AI-Based Intelligent Framework for Enhancing Phishing Attack Detection

Abstract

Phishing has become one of the most widespread and quickly evolving cyber threats that capitalise on human and system-based vulnerabilities in emails, websites and SMS. Conventional defence controls, such as blacklists, rule-based filters, and signature matching, cannot identify advanced, obfuscated, and zero-day phishing attacks. Recent developments in Artificial Intelligence (AI) and classical machine learning have shown promise for scalable and interpretable phishing detection. The research paper proposes a lightweight and real-time phishing detection framework based on TF-IDF textual feature extraction and Naive Bayes and Random Forest classification. The model incorporates the discriminative value of terms in URLs, email messages, and SMS messages, and thus it can accurately separate legitimate and malicious communication. Evaluation of experimental phishing benchmark datasets demonstrates that high classification, inference speed and generalization with low computational cost can be achieved. The simplicity and openness of the framework render it applicable to academic settings, resource-constrained systems, and real-time deployment and it provides an efficient alternative to other more complicated deep learning models. The contribution of this work to the development of accessible, interpretable, and adaptable AI-based phishing detection systems for contemporary cybersecurity issues is significant.