Authors: Udith A, Harsha H S, Jayanth B R, Prathibhavani P M
Abstract: – In today’s fast-changing digital environment, phishing attacks are a major cybersecurity concern. These attacks use deceptive messages to trick users into revealing sensitive information or installing harmful software. Historically, such attacks have involved widespread spam campaigns that target many users with malicious URLs or files designed to bypass standard security measures. To address the increasing so- phistication of these threats, this research introduces an intelligent, real-time framework for detecting phish- ing URLs using machine learning. A gradient boosting classifier was specifically chosen to systematically examine and distinguish phishing URLs from legitimate ones. The approach relies on a broad suite of lexical, structural, and host-based feature extraction. The classifier outperforms traditional methods—including support vector machines, decision trees, random forests, and neural networks—demonstrating both higher accuracy and lower false positive rates. These results validate the system’s capacity for timely and effective phishing detection. The work underscores the promise of sophisticated machine learning methods for enhancing digital trust and reinforcing cyber defense architectures.
DOI: http://doi.org/10.5281/zenodo.16408435
