A Layered Approach for Pattern Recognition in Large Dataset Using Meta modeling with Classification Techniques

Security of computers and networks that connect them is increasingly becoming of great significant. Intrusion detection is the act of detecting actions that attempts to compromise the clandestinely, credibility or availability of a network resource. It is an important attribute of defensive measure protecting computer system and network traffic from abuses. Here, we are focusing on two important aspects of intrusion detection; one is accuracy and other is performance. In the paper it is demonstrated that high attack detection accuracy can be achieved by using meta-modeling techniques in combination with classification techniques and high performance is attained by the layered approach. To test the results we have used NSL-KDD datasets; and also applied PCA for feature reduction that results in a significant improvement on learning algorithms. In this paper, we have designed and evaluated the combinational models for intrusion detection mechanism, and later we compared those models with each other and tried to find which is more accurate and appropriate to detect intrusion. We have applied meta-modeling because it gives better classification performance than any individual classifier. Our research has shown that the combination of meta-modeling algorithms with SVM gives better overall accuracy than any other combinational model. Index Terms—Meta-modeling techniques, Classification techniques, Layered approach, PCA.