G-Net: A Novel Intrusion Detection Model for VANETs Based on Convolutional Neural Network

Abstract

Vehicular Ad-hoc Networks (VANETs) play a critical role in intelligent transportation systems by enabling seamless communication between vehicles and infrastructure. These networks optimize traffic management, enhance road safety, and support the development of autonomous driving technologies. However, their reliance on wireless communication makes them vulnerable to various security threats, necessitating robust intrusion detection systems (IDS). A network intrusion detection model based on deep learning has become a significant area of research to the growth of cyber threats and the limitations of traditional detection methods. Deep learning techniques, such as convolutional neural networks (CNNs), offer advanced capabilities for feature extraction, enabling more accurate detection of anomalies and malicious activities within network traffic. In this paper, we propose G-Net, a novel network intrusion detection model using convolutional neural networks CNNs for multi-class classification. This model is inspired by the gates mechanisms used in recurrent neural networks (RNNs) such as Gated Recurrent Units (GRU) and Long Short-term Memory (LSTM) networks. The proposed G-Net model allows to selectively focus on significant features in network traffic data while filtering out irrelevant information. The performance of the proposed intrusion detection model is evaluated on KDD-NSL dataset and compared with other existing works. Empirical results show that the proposed model achieved an accuracy of 98,57%, precision of 98,64%, recall of 98,57% and an F1-score of 98,58%, for 5-class classification using NSLKDD dataset. This model is able to automatically identify and classify several types of intrusions, including DoS, probe, R2L and U2R attacks. Therefore, it can be an effective and reliable solution for the intrusion detection.