A Comprehensive Machine Learning Framework for Phishing URL Detection: Dataset Integration, Feature Extraction, and Model Evaluation

Phishing is a persistent cybersecurity issue that uses social engineering to trick people into disclosing private information, including financial information and login passwords. Conventional detection techniques, such as rule-based heuristics and blacklist-based strategies, have trouble identifying recently developed and evolving phishing assaults, which frequently results in high false-positive rates and decreased detection accuracy. To overcome these constraints, flexible detection system is required to enhance real-time phishing detection. This paper suggests a methodology for detecting phishing URLs that is based on machine learning (ML) and examines structural, lexical, and content-based characteristics such as HTTPS usage, domain composition, URL length, and special characters. The accuracy, precision, recall, and F1-score of several machine learning models, including Support Vector Machine (SVM), XGBoost, Random Forest, and Decision Trees, are used to assess their performance. Results from experiments show that the XGBoost model performs better than traditional techniques, with an F1-score of 0.5024 and a precision of 85.1%, making it a very dependable phishing detection tool. Our method greatly improves classification accuracy and adaptability by utilizing ensemble learning, allowing for more effective real-time phishing detection. This study offers a scalable and adaptable way to improve cybersecurity defenses and lower the dangers of phishing assaults. The results support the development of proactive and robust cybersecurity frameworks by validating the efficacy of machine learning in automated phishing detection.