Comparison of association rule mining with pruning and adaptive technique for classification of phishing dataset

摘要

Phishing is an e-mail fraud method in which the perpetrator sends out legitimate-looking email in an attempt to gather personal and financial information from any online users. The hackers then steal this personal information for their own purposes, or sell the information to any other criminal parties. There are various techniques to detect phishing websites. However, Classification Data Mining (DM) Techniques can be a very useful tool in detecting and identifying e-banking phishing websites. This paper proposes two algorithms to overcome the difficulty and complexity in detecting and predicting e-banking phishing website. First approach known as Multiclass Classification based on Association Rule (MCAR) based on using association and classification Data Mining algorithms. The algorithms used to characterize and identify all the factors and rules in order to classify the phishing website and the relationship that correlate them with each other. Another approach called Adaptive Boosting (Adaboost) is also a powerful classifier in which final classification is based on vote of weak classifiers. This paper proposes the modification to original MCAR algorithms with redundant rule pruning technique to reduce number of rules generated in final classifier and minimizes redundancy, hence improves the accuracy. In this paper Adaboost classifier is first time applied to detect phishing website. Results are compared with respect to time, accuracy of MCAR and Adaboost. We also analysed MCAR for generated rule to discarded rule with support and confidence values as a result of pruning techniques applied. The experimental result demonstrates the feasibility of using Associative Classification techniques in real applications and its better performance improvement as compared to Adaboost and other classifications algorithms.