International Journal of Modern Engineering Research (IJMER)
With the rapid growth of internet communication and availability of techniques to intrude the network, network security has become indispensable. In this paper, the authors propose a multilevel classification technique for intrusion detection that uses intelligent agents and a combination of decision tree classifier and enhanced multiclass support vector machine algorithm for the implementation of an effective intrusion detection system in order to provide security to wireless sensor networks. The main advantage of this approach is that the system can be trained with unlabeled data and is capable of detecting previously "Unseen" attacks using agents. Verification tests have been carried out by using the KDD cup'99 data set.