Anomaly Based Network Intrusion Detection Using Ensemble Classifiers

Abstract

Network Intrusion Detection System is extensively utilized for protection and reducing the damages of information system. It protects threats and vulnerabilities in computer network. Due to the rapid growth of computer network communications, network intrusion is significantly increased and the intrusion detection is considered as a major issue in nowadays. For secure the communication, it is necessary to identify network attacks or malicious activities in network environment. To detect the intrusion in the network various methods have been proposed in past and effective analysis based on ensemble machine learning methods have been done to detect any types of anomalous events occurred in the flow of network traffic. In the learning process, ensemble methods are known to perform well. Investigating the best ensemble approach is crucial for creating an effective network intrusion detection system. In this paper, we used Bagged Naïve Bayes-Decision Tree (BNBDT) and Random Forest ensemble learning techniques and also used four base classification algorithms which are Naïve Bayes, KNN, Decision Tree and Logistic Regression on NSL-KDD network attack dataset for detecting the anomaly in network traffic and compared the performance of ensemble classifiers with the base classifiers. The proposed ensemble method provides better accuracy and relatively low false alarms rate than the other base classifiers.