Combining Machine Learning Algorithms to Detect Phishing URLs: A Stacking Approach

With the growth of digital technologies and the increasing use of the internet, phishing attacks have become one of the most significant security threats. These attacks aim to gain access to sensitive user information and cause financial and security damages. Accurately and promptly detecting these attacks has become a major challenge due to their increasing complexity. This article examines the use of machine learning models for detecting phishing URLs. A review of previous research shows that basic algorithms can be effective in detecting these attacks, but they have limitations, such as low capability to handle complex data. To improve accuracy and performance, hybrid algorithms have been proposed that combine multiple models to enhance detection accuracy. The proposed model in this study is designed using a hybrid approach to address the weaknesses of the basic algorithms and improve detection accuracy.  This hybrid model utilizes Extreme Gradient Boosting and Random Forests as base models, with Logistic Regression as the final model. The study employs a dataset of labeled phishing and legitimate URLs with features extracted from URL structure and behavior to train and evaluate the model. Experimental results indicate that the proposed hybrid model achieves higher accuracy and precision compared to using the base models alone. The application of this model can effectively contribute to enhancing cybersecurity and preventing phishing attacks.