Enhancing Network Intrusion Detection Using Effective Stacking of Ensemble Classifiers With Multi-Pronged Feature Selection Technique

Information security depends on Network Intrusion Detection (NID), which properly identifies network threats. This work explores simulating a NID system by stacking ensemble classifiers with various feature selection methods. We used the NSL-KDD dataset for the investigation. The binary classification performance integrates all assaults into one class and treats another target variable as usual. Random Forest (RF), Bagging Classifier (BC), and Extra Tree Ensemble (ETE) using Logistics Regression (LR) as a meta estimator. Chi2-square (Chi2) and Recursive Feature Elimination (RFE) are also used for feature selection. The experimental findings suggest that the proposed technique is ideal for developing a classification model with high accuracy and outperforms standard machine learning classification methods. The Chi 2 attribute selection strategy with modest characteristics has the greatest accuracy of 99.84 %. A careful comparison with prior research shows that our technique produces state-of-the-art performance with reduced computing costs.