A Multi-Stage Intrusion Detection Approach for Network Security

Nowadays, the massive increment in applications running on a computer and excessive in network services forces to take convenient security policies into an account. Many methods of intrusion detection proposed to provide security in a computer system and network using data mining methods. These methods comprise of the outlier, unsupervised and supervised methods. As we know, each data mining method is not able to find different types of attacks. So, for removing this vulnerability, we are using Multi-Stage Intrusion Detection Method that containing outlier, unsupervised and supervised detection approaches for improving the performance and detection accuracy by reducing the false alarms for detection of known and unknown attacks. We have used NSL-KDD, KDD Corrected and GureKDD dataset in our experiment. We have compared our proposed outlier method GBBK with GBBK method and our method gives the same result with the less time complexity. The Unsupervised classification algorithm k − point performing the unnecessary comparison of objects iteratively by reducing number of attributes every time up to the threshold that is improved and named as k− point. Empirically, the proposed scheme compared with existing methods, and the results shows that the proposed method outperform in term of time complexity and detection accuracy.