Performance of Ensemble Learning Techniques for Network-Based Intrusion Detection System (NIDS): A Comparative Study

Network security has become a major concern due to the increasing reliance on digital technologies and interconnected networks. Network-based Intrusion Detection System (NIDS) enhances cyber security by monitoring network traffic, swiftly detecting and preventing intrusions, safeguarding digital assets, and ensuring data integrity. Many researchers have used Machine Learning (ML) and Ensemble Machine Learning (EML) techniques to design NIDS. However, in these research articles, an exhaustive study is not shown to find out the best ML algorithms for anomaly-based NIDS where different types of attack data (TCP-SYN Flood, Port Scan, Flow Table Overflow, Black Hole, and Traffic Diversion) pass through the networks. Therefore, this research article presents a comparative analysis of the performance of standard and ensemble machine learning classifiers to find the best ML/EML algorithm for NIDS. Different types of EML classifiers, such as Categorical Boosting (CatBoost), Extreme Gradient Boosting (XGB), Light Gradient Boosting Machine (LightGBM), Random Forest (RF), and Extra Trees (ET), have been applied to the multiclass UNR-IDD 2023 dataset with standard ML classifiers such as Decision Tree (DT), K-Nearest Neighbors (KNN), Naive Bayes (NB), Linear Discriminant Analysis (LDA), and Support Vector Machine (SVM). The results are compared among all classifiers according to accuracy and F1-score of confusion matrix. The performance of this work has been assessed by splitting and 10-fold cross-validation. The experimental results show that DT shows 93.89% accuracy, which is the highest among all standard ML classifiers, whereas the XGB ensemble classifier achieves the best accuracy of 96.93%. This analysis reflects that the EML classifiers are more efficient and robust for NIDS.