CYBERSECURITY ANALYSIS OF PHISHING ATTACKS USING MACHINE LEARNING-BASED DETECTION AND PREVENTION TECHNIQUES

Abstract

A new generation of cyber-attacks have evolved into an ever-present challenge for all users. These types of attacks can take advantage of either a user's lack of knowledge or the weaknesses within Web applications. Due to the rapidly growing number of users accessing digital services (such as banking, shopping and cloud services) via their computer devices, there is now greater potential for these types of attacks to succeed. However, since most traditional detection methods rely heavily on blacklists and heuristics, they may be ineffective when faced with newer forms of sophisticated phishing techniques. This paper proposes a machine-learning framework for identifying and preventing phishing attacks by extracting discriminative features from URLs and Web content. Supervised learning algorithms were also tested including Decision Trees, Random Forests, Support Vector Machines (SVM), and Logistic Regression. Evaluation criteria used to compare the performance of each model included common measures of evaluation accuracy, precision, recall, and F1 score. Results demonstrated that the best performing algorithm was the ensemble method Random Forest, which provided better classification results than other methods. Additionally, it produced fewer false positive classifications. Finally, the authors proposed several preventative methods including real-time filtering mechanisms and browser level security enhancements to further reduce risk associated with phishing attacks.

Keywords- Cyber Security, Phishing Attacks, Machine Learning, Phishing Detection & Prevention