A Comparative Study of Ensemble Learning Techniques and Classification Models to Identify Phishing Websites

The advent of the internet, as we all know, has brought about a significant change in human interaction and business operations around the world; yet, this evolution has also been marked by security issues, including phishing attacks that represent one of the biggest problems to internet users, leading to financial loss and identity theft. The ability of Machine learning and ensemble learning models to process large datasets and complex relationships, and to learn from data have made it easier to detect phishing websites, which have become one of the major problems in modern-day security findings. In this study, a comprehensive analysis of various ensemble techniques is carried out, particularly focusing on algorithms like Random Forest, Gradient Boosting, and AdaBoost, in addition to traditional classification techniques like Logistic Regression, Decision Trees, and Support Vector Machines (SVM). In order to evaluate the effectiveness of these machine learning and ensemble models, the benchmarks dataset having phishing and normal site samples, the study assesses the performance of the mentioned models using distinct evaluation metrics, including accuracy, precision, recall, F1-score, and AUC-ROC. The study focuses its attention on the performance of the Random Forest and Gradient Boosting ensemble models compared to their single classifier counterparts. The findings revealed that ensemble techniques have a better performance in terms of true positive rate, false positive rate, and overall performance. Consequently, the research reinforces that these ensemble learning methods possess the capability of providing strength, flexibility, and efficiency under practical conditions of application. However, there are still some areas for improvement in developing and applying more advanced algorithms.