Phishing Website Detection using XGBoost and Catboost Classifiers

Abstract

Malicious websites infecting users' devices has become a common phenomenon. Users seldom pay attention to the URL details and ultimately fall prey to these websites, which results in personal data theft. Attackers try to mimic legitimate URLs making it difficult to identify them. Therefore, it is imperative to identify these websites. In this paper we evaluated the performance of XGBoost and Catboost: tree-based classifiers in detecting these phishing websites. Both the classifiers performed well in terms of accuracy. The result particularly shows the XGBoost performing slightly better than Catboost. We analyzed these classifiers over two datasets. To strengthen this outcome, k-fold validation as well as train-test validation are used in this study. Furthermore, we compared the performance of XGBoost and Catboost with other conventional classifiers.