International Journal of Scientific Research and Engineering Development

Phishing attacks remain one of the most pervasive and economically devastating threats in the cybersecurity landscape, accounting for over 36% of all reported data breaches globally in 2024. This paper presents a novel real-time phishing detection and prevention system leveraging a hybrid ensemble of classical machine learning algorithms, deep learning architectures, and transformer-based natural language processing models. The proposed system integrates multi-modal feature extraction pipelines — encompassing URL structural analysis, domain reputation intelligence, HTML/JavaScript content inspection, and certificate transparency logs — to construct a rich, 287-dimensional feature vector for each candidate URL or email. An adaptive stacking ensemble combining XGBoost, LightGBM, BiLSTM, and a fine-tuned BERT variant achieves an accuracy of 99.6%, precision of 99.4%, and recall of 99.7% on a benchmark dataset of 2.8 million URLs. Real-time classification latency averages 89 milliseconds per sample, satisfying production-grade deployment requirements. We further introduce a continuously updating threat-intelligence feedback loop that allows the model to adapt to zeroday phishing campaigns without full retraining. Experimental results on three independent validation datasets outperform existing state-of-the-art baselines by 1.5–3.2 percentage points in F1-score.

Mr.M.Mohanasundharam, Yogaraj.G, Saravanapriyan.B, Sathishkumar.V, Srisanjay.R,"Real-Time AI/ML-Based Phishing Detection and Prevention System" International Journal of Scientific Research and Engineering Development, V9(2): Page(2094-2098) Mar-Apr 2026. ISSN: 2581-7175. www.ijsred.com. Published by Scientific and Academic Research Publishing.