Mitigating Email Phishing Attacks using Convolutional Neural Networks

Abstract

Phishing detection has gained huge attention from both academia and industry. Damages and data breaches affecting private and governmental entities caused by email phishing attacks needed an immediate solution. The diversity of attack patterns and mediums associated with phishing attacks made the development of an optimal solution challenging. Also, attackers usually make legitimate looking content using legitimate wording or legitimate looking URLs and websites. Many of the existing phishing solutions requires manual feature extraction that requires expert domain knowledge and thoughtful selection of valuable features to be efficient. Additionally, most effective phishing solutions suffered from large computational costs. In this paper, we propose CNNPD, an email phishing detection framework based on Convolutional Neural Network (CNN). CNNPD marks incoming emails into phishing or benign. Testing the framework on an email dataset shows promising performance in terms of accuracy, precision, and recall when compared to similar approaches.