Gravitational Search Based Feature Selection for Enhanced Phishing Websites Detection

Abstract

Phishing is one of the most prominent online security threat that steals the user credentials through fraudulent websites by masquerading as a legitimate website. Many online users have experienced the monitory losses due to this fraudulent attempt of getting financial credentials of user via illegitimate email. The existing phishing detection strategies mainly consider the web page features and URL characteristics. Though numerous phishing detection strategies are proposed in the literature, very few studies consider the feature selection scheme that eliminates the irrelevant or insignificant features for phishing websites detection problem. In this paper, we explored the significant features for detecting the phishing websites based on their relevance to the detection accuracy. The Gravitational Search Algorithm (GSA) is deployed as the feature selector tool to choose the most significant features from the benchmark phishing datasets. Then, the efficiency of the selected features are evaluated using different classification algorithms. From the experimental results it is stated that the features that are selected using GSA performed better than the other feature subsets for detecting the phishing websites.