Machine Learning-Based Detection of Phishing in COVID-19 Theme-Related Emails and Web Links

Abstract

During the COVID-19 epidemic phishing dodges increased in frequency mostly the links provided current updates about COVID-19 hence it became easy to trick the victims. Many research studies suggest several solutions to prevent those attacks but still phishing assaults upsurge. There is no only way to perform phishing attacks through web links attackers also perform attacks through electronic mail. This study aims to propose an Effective Model using Ensemble Classifiers to predict phishing using COVID-19-themed emails and Web Links. Our study comprises two types of Datasets. Dataset 1 for web links and Dataset 2 for email. Dataset 1 contains a textual dataset while Dataset 2 contains images that were downloaded from different sources. We select ensemble classifiers including, Random Forest (RF), Ada Boost, Bagging, ExtraTree (ET), and Gradient Boosting (GB). During the analysis, we observed that Dataset 1 achieves the highest accuracy rate as compared to Dataset 2 which is 88.91%. The ET classifier performs with an accuracy rate of 88.91%, a precision rate of 89%, a recall rate of 89%, and an f1 score of 89% which is better as compared to other classifiers over both datasets. Interesting concepts were found during the study.