Research and Construction of Multi-model Fusionbased Malicious URL Detection Method

While the rapid development of the Internet brings convenience to people’s lives, it also breeds network security threats that use malicious URL (Uniform Resource Locator) as an entry point, such as phishing attacks and malware distribution. These attacks may lead to data leakage, system damage, and financial loss, therefore, effective detection and blocking of malicious URLs are crucial for network security. In this paper, we compare the performance of several mainstream supervised learning methods for malicious URL detection and construct a two-layer stacked integration model that combines random forests, extreme trees, and XGBoost as base models, and decision trees as metamodels. The experimental results show that the model in this study achieves 0.92 accuracy, precision, recall, and F1 score under cross-validation with an AUC value of 0.95, The experimental results show that the model performs well in terms of accuracy, balance, and robustness. This multi-model fusion method not only improves the detection accuracy but also enhances the system’s ability to process large-scale data, which provides strong support for building an efficient and reliable malicious URL detection system.