Feature selection for phishing detection: a review of research

Web services motivate phishers to evolve more deceptive websites as their never-ending threats to users. This intricate challenge enforces researchers to develop more proficient phishing detection approaches that incorporate hybrid features, machine learning classifiers, and feature selection methods. However, these detection approaches remain incompetent in classification performance over the vast web. This is attributed to the limited selection of the best features from the massive number of hybrid ones, and to the variant outcomes of applied feature selection methods in the realistic condition. In this topic, this paper surveys prominent researches, highlights their limitations, and emphasises on how they could be improved to escalate detection performance. This survey restates additional peculiarities to promote certain facets of the current research trend with the hope to help researchers on how to develop detection approaches and obtain the best quality outcomes of feature selection.