Adaptive Shields for Network Intrusion Detection via Gradient Boosting

Abstract

Network Intrusion Detection Systems (NIDS) are essential for protecting computer networks because they keep an eye on traffic and spot harmful activity. The HC-DTTSVM (Hierarchical Clustering Decision Tree Twin Support Vector Machine) technique and Gradient Boosting Decision Tree (GBDT) technology are combined in this study to present a novel way to improve NIDS accuracy. To improve the speed of the HC-DTTSVM method, GBDT is used as a feature extractor to automatically find important patterns in network traffic. The suggested approach uses a step-by-step training procedure that begins with GBDT and progresses to Artificial Neural Networks (ANNs) and Twin Support Vector Machines (TWSVMs). Using metrics like accuracy, precision, recall, false positive rate (FAR), F1-score, and G-mean, a thorough analysis shows that this combination strategy performs better than a number of deep learning-based methods. The outcomes demonstrate how well GBDT and HC-DTTSVM operate together to categorize network intrusions, indicating that this combination has the potential to be a reliable NIDS solution. In addition to demonstrating the suggested method's ability to increase detection accuracy, this study makes recommendations for future research directions, such as investigating sophisticated ensemble methods, utilizing deep learning models, improving feature engineering techniques, and validating the approach across a variety of datasets to guarantee generalization and robustness.