PhishGuard: A Hybrid Ensemble Framework for Detecting Malicious URLs
Joel Mathew, Lekshmi S Nair, Hema P Menon, P Asmi · 2025
Phishing attacks remain a persistent and evolving threat to internet users, where malicious URLs deceive users into sharing sensitive information. This article proposes an intelligent URL detection system that accurately identifies phishing links using handcrafted and textual features. The system utilizes machine learning (ML) and deep learning (DL) models for enhanced phishing website detection, employing CNN, CNN-LSTM, Random Forest, and XGBoost to ensure robustness and accuracy. A true hybrid prediction function is implemented that invokes all models, collects their predictions, and determines the final verdict using majority voting. This ensemble approach improves prediction reliability and reduces false positives. The paper also features a user-friendly web interface, powered by Flask, which provides real-time analysis and model-specific transparency, enhancing user trust.