Phishing Website Detection Using Improved Multilayered Convolutional Neural Networks

Abstract

The internet has become an essential part of many fields: Communication, entertainment, commerce, industrial production, agriculture, etc. Unfortunately, online users are vulnerable to various attacks; this could lead to financial damages and loss of personal information. Phishing is seen as an internet threat and a cybercrime where anyone can capture personal information and data by posing as a reliable source. Data may include passwords to access confidential private or industrial repositories, emails, banks, financial information, etc. The prediction task is one of the crucial aspects of modern security systems, including anti-virus, firewall, and anti-spyware software. Currently, there is no availability of a single technique that can effectively detect every phishing attack. This study proposes a novel intelligent approach, phishing Prediction, using machine learning and deep learning to accurately predict phishing websites. We apply a pre-processing pipeline and develop the model using four machine learning models namely decision tree, Naive Bayes, support vector machine random forest, and Convolutional Neural Network (CNN) as a deep learning model. The UCI machine learning repository dataset comprised 11,055 websites, including lists of 4898 phishing and 6157 legitimate websites. The multilayered CNN has achieved the highest accuracy of 99.1% among all the listed algorithms, showcasing a precision of 97, a recall of 96%, and an F1-score of 96%.