International Journal of Scientific Research and Engineering Development

Phishing detection is crucial for protecting users from cyber threats, especially as online transactions and digital communications continue to grow rapidly. Fraudulent websites are designed to mimic legitimate platforms, making it difficult for users to distinguish between genuine and malicious URLs. Machine learning techniques, particularly ensemble learning models such as Random Forest and XGBoost (Extreme Gradient Boosting), have demonstrated significant potential in improving phishing detection accuracy. A comparative analysis of cost-sensitive Random Forest and cost-sensitive XGBoost is conducted to evaluate their effectiveness in detecting phishing URLs. URL-based features, including length, entropy, presence of suspicious keywords, number of special characters, domain characteristics, and HTTPS usage, are extracted and used as input for training and testing both models. Since phishing datasets are typically imbalanced, cost-sensitive learning is incorporated to assign higher misclassification penalties to phishing instances. Results indicate that while both models enhance detection performance compared to conventional approaches, XGBoost slightly outperforms Random Forest in terms of recall, F1-score, and ROC-AUC, whereas Random Forest provides stable and interpretable results. The findings contribute toward developing reliable and scalable phishing detection mechanisms for strengthening cybersecurity systems.

A.Beulah, Mrs.M.Saranya,"Comparative study of URL based Phishing Detection Using Machine Learning" International Journal of Scientific Research and Engineering Development, V9(2): Page(689-695) Mar-Apr 2026. ISSN: 2581-7175. www.ijsred.com. Published by Scientific and Academic Research Publishing.