Recursive Feature Elimination Algorithm for Intrusion Detection Systems

Identification of the attacks on computer networks has significant importance in security research. In this paper, a set of machine learning algorithms are investigated for the design of intrusion detection systems. The detection performance of the proposed intrusion detection system is improved by the application of the recursive feature elimination method to rank the entire input features and generate various feature combinations based on their importance. The cross-validation procedure is also implemented for the machine learning techniques on the validation dataset using a larger number of feature combinations. The high classification performance of the proposed algorithm for the intrusion detection system implies a better capability of application in the practical environment of computer networks.