ConvGRU: A Lightweight Intrusion Detection System for Vehicle Networks Based on Shallow CNN and GRU

ConvGRU: A Lightweight Intrusion Detection System for Vehicle Networks Based on Shallow CNN and GRU

The rapid proliferation of connected vehicles has significantly expanded the attack surface of the Internet of Vehicles (IoV), introducing severe security risks. In such resource-constrained environments, developing lightweight solutions is crucial to ensuring real-time detection and efficient deplo...

Full description

Saved in:
Bibliographic Details
Main Authors: Shaoqiang Wang, Jiahui Cheng, Yizhe Wang, Shutong Li, Lei Kang, Yinfei Dai
Format: Article
Language:English
Published: IEEE 2025-01-01
Series:IEEE Access
Subjects:
Internet of Vehicles
intrusion detection
convolutional neural networks
temporal features
lightweight
Online Access:https://ieeexplore.ieee.org/document/10975752/
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:The rapid proliferation of connected vehicles has significantly expanded the attack surface of the Internet of Vehicles (IoV), introducing severe security risks. In such resource-constrained environments, developing lightweight solutions is crucial to ensuring real-time detection and efficient deployment. To address these challenges, this study proposes ConvGRU, a lightweight vehicular network intrusion detection model that integrates a shallow Convolutional Neural Network (CNN) with a Gated Recurrent Unit (GRU). By employing optimizations such as small convolutional kernels and depthwise separable convolutions, the model significantly reduces the number of parameters and computational overhead, making it well-suited for resource-limited IoV environments. The shallow CNN effectively captures spatial features, while the GRU extracts temporal dependencies, enhancing the model’s generalization ability. ConvGRU achieves an accuracy, precision, recall, and F1-score exceeding 0.99 on the HCRL-Car-hacking, OTIDS, and CICIDS-2018 datasets, with only 112.55K parameters and a memory footprint of merely 0.43 MB. Experimental results demonstrate that this intrusion detection solution substantially improves malicious traffic detection accuracy while ensuring efficient operation in resource-constrained vehicular environments.
ISSN:2169-3536

Similar Items