Abstract
Command, Control, Communications, Computers, and Intelligence (C4I) systems have been used in different organizations so the security of such systems is very essential. One of the solutions is a best intrusion detection system because the detection is the first step to prevent such intrusions. Several Intrusion Detection Systems (IDSs) with different classification techniques are available. However, these systems suffer from such issues that lead to low of detection rate (DR) and high of false alarm (FA). These issues can be happened due to the select inappropriate technique to classify data into normal or abnormal pattern. The effectiveness of IDSs depends on mainly on the optimal classifier architecture to detect intrusion with higher level of accuracy. In this paper, architecture of Modular Neural Network (MNN) to classify data into normal or abnormal pattern is proposed, which increased detection rate and decreased false alarm. The NSL-KDD dataset is used for the experiments. Moreover, Principle Component Analysis (PCA) is applied in order to transform features. The performance of the proposed approach was analyzed and then compared with other related work, and the results show that, the proposed approach outperforms the other approaches by increases the DR and decreases FA.