A Novel Architecture Based on Weight Freezing and Random Forest for Website Phishing Detection

Phishing is a severe cybersecurity threat that continues to cause significant economic losses and breaches of privacy globally. This paper presents a novel architecture for effective phishing website detection, integrating a unique weight freezing technique in the Random Forest algorithm with a feature pyramid scheme for feature selection. This innovative combination enables the model to focus on important features and discard irrelevant or redundant ones, thus improving detection accuracy while preventing overfitting. Using a dataset of 2456 instances with 30 attributes, the proposed model achieved an accuracy of 97.4%, an MSE of 0.05, and an F1 score of 96.2%, outperforming previous state-of-the-art methods. The promising results suggest the model’s significant potential in the cybersecurity field, providing a robust tool against the escalating threat of phishing.