Robust Feature Selection Technique for Intrusion Detection System
Pullagura Indira Priyadarsini, Manikonda Srinivasa Sesha Sai, Akula Suneetha, M. Santhi · International Journal of Control and Automation · 2018
Intrusion Detection System (IDS) is one of the vital steps in network defense mechanism since the intrusive behavior of network data is a bit bewildering.Besides enormousness of the data, noisy or irrelevant features cannot be ruled out using classic data mining techniques.Feature selection is of substantial prominence in pattern recognition which is potent in augmenting learning efficacy, increasing generalization effect, and achieving data visualization.Feature selection process with the Fuzzy system is strong and effective.A novel method is proposed by exploiting the fuzzy system for feature selection.It can be implemented for reducing the computational complexity and improving the classification accuracy in IDSs.In this paper, a Robust Feature selection (RFS) algorithm is given which was an ensemble of three filtering methods: Euclidean distance, chi-square distance, and correlation coefficient.The ensemble of the three filtering methods is done using fuzzy aggregation operator.The tests made on KDD cup 99 data set, ensured good results and generated a greater proportion of recall and precision when compared to other feature selection methods.The average area under the curves (AUCs) will be given as 0.889 which can be a pretty good fit for the proposed algorithm.