Research on Situational Awareness and Threat Assessment Model Using Reinforcement Learning and GNN

This paper proposes an innovative situational awareness and threat assessment model combining reinforcement learning and GNN. The model uses GNN to model the in-vehicle communication network topology of intelligent connected vehicles and capture the interactive relationship between vehicles and infrastructure to achieve global situational awareness. Reinforcement learning (RL) is used to automatically adjust the threat detection strategy according to the real-time network status and optimize the threat assessment process. By combining the global information processing capability of GNN and the dynamic strategy optimization of RL, the system can more effectively identify various potential attacks and respond quickly to changing network environments. The data set generated by the in-vehicle network simulation platform is used to simulate normal communication traffic and various network attacks (such as denial of service attacks, data tampering attacks, etc.). The proposed model improves the threat detection rate by 20% compared with the traditional method, reduces the false alarm rate by 10%, and shortens the threat assessment response time to less than 5 seconds. The model not only improves the accuracy and real-time performance of security detection, but also enhances the system's anti-attack capability and adaptability, providing an innovative solution for the communication security of intelligent connected vehicles.