International Journal of Advanced Research in Computer and Communication Engineering

Abstract: Phishing, a deceptive cyber-attack technique aimed at extracting confidential information, has evolved into one of the most significant threats in modern cyberspace. Traditional detection systems relying on static rules, blacklists, or manual inspection fail to recognize rapidly evolving and AI-generated phishing attempts. This research presents an Artificial Intelligence (AI)-driven phishing attack detection and prevention framework that integrates Natural Language Processing (NLP) with Deep Learning (DL) to enhance real-time recognition accuracy. The proposed hybrid model combines Bidirectional Encoder Representations from Transformers (BERT) for contextual text understanding and a Convolutional Neural Network (CNN) for URL pattern analysis. Additionally, the integration of Explainable AI (XAI) techniques such as LIME and SHAP provides interpretability for each classification decision. Through experimentation on benchmark datasets like PhishTank and Kaggle, the system achieved an overall performance accuracy of 96.4%. The model exhibits strong adaptability, continuous learning capabilities, and superior resilience against zero-day phishing threats. This study contributes to a transparent, adaptive, and intelligent defense framework for the next generation of cybersecurity systems.

To address these limitations, this research proposes an AI-Powered Phishing Attack Detection and Prevention System that integrates Natural Language Processing (NLP) and Deep Learning (DL) for intelligent, adaptive, and explainable threat detection. The system employs BERT (Bidirectional Encoder Representations from Transformers) to analyze the semantic and contextual meaning of email or web content, while a Convolutional Neural Network (CNN) model examines the structural and lexical characteristics of URLs. By combining these two analytical perspectives, the model forms a hybrid AI engine that significantly enhances accuracy and resilience against zero-day phishing attacks.
A key innovation of this work lies in its Explainable AI (XAI) component, which utilizes tools such as LIME and SHAP to interpret the model’s decisions. This transparency allows users and cybersecurity analysts to understand the reasoning behind each detection result, thereby improving trust and system reliability. The system also integrates a real-time browser extension and interactive web dashboard that proactively prevents users from accessing malicious domains and provides analytical visualizations of phishing trends.

Keywords: Artificial Intelligence, Cybersecurity, Phishing Detection, Deep Learning, NLP, Explainable AI, BERT, CNN

Downloads: | DOI: 10.17148/IJARCCE.2025.141011