Classification of Phishing Websites using Machine Learning Techniques

Abstract

Phishing detection is a momentous problem which can be deliberated by many researchers with numerous advanced approaches. Current anti-phishing mechanisms such as blacklist-base anti-phishing, Heuristic-based anti-phishing does suffer low detection accuracy and high false alarm. There is need for efficient mechanism to protect users from phishing websites. The purpose of this study is to investigate the capability of 6 machine learning algorithms i.e. Multi-Layer Perceptron (MLP), Support Vector Machine (SVM), Random Forest (RF), Decision Tree (DT), Logistic Regression (LR) and Naïve Bayes (NB) to classify phishing and non-phishing websites. These algorithms were trained with two different groups of training in WEKA environment and then were tested in terms of accuracy, precision, TP rate, and FP rate on a 3 different sets of dataset which contains dissimilar portion of phishing and non-phishing instances. Results presented that Naïve Bayes classifier has better detection accuracy between other classifiers for predicting phishing websites while Multi-Layer Perceptron gave worst result in terms of detection accuracy. The result also showed that Support Vector machine has better FP rate between other classifier. In addition, Random Forest, Decision Tree, and Naïve Bayes can classify all phishing websites as phishing correctly. It means that TP rate is 100% for these classifiers. In conclusion this paper suggests using NB as the best classifier for predicting phishing and non-phishing websites.