Anomaly based intrusion detection using filter based feature selection on KDD-CUP 99

DoS, probing, phishing, website defacements etc. are the major problems being faced by the network users these days. It has led to exposing of the network resources to the attackers. Hence, it is really important to identify whether a connection is secured or not. This paper proposes an algorithm that discriminates the anomalous from normal connection. Solving DoS problem facilitates the user in faster access of network services. It thus strengthens our present day Intrusion Detection Systems (IDS) by quickly classifying the attack. Datasets like KDD-CUP 99, Kyoto 2006+, iCTF, DEFCON, NSL-KDD are available for experiments. However, most of them are not license-free and some are unlabeled. Here we have carried out the experiments on KDD-CUP 99 dataset. The major contribution of the paper is to determine the most appropriate feature selection algorithm to select the relevant features from 41 attributes of the connection vector. We have used statistical techniques to classify the instances of KDD-CUP 99. The results are analysed for various models by comparing their accuracy, detection rate, FAR etc. Empirical evaluation validates the superiority of the proposed algorithm against other state-of-the-art methods.