Deep Multimodal Learning for Real-Time Ddos Attacks Detection in Internet of Vehicles
Mohamed Ababsa, Soheyb Ribouh, Abdelhamid Malki, Lyes Khoukhi · 2025
The advancement of Intelligent Transport Systems (ITS) and the Internet of Vehicles (IoV) enhances road safety and traveler comfort by enabling safer, more efficient transportation networks. However, these technologies are vulnerable to a range of security threats that malicious actors could exploit. One of the most severe threats to IoV is the Distributed Denial of Service (DDoS) attack, which could disrupt traffic flow, disable vehicular communication, or even cause accidents. This paper proposes a novel Deep Multimodal Learning (DML) approach to detect DDoS attacks in IoV, strengthening cybersecurity in intelligent transport systems. Our DML model integrates Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) networks, enhanced by Attention and Gating mechanisms, alongside a Multi-Layer Perceptron (MLP) with a multimodal intermediate fusion architecture. This innovative method leverages the Framework for Misbehavior Detection (F2MD) to generate a synthetic dataset and deploy our model, enabling real-time DDoS detection and mitigation while overcoming the limitations of the Vehicular Reference Misbehavior (VeReMi) dataset. The proposed approach is evaluated in real-time across various simulated real-world scenarios with differing attacker densities. Our DML model achieves an average accuracy of$\mathbf{9 6. 6 3 \%}$, outperforming classical Machine Learning (ML) and state-of-the-art approaches, demonstrating significant efficacy and reliability in safeguarding vehicular networks against malicious cyberattacks.