An Intrusion Detection System for IoT Using Deep Learning and Optimization Techniques | Chapter 2 | Contemporary Perspective on Science, Technology and Research Vol. 4

Internet of Things (IoT) has got more attention in the research field of computer science. The extreme increase in the IoT appliances across various factors, such as smart industries and health care appears with extensive security hazards. It is not only restricted to the attacks on confidentiality but also broadens to the attacks on performance and availability of the network. Hence, an intrusion detection mechanism is mandatory for identifying the attacks on IoT to offer effectual protection and security. Even though various intrusion detection methods are developed, achieving higher classification performance still results a challenging task. Therefore, an effective intrusion detection method is developed using the proposed Competitive Swarm Henry Optimization (CSHO)-based Deep Maxout network to find intruders in the IoT environment.

The process of detection strategy is carried out with the information captured by nodes distributed in network. Routing plays an essential role in transferring data from IoT devices to the base station (BS) to accomplish the task of intrusion discovery. It is the common approach used for increasing the energy efficiency in network communication. Here, the selection of optimal routing path is made using the FGGWO. The data received at BS undergoes the features selection phase and intrusion detection phase for detecting the intruders in IoT network. Here, deep learning classifier is used to identify network intruders.

The feature selection is mainly used to increase the computational efficiency and learning performance. The feature selection phase plays an essential factor in reducing dimension of data and to prevent overfitting. The input data Di is fed to feature selection module for selecting features using the Tversky index. After the selection of unique features, the mechanism of intrusion discovery is done with Deep Maxout. The training of Deep maxout network is done by CSHO algorithm. The working principle of HGSO depends on Henry's law. It has the facility to balance the exploration and exploitation phase by reproducing the gas huddling behavior.

The proposed solution offers security as a service and provides evidence in terms of scalability. However, the proposed method offers better results using the metrics, like energy, F-measure, precision, and recall as 0.1610, 0.9001, 0.9052, and 0.8993, respectively.

Author(s) Details:

Mythili Boopathi,
School of Information Technology, Vellore Institute of Technology, Katpadi, Vellore 632014, India.

Please see the link here: https://stm.bookpi.org/CPSTR-V4/article/view/13135