Intrusion Detection System Using Bagging with Partial Decision TreeBase Classifier

Abstract

Abstract Intrusion Detection System has become an essential part of the computer network security. It is used to detect, identify and track the intruders in the computer network. Intrusion Detection Technology which provides highest classification accuracy and lowest false positive is required. Many researchers are involved to find out and propose Intrusion detection technology which provides the better classification accuracy and less training time. The traditional Intrusion Detection system exhibits low detection accuracy and high false alarm rate. Now a day,an Ensemble method of machine learning is widely used to implement intrusion detection system. By analyzing Ensemble method of machine learning and intrusion detection system in this paper, we make use of Bagging Ensemble method to implement Intrusion Detection system. The Partial Decision Tree is used as a base classifier due to its simplicity.The selections of relevant features are required to improve the accuracy of the classifier. The relevant features are selected based on their vitality for each type of attacks. The dimension of input feature space is reduced from 41 to 15 features using Genetic Algorithm.The proposed intrusion detection system is evaluated in terms of classification accuracy, true positives, false positive and model building time. It was observed that proposed system achieved the highest classification accuracy of 99.7166% using cross validation. It exhibits higher classification accuracy than all classifiers except C4.5 classifier on test dataset.The Intrusion Detection system is simple and accurate due to simplicity of Partial Decision Tree.