Intrusion Detection Neural Network Model Based on Interval Type-2 Fuzzy C-Means Clustering

Intrusion Detection System (IDS) is a critical research field in the age of internet, therefore a significant number of techniques have been employed for IDS. However, because of the inherent drawbacks of IDS datasets, these techniques are not very successful in identifying all types of intrusion. In this paper, we propose an intrusion detection neural network model based on interval type-2 fuzzy c-means clustering (IDNN-IT2FCM) to help IDS improve detection rates. Firstly, we utilize interval type-2 fuzzy c-means clustering (IT2FCM) to cluster the training set into different training subsets, which makes intrusion detection network learn subset more quickly, robustly, and precisely. Secondly, we design two criteria to decide the cluster belongingness of low-frequent attack samples in the training set and pre-classify the samples of the testing set. Finally, all groups of the testing set are classified by the neural network. Compared with other classification approaches, the proposed method can obtain the satisfying results on NSL-KDD dataset.