Malicious node detection in IoT networks using artificial neural networks
Kazi Kutubuddin Sayyad Liyakat
Abstract
Increasing the rate at which people interact with their gadgets is the primary objective of Internet of Things (IoT). IoT infrastructures are defenceless against hacker attacks in this day and age, which can frequently have unfavourable repercussions. This is despite the fact that IoT structures offer a multitude of benefits. This is a consequence of the continual development of the Internet of Things ecosystem. The IoT is a network that can configure itself automatically. It is possible for rogue nodes to launch attacks against networks in a variety of different methods. A denial of service assault, often known as a DoS attack, can be initiated by malicious nodes via the transmission of massive numbers of packets to a target node. It is necessary to implement a threshold-based technique that makes use of machine learning (ML) methods in order to identify these rogue nodes that are present in our network. An attacker node can be located with the assistance of the suggested approach, which involves monitoring delays in route paths and sending an alarm to the system if the delays surpass a specific threshold. In the past, it has been demonstrated that machine learning, or more specifically, a subset of this technology known as deep learning (DL), has the ability to deliver extraordinary results when it comes to the allocation of diverse data of varying quantities. Through the utilisation of artificial neural networks (ANNs), this chapter presents a method for spotting malicious nodes in Internet of Things (IoT) environments. An NS2 platform will be utilised in order to carry out the techniques that have been suggested. Using a number of different measures, such as accuracy, delay, and packet loss, we evaluate the technique that has been proposed and demonstrate that our system functions effectively. This research, in general, makes two primary contributions to the field. It does two things: first, it classifies the typical and hazardous behaviour of Internet of Things devices in networks, and second, it provides a method for identifying malicious nodes with an accuracy rate of 82.5%.