Phishing Detection System through Hybrid Machine learning Model

Abstract

Phishing attacks, introduced around 1996, have become a major cybersecurity threat by using fake emails and websites to steal sensitive information. Despite many detection techniques, a complete solution is still lacking. This study applies machine learning to detect phishing by analyzing URL features from a dataset of over 11,000 websites. The URLs are converted into vector form for model training and testing. Multiple algorithms such as Decision Tree (DT), Logistic Regression (LR), Random Forest (RF), Naive Bayes (NB), Gradient Boosting (GBM), KNN, and SVC are evaluated. A hybrid model called LSD, combining LR, SVC, and DT with voting mechanisms, is proposed. Feature selection using the Canopy method and optimization through cross-validation and Grid Search enhance performance. Results show that the LSD model achieves higher accuracy and efficiency compared to other models in detecting phishing URLs