Enhanced Multi-Threat Detection Classification using Light GBM, Cat Boost and Random Forest Models

Effectively identifying and categorizing dangerous URLs has become essential for cyber security defense due to the rise in malicious online activity. In order to classify multithreat URLs as safe, phishing, or malware-related, this study suggests a machine learning-based method. We created a classifier with a strong 96% accuracy rate using ensemble learning models, namely Light GBM, Cat Boost, and Random Forest. Each model's capacity to handle unbalanced data, adjust to various threat types, and provide reliable detection accuracy was assessed. Our method highlights the advantages of ensemble approaches in managing intricate threat classifications by optimizing feature selection and hyper parameters. The viability of implementing these effective models in actual cyber security systems to successfully counter URL-based threats is demonstrated by this study.