A Unified Multi-Channel Phishing Detection Framework Using Machine Learning and Rule-Based Threat Analysis

Phishing attacks continue to evolve across multiple communication channels including emails, SMS messages, URLs, and websites. This paper presents a unified phishing detection framework that combines machine learning and rule-based threat analysis to identify phishing attempts across different attack vectors. The proposed system employs Calibrated Logistic Regression for SMS and email detection and Random Forest classification for URL and website analysis. Experimental evaluation demonstrated accuracies of 96.32% for SMS detection, 96.59% for email detection, and 98.67% for URL detection. The framework further incorporates threat scoring, explainable security recommendations, and unified risk assessment. Results indicate that the proposed approach provides accurate, interpretable, and practical phishing detection suitable for real-world deployment.