Website Phishing Detection Using Machine Learning Classification Algorithms

Abstract

The growth and development of internet technology have made life a lot better and easier. It has increased the convenience of performing essential transactions with the click of a button using smart devices. Convenience comes with a price, the menace of cybercrimes. In recent years, the incidence of phishing websites has been one of the serious cyber security threats as it leads to the leakage of sensitive personal information. Phishing attacks have become a matter of major concern, and diligent actions and measures must be taken to curtail them. Detection of malicious websites can prevent phishing attacks to a great extent. In this paper, a URL feature-based website phishing detection technique is proposed to predict whether the websites are phishing or not with high accuracy. For this, we explored eight existing machine learning classification algorithms like extreme gradient boosting (XGBoost), random forest (RF), adaboost, decision trees (DT), K-nearest neighbours (KNN), support vector machines (SVM), logistic regression and naïve bayes (NB) to detect malicious websites. From the results it is obvious that XGboost exhibited the highest accuracy of 96.71%. Therefore, based on the implementation outcomes on the phishing website dataset, XGBoost performs better than the other classification algorithms in classifying the websites as phishing or legitimate.