Fuzzy-Based Extreme Gradient Boosting for 5G Network Security

Abstract

In this work, an Intrusion Detection System (IDS) is developed for the SDN by including well-known machine learners and tree-based algorithms. The entire process is done as 1) Data preprocessing, 2) Feature extraction, 3) Dimensionality reduction, 4) Classification. The well-known NSL-KDD dataset is considered for this research. The Random Forest classifier aids in the feature extraction, and the Principal Component Analysis (PCA) is used for the dimensionality reduction. A Fuzzy-XGBooster classifier is proposed in this work, and it handles the classification part, and detects the normal and the anomaly class. The implementation part is done on the NSL-KDD dataset, and the performance is evaluated on several metrics. The proposed Fuzzy-XGBoost classifier achieved higher performance rate with the values of 0.999246 for accuracy, 0.998859 for precision, 0.998716 for recall, 0.998788 for F1 measure, 0.999485 for specificity, and 0.000515 for false alarm rate, respectively. Again, for the metrics Matthews Correlation Coefficient (MCC), Negative Predictive Value (NPV), False Positive Rate (FPR), False Negative Rate (FNR), Positive Predictive Value (PPV), Root Mean Square Error (RMSE), and Mean Absolute Error (MAE) the proposed Fuzzy-XGBoost classifier has achieved suitable values of 0.9981, 0.9993, 0.000602, 0.001317, 0.9988, 0.029031, and 0.000843 respectively.