A Novel Model Based on Ensemble Learning for Phishing Attack

Abstract

With the increase in the speed of the internet environment and the development of the infrastructures used, people have started to perform most of their work online. As much as this makes life easier, it also increases the possibility of being attacked by malicious people. Attackers can activate a phishing attack that aims to steal information from victims by creating copied, fake websites. While this attack is very old and somewhat simple, it can still be effective due to low IT literacy. People can enter their information on these fake websites out of spontaneity or ignorance or good intentions and be exposed to Phishing attacks. The compromise of a user's account information also puts at risk the security of the organization or institution to which it is connected. In this study, we propose a new machine learning-based ensemble model with feature selection methods to detect phishing attacks. Also, an ablation study is presented to measure the effect of different feature selection methods. The proposed model which we named as NaiveStackingSymmetric (NSS) is analyzed using the widely used accuracy (ACC), the area under curve (AUC), and F-score metrics as well as the polygon area metric (PAM), and it is shown that it outperforms other studies in the literature using the same dataset.