DESIGN AND IMPLEMENTATION OF A CLOUD-BASED PHISHING DETECTION MODEL

ABSTRACT Phishing attacks remain a critical cybersecurity challenge, leveraging deceptive URLs to compromise sensitive data. Traditional detection systems, reliant on static blacklists and heuristic rules, often fail to address zero-day phishing threats. This project presents a scalable cloud-based phishing detection model that combines machine learning with real-time analysis to identify malicious URLs dynamically. The proposed solution employs a feature-based classifier (e.g., Xgboost) trained on lexical, host-based, and content-based features (e.g., URL length, domain reputation, SSL certificates). Hosted on a cloud platform (AWS/GCP), the model processes user-submitted URLs via a REST API, ensuring low-latency responses. The XGBoost model was evaluated on the test set (3,162 samples) and achieved 98.6% accuracy, 98.5% precision, 98.8% recall, and a 98.6% F1-score on the test set (3,162 samples), surpassing the literature benchmark of 90% (Sahingoz et al., 2019) and meeting project targets.