Wireless Sensor Network-based Intrusion Detection Technique using Deep Learning Approach of CNN-GRU

In a WSN, sensors are always collecting information and transmitting it to other nodes. Their primary uses are in fields like defense, smart city development, and agricultural monitoring. The WSN needs to perform at a premium for these uses. Yet, there are many potential security concerns that could compromise a WSN's performance. Any interference with the WSN could severely degrade its functionality and lead to catastrophic failures. As a result, having the ability to quickly identify and stop intrusions is crucial. The goal of this proposed is to use a GRU-CNN model for quick detection and prevention of any intrusion. After receiving the input, the proposed method uses normalization and discretization for preprocessing the data, PSO (Particle Swarm Optimization) for feature extraction, CFS for feature selection and finally training the model by CNN-GRU. To better detect intrusions in wireless sensor networks, a GRU-CNN hybrid neural network model was presented; although the CNN module is responsible for extracting the feature vector from other high-dimensional data, the GRU module is responsible for from time sequence data to extract the feature vector.

I. INTRODUCTION

WSN development offers a promising alternative for keeping tabs on critical infrastructure, and it has been recommended for use in fields as varied as traffic monitoring, building monitoring, healthcare, and combat surveillance [1]. Every attempt to disrupt the operation of essential services has the potential to be exploited for either terrorist intentions or illegal financial gain. Network security is vital in these environments and must be implemented in a way that accounts for the constraints of these systems so that their weaknesses and vulnerabilities may be fixed. In this proposed approach to investigate the problem of incorporating intrusion detection into WSN and present a novel solution based on mobile agents to solve it. Its ease of installation and operation is a major selling advantage. Yet, creating a safe network of sensors is a major challenge. Because to advancements in wireless communications, micro-electromechanical systems, and digital electronics, low-cost sensor devices, tiny may now be mass-produced. Storage space, Size, processing power, battery life, range, and data speed are all factors that define these gadgets. These devices are well suited for large-scale monitoring and control of physical events because of the range of transducers that may be attached to them. In [2], the authors of a clustering mutual coordination detection model that accounts for the quirks of wireless sensor networks reported their work. The percentage of accurate detections and the number of false positives have both gone up. Unknown assault detection is also quite poor. Intrusion detection is evolving in response to the rise of networked environments. While intrusion detection's initial focus was on single systems [3], the industry has recently switched its emphasis to networks [4]. In Network-oriented Intrusion Detection Systems, the source of events (and the object of analysis) is a distributed system consisting of many hosts and network links (NIDSs). Threats affecting several hosts and the network as a whole can be detected by NIDSs. The widespread adoption of WSNs in recent years has greatly aided human endeavors. While there are many advantages to WSN, there are also new dangers to people's safety and property. This has led to an increase in attention paid to the problem of "sensor network security" in the discipline of computer science. The purpose of network intrusion detection systems (IDS) is to, as the name implies, detect intrusions into a network [5][6]. Intrusion detection systems (IDS) are designed to protect sensitive data and physical assets by monitoring and analyzing data from a system's or networks most vital nodes in the event of an attack. Using a network-based intrusion detection system (NIDS) to monitor and analyze network traffic is essential for protecting a system from network-based attacks. Each incoming packet is inspected by a network intrusion detection system (NIDS) for irregularities. Depending on the severity of the threat, the system may provide an alert to administrators or prevent access to the network from the offending IP address. Snort is a network intrusion detection and prevention system developed as open-source software (NIPS). Snort is a multi-purpose network intrusion prevention system that may also act as a packet recorder and sniffer. Using wireless networks, the sensor nodes collect data and relay it to a sink or base station. When compared to traditional networks, WSNs have severely constrained resources in the areas of energy, communication range, and computational capacity. Thus, in the design of WSNs, efficiency and data centralization are given high importance [7].