International Journal of Advanced Research in Computer and Communication Engineering

Abstract: Phishing attacks pose a serious cyber security threat by imitating legitimate websites to steal sensitive data. This study presents a hybrid phishing detection system integrating Machine Learning (ML), Deep Learning (DL), and Ensemble Learning (EL). Feature selection techniques such as Information Gain, Gain Ratio, and Principle component Analysis (PCA) are applied to extract the most relevant indicators from a dataset of 11,055 URLs. ML classifiers (SVM, DT, KNN), EL models (RF, XGBoost, AdaBoost), and DL architectures (LSTM, GRU, CNN) are used. A hybrid model fuses LSTM and GRU outputs, processed by ensemble classifiers and finalized by a meta-classifier. The model captures both structural and sequential URL features, improving accuracy, reducing false positives, and enabling real-time adaptability. The framework can be deployed in email clients, browsers, or gateways to safeguard users from phishing threats. This scalable and intelligent system outperforms individual models and adapts to evolving phishing tactics, contributing to a more secure online ecosystem.

Keywords: Phishing, Machine Learning, Deep Learning, Ensemble Learning, Hybrid Model, Cyber security

Downloads: | DOI: 10.17148/IJARCCE.2025.14803