A Feature Selection Comparative Study for Web Phishing Datasets

Abstract

Among the plethora of cybercrime techniques employed by criminals, Phishing is by far the most extensively implemented technique. Phishing attacks are performed with the motive of monetary gains or theft of sensitive or intellectual data leading to major losses to both organizations and individuals. In this paper, we talk about the detection of Web Phishing attacks using Machine Learning. A comparative study is made between different Machine Learning Algorithms (that can be used for binary classification) and Feature Selection Techniques when applied to Phishing datasets. The goal of this experiment is to obtain similar/comparable accuracies while achieving a significant reduction in the number of features. We use the F1_Score and Time of execution as metrics to evaluate improvements in the overall system. Results are then rendered in the form of tables and graphs that demonstrate the same.