| Popis: |
Connected and Autonomous Vehicles (CAVs) enable different functionalities and capabilities such as navigation and path planning, automated driving assistance, cruise control, low-carbon transportation, and independent decision-making in real-world scenarios. However, the increased usage of CAVs reduces the possible vulnerabilities in the Internet of Vehicles (IoVs) framework, exposing it to cyber-attacks. An Intrusion Detection System (IDS) is a method to report network attacks by Autonomous Vehicles (AVs) without authorization and encryption techniques for internal and external vehicular transmission. To alleviate this risk, the lightweight IDS system must identify attacks on vehicular systems. Deep learning (DL) approaches promising algorithms for intrusion detection in CAVs, which leverage their capability to automatically extract and learn complex patterns from intricate and vast datasets. DL-based IDS can efficiently identify anomalous activities indicative of system vulnerabilities or cyberattacks by analyzing vehicle behaviour, network traffic, and sensor data in real-time, thus ensuring the integrity and security of CAV operations. Bio-inspired optimization algorithms have recently been used for feature selection (FS) and hyperparameter tuning processes, commonly stimulated by physical ideologies, evolution theories, and specific characteristics of living beings, to solve optimization issues competently in different applications. Therefore, this study introduces a novel Planet Optimization Algorithm with a Deep Ensemble Learning-based Intrusion Detection (POADEL-ID) technique for CAV networks. The presented POADEL-ID model concentrates on identifying intrusions in the CAV networks. In the POADEL-ID technique, linear scaling normalization (LSN) is utilized for the data scaling process. Besides, the high-dimensionality problem is resolved by the POA-based FS approach. Moreover, intrusion detection is performed by using an ensemble of three techniques, namely long short-term memory (LSTM), convolutional autoencoder (CAE), and kernel extreme learning machine (KELM). The multi-verse optimization (MVO) technique is implemented for the hyperparameter tuning process to enhance the |
