Zinah Sattar Jabbar Aboud, Rami Tawil, Mustafa Salam Kadhm
Abstract
Intrusion Detection (ID) in network communication and Wireless Sensor Networks (WSN) is a big challenge that has grown with the rapid development of these technologies. Various types of intrusion attacks may occur to the transferred data of such networks and various ID methods and algorithms have been proposed. One powerful tool used in this field is Machine Learning (ML), which has achieved satisfied detection results. However, these results with the available ID datasets can be further improved. This paper proposes an accurate approach for ID in the network and WSN using ML methods including chaotic map, Nomadic People Optimizer (NPO), and Support Vector Machine (SVM). The proposed approach has five main stages which are: data collection, pre-processing, feature selection, classification, and evaluation. An improved version of NPO based on chaotic map and Cauchy mutation called CNPO is proposed. The proposed scheme uses chaotic maps to initialize the population and Cauchy mutation for solution distribution. Besides, the proposed fitness function based on SVM is proposed. The CNPO is employed for the feature selection task. The proposed approach was evaluated in two datasets, NSL-KDD, and WSN-DS, with accuracy of 99.97% and 99.99, respectively.
أ.م.د مصطفى سلام كاظم | Mustafa Salam Kadhm | 15401