Improving the accuracy of Search Engine based anti-phishing solutions using lightweight features

Abstract

A drawback of the Search Engine (SE) based anti-phishing technique is that authentic websites that are newly launched over the web are classified as phishing websites due to their low ranking. This causes a noticeable reduction in the True Negative Rate (TNR) of such schemes. Some SE based techniques use other complex anti-phishing techniques or features in combination to increase the TNR which results in higher accuracy. However, such an addition dilutes the importance of SE based techniques, which are intended to make anti-phishing lightweight, client side implementable and efficient. In this paper, we study a set of lightweight features including https, digits in domain name and Google Safe Browsing (GSB) API check, and the effect of their addition to the results of a SE based solution. An eventual effect on the accuracy of the existing SE based schemes is also reflected in this paper. A total shift of nearly 96% from False Positive Rate (FPR) to TNR is recorded with a slight decrease of 0.1% in True Positive Rate (TPR), thus increasing the accuracy by 3.28%.