Tree-based Ensemble Machine Learning for Phishing Website Detection

Abstract

Phishing remains a prevalent and perilous cyber threat in the digital age, exploiting human vulnerabilities to deceive individuals into disclosing sensitive information. This paper presents a method to achieve high accuracy in phishing website detection using Tree-based Ensemble Machine Learning techniques. Through rigorous experimentation and evaluation, we identified RandomForest and ExtraTrees as the top-performing models, achieving accuracy, precision, recall, and F1 scores all exceeding 98%. Additionally, our study highlights the significance of feature selection techniques in enhancing model performance, with thresholding methods proving effective in retaining relevant features for classification. By addressing imbalanced datasets and optimizing hyperparameters, our models demonstrate robust detection capabilities against phishing attacks. These findings contribute to the advancement of cybersecurity measures and underscore the potential of ensemble machine learning in combatting online threats, ultimately enhancing internet user security.