An Attention Mechanism Spatio-Temporal Graph Framework for Misbehavior Detection in IoV

In recent years, there has been a growing interest in applying Deep Learning (DL) techniques for Misbehavior Detection in the Internet of Vehicles (IoV). Despite this, network anomalies caused by various attack types continue to pose significant challenges, such as incomplete misbehavior detection, inefficient multi-model detection, and insufficient accuracy. To address these issues, this paper introduces a novel detection framework, specifically designed for the Internet of Vehicles (IoV), which is called Attention Mechanism Spatio-Temporal Graph Framework (ATG). In contrast to traditional DL models, the framework is a multi-classification model that integrates Graph Neural Network (GNN), Gated Recurrent Unit (GRU) and Spatial Attention Mechanism. The proposed scheme enhances feature engineering through the Attention Mechanism, utilizes a node aggregation algorithm within the GNN, and achieves global detection via the GRU. Experimental results demonstrate the model's performance in different environments using the dataset generated in Veins, and provide a comparison with baseline methods. Under typical conditions, the model achieves an accuracy of 94.6%, a precision of 88.8%, a recall of 90.9%, and an F1-Score of 89.1%. ATG shows a precision improvement of 1-10% and a recall improvement of 1-16% compared to other schemes, demonstrating enhanced security and reliability.