An Ensemble Model for Detecting Phishing Attack with Proposed Remove-Replace Feature Selection Technique

Abstract

Phishing E-mail is an unwanted E-mail, sent by the phishers to get the sensitive information like password, account number and credit card details from users. This research work emphasizes on constructing an ensemble Machine Learning (ML) based model to detect phishing E-mail with Remove-Replace Feature Selection Technique (RRFST). RRFST reduces features from original feature space by selecting a feature in random manner and removes it if accuracy is being unchanged or increased otherwise feature is being replaced to its original feature space. Classifiers were developed using two Decision Tree (DT) algorithms i.e. C4.5 and Classification and Regression Tree (CART) along with ensemble of these two as an efficient classification model with reduced feature subset obtained through RRFST. Empirical results indicate that proposed FST produces remarkable performance of 99.27% accuracy using ensemble of C4.5 and CART with only 11 features. Research outcome of proposed FST is also investigated with two existing FSTs: Info Gain (IG) and Gain Ratio (GR).