DEVELOPMENT OF A PHISHING LINK DETECTION SYSTEM USING HYBRID MACHINE LEARNING ALGORITHMS

ABSTRACT Driven by the escalating threat of phishing attacks and persistent gaps in user education, this research presents the development of a hybrid web-and-desktop phishing detection system powered by machine learning. The proposed solution delivers real-time analysis of suspicious URLs through single-link scanning, batch processing, and optional sandbox previews, and reinforces user awareness with an interactive quiz module. Implemented as a responsive web application and packaged desktop client via Electron.js, the system is built using HTML, Tailwind CSS, JavaScript (Chart.js), Python, Flask, JWT authentication, SQLite3 (via SQLAlchemy), Pandas, Scikit-Learn, and Joblib. It persistently logs scan histories, visualizes user metrics on a unified dashboard, and provides clear, model-driven explanations for flagged links. By combining instantaneous ML-driven detection, safe sandboxing, contextual insights, and hands-on training, the proposed system empowers users to recognize and avoid phishing threats across environments, strengthening everyday cybersecurity.