Feature selection based supervised learning method for network intrusion detection

Abstract

Supervised learning is one of the data mining phenomena where a knowledge model is built for artificial intelligence. Learning from training samples has its advantages in predictive solutions. Such solution is essential for network intrusion detection problems. Networks of all kinds do have problem of intrusions as they are exposed to public communications in one way or other. Intrusions over a network are in the form of network flows that need to be analyzed. Manual observation of the flows and detecting intrusions is very time taking. Therefore it is essential to have an automated system for quickly detection of intrusions to safeguard network systems. There are many intrusion detection systems found in the literature. However, there is need for faster algorithm that makes sense in helping network administrators with accurate knowledge presented. Towards this end we proposed a framework with a feature subset selection mechanism to speed up detection process and improve accuracy of the same. The feature subset selection algorithm and Support Vector Machine (SVM) work together in order to have a faster detection system. Benchmark datasets like KDD and NSL-KDD are used for experiments. The empirical results showed that the proposed SVM-FSS framework shows better performance over the state of the art framework.