A Network Intrusion Detection System Based On Ensemble CVM Using Efficient Feature Selection Approach
T Divyasree, K. K. Sherly · Procedia Computer Science · 2018
The growth and development of web services is drastic these days. Many things which we did manually once can be done through the Internet today. But with these types of services, the concern for security also increases. Many efficient Intrusion Detection systems proposed by researchers are in practice; but still hackers manage to attack the systems successfully. This paper proposes an efficient intrusion detection system using Ensemble Core Vector Machine (CVM) approach. CVMs are algorithms which work on the basis of Minimum Enclosing Ball concept. It detects the attacks like: U2R attack, R2L attack, Probe attack and DoS attack. For each type of attack, a CVM classifier is modeled. KDD Cup’99 dataset is used for training and testing the classifiers. This approach uses Chi-square test for selecting the relevant features for each attack and a weighted function is applied to these features for the dimensionality reduction. The test results verify that this model achieves high efficiency in all the four attacks with less computation time compared to the existing approaches.