Phishing URL Detection: A Basic Machine Learning Approach

Abstract

Phishing is currently one of the most trending scam attacks in information transmission worldwide. Such occurrences can cause severe network disruption to corporate organizations, financial and/or educational institutions and even individual network subscribers. Phishing attacks would normally involve developing a malicious webpage mimicking a legitimate website. This lures unsuspecting users into giving out personal information, such as banking details, social security numbers, passwords, usernames, etc. A phishing fraudster sends an e-mail or text message to a target webpage that contains its Universal Resource Locator (URL). The result could be monetary loss, data breach, intellectual property theft, damaged reputation, and loss of customers. In this paper, we propose a simple phishing detection approach. Moreover, because the impostor is always evolving attacking techniques in a bid to evade detection, our method addresses this using a supervised machine learning system. From the website content, we extract the stochastic dynamical patterns and then use this to predict the authenticity of the website. The Extreme Gradient Boosting (XGBoost) algorithm is used to model the website features to obtain a better prediction result. The proposed technique can detect phishing websites with an accuracy of 86.6%.