Phishing URL Detection with Gradient Boosting Classifier

Abstract

Phishing attacks continue to pose a serious and persistent risk to internet security, jeopardizing people's and businesses' financial stability and privacy. The precise and timely identification of these malicious websites remains a crucial challenge in the field of cyber security. This paper provides a new strategy employing Gradient Boosting Classifiers (GBCs) to overcome this issue. Phishing websites are purposely built to mimic legitimate sites to fool users into providing sensitive information. As a result, many malicious sites exhibit tiny traits that identify them from real ones. Traditional rule-based systems typically struggle to adequately identify these complex variances. In contrast, Gradient Boosting provides an ensemble learning framework that utilises the combined strength of weak classifiers, resulting in a robust model capable of recognising these elusive characteristics. Our experimental findings show that our suggested strategy performs better than others in a number of parameters, including accuracy, precision, recall, and F1-score. Crucially, our model demonstrates exceptional resistance to aggressive tactics that are frequently employed to hide the actual purpose of phishing websites. This study represents a substantial step forward in the fight to improve internet security and shield consumers from the constant danger of phishing scams.