Hybrid Grasshopper Whale Optimization Algorithm Based on Faster Recurrent Convolutional Neural Network for Intrusion Detection System in Internet of Things

Intrusion Detection System (IDS) is one of the general Deep Learning (DL) techniques which are utilized to find and identify the outliers to prevent adversarial attacks, fraud, intrusions in Internet of Things (IoT). The proposed hybrid optimization technique known as GWOA by integrating the Grasshopper Optimization Algorithm (GOA) and the Whale Optimization Algorithm (WOA). For better IDS performance, this GWOA can be utilized into the Faster Recurrent Convolutional Neural Network (FRCNN) model. This proposed method utilized the BoT-IoT dataset to assess the efficacy of this GWOA-based strategy. The main objective was to improve the IDS's balance between exploration and exploitation. This system makes use of the FRCNN to categorize incursions precisely and effectively, improving the overall detection capabilities. When compared to measures of evaluation like precision, recall, accuracy, and f1-score, the suggested GWOA-based FRCNN model's results show a considerable improvement. Comparing the model to other intrusion detection methods, such as Deep Neural Network (DNN), Long Short-Term Memory (LSTM) and IDS-SIoEL. It attains values of 93.77% for accuracy, 86.66% for precision, 95.87% for recall, and 91.03% for f1-score.

I. Introduction

In recent years, Intrusion Detection Systems (IDS) are often utilized to improve the security of device communication and detect network intrusions. These IDSs serve a critical role in ensuring secure internet connectivity and rapidly alerting system administrators when harmful actions are detected [1]. there has been a growing trend in the adoption of cloud computing (CC) in both the educational and commercial sectors. It is generally recognized for its utility in data storage and retrieval in cloud systems [2]. DL-based algorithms in particular confront issues such as high model complexity and big model size [3]. Efforts have been made to integrate cybersecurity systems with modern communication technologies, particularly in industries such as manufacturing, where monitoring and control capabilities for industrial systems are critical [4]. As wireless technology and equipment become more control, guaranteeing the security of wireless communication channels becomes increasingly important. These channels carry serious business data such as remote control, telemetry, and remote signaling, needing protection against malicious node activity and unauthorized access [5]-[6]. The limitation of energy resources, particularly in devices with low battery capacity, presents a substantial challenge to network performance. As a result, reducing energy consumption is a critical requirement for attaining quality of service (QoS) in the Internet of Things (IoT) setting [7]. The Internet of Things has the potential to boost production and efficiency in smart sectors through intelligent decision-making and remote management. However, the rapid expansion of IoT networks has generated uncertainties about security and privacy [8]. The major contributions are discussed as follows; •

The preprocessing is done after collection data from BoT-IoT dataset which improves the model performance and fed to the feature selection.

•

The hybrid Grasshopper Optimization Algorithm - Whale Optimization Algorithm (GWOA) is utilized for the feature selection process and Faster Recurrent CNN (FR-CNN) is utilized for classification.

•

The proposed GWOA based on FRCNN model was evaluated by utilizing evaluation metrics like accuracy, precision, recall, specificity, and f1-score.