Analysis of Hybrid Intrusion Detection System Based on Data Mining Techniques

Abstract

Abstract-The rapid growth of network based activities makes computer security is a more crucial issue. Many security methods are developed and used, but they are unfit to detect novel intrusions. Therefore, we propose a hybrid intrusion detection framework based on data mining classification and clustering techniques. In the proposed hybrid framework, improves the detection rate by taking the advantages of misuse and anomaly detection. In case of misuse detection, intrusion patterns are built automatically from a training data by the use of the random forest classification method. Then comparing this pattern against network activities to detect intrusions. In case of anomaly detection, the network activities processed to several clusters using weighted k means technique to detect novel intrusions. The whole process is evaluated over KDD'99 dataset.