What question did this study set out to answer?

The aim is to develop a Hybrid Deep Learning Intrusion Detection System for Vehicular Fog Computing Networks that enhances security.

July 3, 2026Open Access

Hybrid deep learning and graph-based intrusion detection for vehicular fog computing networks: leveraging transformers, GNNs, and GANs for enhanced security

Key Points

The aim is to develop a Hybrid Deep Learning Intrusion Detection System for Vehicular Fog Computing Networks that enhances security.
Integrates transformers for temporal modeling of network traffic.
Utilizes graph neural networks to capture spatial relationships among vehicles and fog nodes.
Employs generative adversarial networks for unsupervised detection of unknown attack behaviors.
Achieved detection accuracy of 98.5% with a false positive rate of 1.2%.
Demonstrated a low-latency detection time of approximately 45 ms.
Improved detection accuracy by 5% and reduced false positives by over 30% compared to baseline models.

Abstract

Vehicular Fog Computing Networks (VFCNs) have emerged as a critical component of intelligent transportation systems, enabling latency-sensitive and safety-critical applications such as autonomous driving, real-time traffic management, and vehicle-to-everything communication. However, the highly dynamic topology, distributed architecture, and rapid mobility of VFCNs expose them to sophisticated cyber threats, including Distributed Denial-of-Service (DDoS) attacks, location spoofing, and zero-day attacks. Existing intrusion detection systems (IDS) often struggle to meet the scalability, accuracy, and low-latency requirements of such environments. To address these challenges, this paper proposes a Hybrid Deep Learning Intrusion Detection System (HDL-IDS) tailored for Vehicular Fog Computing Networks. The proposed framework integrates Transformer-based temporal modeling to capture long-range attack evolution in network traffic, Graph Neural Networks (GNNs) to represent spatial and topological relationships among vehicles and fog nodes, and Generative Adversarial Networks (GANs) to enable unsupervised detection of previously unseen attack behaviors. These complementary learning components are combined through a decision-level fusion mechanism, allowing the system to jointly model temporal, spatial, and anomaly characteristics within dynamic vehicular environments. Extensive experiments conducted on multiple benchmark datasets, including CICIDS 2017, ToN-IoT, and VeReMi, demonstrate that the proposed HDL-IDS consistently outperforms conventional machine learning, deep learning, and existing hybrid IDS approaches. The proposed model achieves a detection accuracy of 98.5%, a false positive rate of 1.2%, and a low-latency intrusion detection system (~ 45 ms), suitable for latency-sensitive vehicular environments, while improving detection accuracy by about 5% and reducing false positives by more than 30% compared to baseline models. These results indicate that the proposed HDL-IDS provides an effective, scalable, and low-latency security solution for safeguarding Vehicular Fog Computing Networks against evolving cyber threats.

Ask AI

Helpful

Bookmark

View Full Paper