Adversarial Attacks on Machine Learning-Based IDS for V2X Networks: A CICIoV2024 Study

With the increasing adoption of Vehicle-toEverything (V2X) communication in next-generation intelligent transportation systems, ensuring cybersecurity in vehicular networks has become a critical challenge. Machine Learning (ML)-based Intrusion Detection Systems (IDS) have been widely proposed to detect anomalous and malicious activities in V2X networks. However, these systems are vulnerable to adversarial attacks, where an attacker subtly manipulates input data to evade detection. This study investigates the impact of adversarial attacks on ML-based IDS using the CICIoV2024 dataset, which includes diverse cyber threats targeting V2X networks. We evaluate Fast Gradient Sign Method (FGSM), Projected Gradient Descent (PGD), Carlini & Wagner (CW), and DeepFool attacks against ML models such as Random Forest (RF), Support Vector Machines (SVM), Convolutional Neural Networks (CNN), and Transformer-based IDS. Our results demonstrate that adversarial attacks significantly degrade IDS performance, reducing detection accuracy by up to 40%, thereby exposing critical vulnerabilities in V2X cybersecurity frameworks. Furthermore, we assess defence mechanisms, including adversarial training and feature squeezing, and analyze their effectiveness in mitigating adversarial threats. The findings highlight the urgent need for robust, adversarially-aware IDS to safeguard V2X networks from evolving cyber threats. Future research directions include realworld deployment, federated learning-based IDS, and adversarially robust AI architectures for V2X security.