An investigation of classification algorithms for intrusion detection system — A quantitative approach

Nowadays security issues are growing in a tremendous rate. So it is expedient to have a mechanism to keep track of its security issues in the network or host. The Intrusion Detection System (IDS) has a critical part for supervising the networks. The false alarm rate and accuracy are the two important factors to be considered in the design of competent IDS. The role of classification algorithms is indispensable in the decision making of IDS. The redundant and irrelevant features of dataset also affects the performance of classifiers which in turn degrading the evaluation of anomaly detection. The proposed work focuses on the detailed study of different classifiers on two feature selection techniques using NSL-KDD dataset, where Random Forest on Principal Component Analysis (PCA) gives the accuracy rate of 99.52% and false alarm rate is 0.48%.