Lightweight Echo State Network for Vehicular Network Intrusion Detection: A Reservoir Computing Approach

Network intrusion detection systems (NIDS) are vital for safeguarding vehicular networks against emerging cyber threats. While many studies propose deep learning models like DNNs, CNNs, LSTMs, and hybrid architectures for accurate intrusion detection, these methods often come with high computational costs, making them unsuitable for lightweight applications. In this work, we introduce a lightweight Echo State Network (ESN) model leveraging the principles of Reservoir Computing to achieve high-performance intrusion detection in vehicular networks. To address the challenge of dataset imbalance and scarcity, we employ synthetic generation methods such as k-means SMOTE, which enhances the robustness of our model by generating realistic synthetic samples. Our approach is trained and tested on HCRL's Car Hacking dataset and the M-CAN dataset for vehicular network intrusions, as well as general network intrusion datasets (NF-ToN-IoT, NF-UQ-NIDS and NF-BoT-IoT), ensuring its robustness across domains. Despite the scarcity of the latest vehicular network intrusion datasets, our model demonstrates exceptional accuracy and efficiency, outperforming traditional models while remaining computationally efficient. This research contributes to the field of network security by presenting a scalable and lightweight NIDS tailored for vehicular technology networks. We have made our work available on GitHub (https://github.com/codewithkhurshed/VNIDS-SEU) to assist researchers in further improving network intrusion detection systems.