Intrusion Detection Approaches for Internet of Things

Abstract

The Internet of Things (IoT) links "everything" to the internet and enables "things" to communicate with one another over wired or wireless networks. The number of IoT applications has significantly increased, such as smart cities, smart homes, wearable's, and healthcare. Security becomes more important as the number of devices connected to the IoT increases due to the types of devices, the volume of data that is sent over the network, the nature of the structure, and the different communication methods (primarily wireless). This fundamental nature of the IoT architecture intensifies the number of attack targets that may affect the sustainable growth of the IoT. Hence, security issues become a critical factor that must be addressed. Therefore, it became necessary to develop an attack detection system to keep pace with the current development of the IoT, as it deals with sensitive information that must be protected and secured. In this thesis, a deep learning approach based on convolutional neural networks is proposed to perform real-time detection of attack behaviors in IoT systems. The UNSW-NB15 dataset was used to train and test the proposed approach. The approach uses binary classification to distinguish attack and normal patterns. The proposed intrusion detection approach passes through two stages. The first stage: After loading the dataset, pre-processing is performed to obtain more accurate results. The second stage is classification by CNN classifier where the experimental result shows the efficiency of the presented approach concerning precision, recall, and f-measure as the detection precision reached 100%. The result of experiments is highly efficient for intrusion detection to distinguishing between normal and attack behaviors that provides a research approach