Anti-phishing model for phishing websites detection: using pruning decision tree

Abstract

As a new form of malicious software, phishing websites appear frequently in recent years, Phishers use spoofed emails and fraudulent web sites to lure unsuspecting online users into giving up personal information. A lot of researches have been conducted to mitigate phishing websites. They used common phishing websites features in their investigation process. Some of these features do not have significant value on accuracy ratio which can affect the performance in terms of computation time. In this study, few new significant features for phishing websites are suggested. Meanwhile, Pruning Decision Tree is used for dectection of phishing websites because it is capable of balancing the computation time and accuracy ratio. Pessimistic Error Pruning is used as a pruning algorithm to prune the decision tree leafs without affecting the accuracy. This study also focused on categorization of phishing websites. The purpose of categorization process is to give significant and specific tips to increase the awareness level among users for each category. This study consists three main phases. First phase focused on dataset gathering, preprocessing, features extraction, dataset normalization and dataset division in order to make the dataset suitable for the classification process. Second phase focused on setup process of Decision Tree with Pessimistic Error Pruning technique. Third phase focueed on evaluation of results in terms of accuracy ratio, false positive and computation time. In addition, third phase focuesd on categorization of phishing websites. The result of this study shows an accuracy ratio of 99.12% before and after Pruning. That means the Pessimistic Error Pruning did not affact the accuracy ratio but it reduced the leafs of Decision tree to affect the computation time positively.