Hybrid-CIID: Hybrid Chaotic Mongoose Optimized Intrusion Detection in IoT Networks via the Convolutional Bi-GRU Model

The Internet of Things (IoT) system has become integral to everyday life and security is paramount to ensure data integrity and privacy. However, traditional Intrusion Detection Systems (IDSs) struggle to detect distributed attacks or emerging threats due to their inability to analyze complex interactions with connected devices. To address these problems, a novel Hybrid Chaotic Mongoose optimized Intrusion IDentification in IoT System (Hybrid-CIID) framework is designed to enhance the security strength of the data in IoT networks. The data gathered from the IoT devices are preprocessed to enhance the quality and reliability of the data. An unsupervised Deep Learning (DL) approach extracts both the structural and temporal features from the IoT data graph to identify the intrusive data patterns. The most relevant features are selected by a hybrid chaotic metaheuristic algorithm, and it is labelled as authentic, suspicious, or threat data to prevent the system from potential attacks. A Convolutional-based Recurrent Network analyzes the labeled data and classifies the intrusions into five distinct classes to enhance security by preventing the systems from unauthorized access. The Hybrid-CIID framework is validated by using CICIDS 2017, CSE-CIC-IDS 2018, and CICIoT2023 datasets, and it is simulated by employing Cloud Simulator (CloudSim 7G). The performance of the Hybrid-CIID framework is assessed by such metrics as detection rate, computational complexity, detection time, False Alarm Rate (FAR), scalability, and security strength. The Hybrid-CIID methodology attains an overall accuracy of 97.84%, whereas the TLBO-IDS, WILS-TRS, and CyberAIBot method achieves 87.47%, 89.60%, and 94.75% for efficient intrusion detection in IoT environments.