A Layered Approach for Intrusion Detection Using Meta-modeling with Classification Techniques

Extensive growth of the internet and increasingavailability of tools and tricks for intruding and attackingnetworks, have prompted intrusion detection to become acritical component of network administration. It is animportant attribute of defensive measure protecting computersystem and network traffic from abuses. Here, we are focusingon two important aspects of intrusion detection; one is accuracyand other is performance. In the paper it is demonstrated thathigh attack detection accuracy can be achieved by usingmeta-modeling techniques in combination with classificationtechniques and high performance is attained by the layeredapproach. To test the results we have used NSL-KDD datasets;and also applied PCA for feature reduction that results in asignificant improvement on learning algorithms. In this paper,we have designed and evaluated the combinational models forintrusion detection mechanism, and later we compared thosemodels with each other and tried to find which is more accurateand appropriate to detect intrusion. We have appliedmeta-modeling because it gives better classificationperformance than any individual classifier. Our research hasshown that the combination of meta-modeling algorithms withSVM gives better overall accuracy than any othercombinational model.