A Hybrid Machine Learning based Phishing Website Detection Technique through Dimensionality Reduction

Phishing attacks are generally launched through emails or websites to acquire unauthorized access to the user’s sensitive information. In recent times, many users face monetary losses due to phishing attacks. The motivation of our study is to present a prudent framework for detecting phishing websites to save users from being affected. Previous works used several supervised machine learning algorithms for classification to acquire higher accuracy for detection of phishing sites. In this paper, we have proposed a hybrid technique comprising of SVM, Decision tree, Random Forest, XGBoost by combining the idea of bagging and boosting. We have used the features of both phishing and legitimate website to mitigate the risk of phishing websites. We have evaluated classification algorithms using a number of feature subsets selected by various feature selection techniques to ascertain the most effective and efficient subset of features. Our hybrid technique achieved an accuracy of 98.28%, outperforming the state-of-the-art techniques.