Abstract: Due to colossal financial losses in recent years, phishing has drawn attention of most of the individuals and organizations in the world of internet. Need for protection against phishing activities through fraudulent emails has increased remarkably. In this paper we propose a hybrid model to classify phishing emails using machine learning algorithms with the aspiration of developing an ensemble model for email classification with improved accuracy. We have used the content of emails and extracted 47 features from it. The processed emails are provided as input to various machine learning classifiers. Going through experiments, it is observed and inferred that Bayesian net classification model when ensemble with CART gives highest test accuracy of 99.32%.

Keywords: Phishing, Machine learning, Email classification, Hybrid model.