Machine learning based framework for network intrusion detection system using stacking ensemble technique

Abstract

Cybersecurity issues are increasing day by day, and it is becoming essential to address them aggressively. An efficient IDS system should be placed to identify abnormal behaviour by dynamically tracing the network traffic pattern. In this work, we proposed a framework for Network Intrusion Detection System using stacking ensemble technique of machine learning, which is testified on Random Forest Regressor and Extra Tree Classifier approaches for feature selections from the subjected dataset. The extensive experimentation has been done by applying 11 states of the art and hybrid machine learning algorithms to select the best performing algorithms. During the investigation, Random Forest, ID3 and XGBoost algorithms are found as best performers among different machine learning algorithms based on accuracy, precision, recall, F1-score and time to increase real-time attack detection performance. Three case studies have been carried out. Our results indicate that the proposed stacking ensemble-based framework of NIDS outperformed compared to the different state of art machine learning algorithms with average 0.99 prediction accuracy.