make them seem legitimate. However, upon detailed examination, it is discovered that dealing with phishing websites is quite a challenging task. To overcome these problems, a method has been proposed and developed by the authors of this paper, for validating websites based on the Behavioral Random Model approach. This approach is defined by eight sets of features, which are in turn based on three types of Heuristics with a random number of inputs and corresponding responses. Subsequently, the most frequently occurring feature-combinations between phishing and legitimate websites were identified, in order to infer the integrity of each site. A tool named Phish Detector was implemented for automating the testing process. The results show the approach to have 100% true negatives and 0.03% false positives in detecting phishing and legitimate websites. The results were then classified on the basis of the classifier algorithms.