Designing a Prediction Model for Phishing Website detection using an Ensemble of Supervised Learning Techniques

Phishing remains a critical issue in cybersecurity, particularly with the increasing occurrence of phishing websites used in data breaches. These malicious sites often trick users into divulging sensitive personal information, leading to identity theft and financial losses. This study presents the design and implementation of a phishing website prediction model using an ensemble of supervised learning techniques. The model combines four different supervised machine learning algorithms—Support Vector Machine (SVM), Multilayer Perceptron (MLP), K-Nearest Neighbor (KNN), and eXtreme Gradient Boosting (XGBoost)—in a stacked ensemble framework to enhance prediction accuracy. Additionally, two unsupervised algorithms, Hierarchical Clustering and K-Means Clustering, were applied independently for comparative analysis. The dataset, sourced from the Mendeley platform, was meticulously preprocessed to remove duplicates, clean the data, and encode features. It was then split, with 80% of the data used for training and the remaining 20% for validation and testing. The individual models achieved classification accuracies of 93.66% (SVM), 93.88% (MLP), 93.13% (KNN), 93.39% (XGBoost), 71.43% (Hierarchical Clustering), and 71.78% (K-Means Clustering). The stacked ensemble model surpassed these, achieving an accuracy of 94.05%. The findings underscore the potential of ensemble learning techniques in effectively predicting and mitigating phishing website attacks.