A Proposed Bucket Based Feature Selection Technique (BBFST) for Phishing e-Mail Classification

Phishing e-mail is a common problem faced nowadays by the e-mail users, which is an attempt to acquire sensitive information like password, credit cards details, etc. by sending malicious e-mail to the users. Classification of these types of e-mail is necessary to protect the e-mail users against harmful activities. This paper proposed to develop a classification model with the help of a new feature selection technique (FST) known as bucket-based feature selection technique (BBFST) in combination of C4.5. As the name suggested, this FST removes the feature one by one from original feature space of phishing e-mail data and puts into the three buckets based upon importance of the features as relevant feature, less relevant feature and irrelevant feature, and a new feature sub set is created. Classification technique C4.5 is then applied with data of new feature subsets and compared with existing FST. Results obtained reveal that BBFST is superior to those of existing FST with 99.008% accuracy with 12 features of phishing e-mail data.