Intrusion detection system using decision tree-based attribute weighted AODE

Abstract: The number and severity of the network attacks have increased in past few years. So for securing the network from different network attacks, intrusion detection system (IDS) plays a key role. Detection of the intrusive activities by using resource intensive intelligent algorithms has been possible because of advancements in computing performance in terms of processing power and storage. In this paper, an efficient data mining algorithm called Naïve Bayes for anomaly based network intrusion detection has been implemented. However, Naïve Bayes assumes attributes are independent of each other, the same may affect the accuracy of the system. To solve this attribute independence issue and to increase accuracy another data mining algorithm named Averaged One Dependence Estimator i. e. AODE is implemented. And to improve the performance of AODE further another algorithm has been proposed named Decision Tree-based Attribute Weighted AODE (DTWAODE). In DTWAODE Decision Tree is used & weight is assigned to each attribute. The weight is set according to its depth in the decision tree building on the training samples. The training sample used is NSL KDD-99 data set. The performance of Naïve Bayes, AODE & DTWAODE is studied and analyzed on the NSL KDD-99 intrusion benchmark data set and the accuracy is calculated.

Keywords: IDS, NSL, KDD, AODE, DTWAODE.