PaciPhish: Intelligent and Interpretable Phishing URL Detection Framework

Phishing is one of the most widespread forms of malicious web-based attacks. One of the most prominent applications is emails containing malicious phishing URLs. In this paper, we introduced PaciPhish, a machine learning (ML) framework for performing Phishing URL Detection (PUD). The methodology utilizes an ML algorithm to detect between "Phishing" and "Not Phishing" URLs. In addition, PaciPhish utilizes Explainable Artificial Intelligence (XAI) to help unearth the dataset features that influence the categorization between "Phishing" and "Not Phishing" URLs. Results indicate efficient performance for PaciPhish in PUD. The highest performance achieved was using LR with A, P, R, and F scores of 0.99, 0.99, 0.99, and 0.99, respectively. The best dataset size was when E<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">good</inf> = E<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">bad</inf> = 300 with A, P , R, and F scores of 0.99, 0.99, 0.99, and 0.99, respectively. Using XAI analysis allowed us to understand the reasoning behind the ML model classifying a sample as "Phishing" versus "Non Phishing". Finally, we observed that the PaciPhish framework also performed better than the existing state-of-the-art frameworks for PUD.