AN IMPLEMENTATION OF IDS IN A HYBRID APPROACH AND KDD CUP DATASET

Abstract

Now in these days due to rapidly increasing network applications the data and privacy security in network is a key challenge. In order to provide effective and trustable security for network, intrusion detection systems are helpful. The presented study is based on the IDS system design for network based anomaly detection. Thus this system requires an efficient and appropriate classifier by which the detection rate of intrusions using KDD CPU dataset can be improved. Due to study there is various kind of data mining based, classification and pattern detection techniques are available. These techniques are promising for detecting network traffic pattern more accurately. On the other hand recently developed the hybrid models are providing more accurate classification. Thus a hybrid intrusion system is presented in this proposed work. That provides a significant solution even when the overall learning patterns are not available in database. Therefore, three different data mining algorithm is employed with system. Proposed system consists of K-mean clustering algorithm for finding the relationship among data in order to filter data instances. The implementation of the proposed classification system is performed using MATLAB environment and performance of designed classifier is evaluated. The obtained results from the simulation demonstrate after filtering steps. On the other hand the classification accuracy is adoptable with low number of training cycles with less time and space complexity.