Deep Learning-Based Intrusion Detection Systems in VANETs: A systematic Literature Review

Abstract

VANETs are key to intelligent transport, offering significant benefits for safety, efficiency, and self-driving cars. However, the wireless nature of VANETs makes them vulnerable to a variety of attacks. A robust security measure is a mandatory requirement. This paper provides a scientific literature review that examines deep learning (DL) methods applied to Intrusion Detection Systems (IDS) within VANETs. We carefully reviewed recent research, identifying a range of deep learning architectures. Based on our findings, various DL models are implemented to enhance the security of VANETs, including Deep Belief Networks (DBNs), Recurrent Neural Networks (RNNs) like LSTMs and GRUs, and Convolutional Neural Networks (CNNs). The findings consistently show that these DL models achieve higher accuracy in detecting attacks than traditional machine learning models. Their performance is typically evaluated using confusion matrix metrics, accuracy, precision, recall and F1-score. Regardless of their effective, some key challenges were discovered. DL models face challenges like large computational demands, privacy concerns, and a scarcity of quality training data. Also, we explore developing trends and future routes, for example Federated Learning (FL), Software-Defined Networking (SDN), and blockchain integration, which seem capable of addressing VANETs’ challenges. In summary, this review indicates deep learning is crucial to strengthening IDS within VANETs. Implementing these advanced techniques is essential for safe and reliable intelligent transport system

VANETs, Deep Learning, Intrusion Detection Systems, Cybersecurity, Road Safety, Systematic Literature Review, Machine Learning