The Machine Learning Based Phishing Detection Techniques as a Unary and Binary Classification Problem

Abstract

in recent studies we have realized most of the phishing attacks the URLs are deployed on target user’s device to capture their sensitive, private and confidential information. In this context the rule based classification system is proposed to deal with the URL classification problem. Recently we have treated this URL classification problem as unary classification problem, now we have collected some of the legitimate URLs to prepare a complete dataset. Further the model is implemented with the help of decision tree algorithms i.e. C4.5 and CART. By using these supervised learning classifiers the experiments has been carried out. The experimental analysis has been carried out based on 50% phish tank data and 50% real world data with the manually assigned class labels. The performance in terms of accuracy demonstrates the performance of classification by using supervised learning and by defining URLs in both the labels. The experimental results confirm the higher detection rate as compared to previously offered methodology.