Optimal Strategy for Moving Target Defense on the Internet of Vehicles Based on Game Theory and Reinforcement Learning

As an important part of the intelligent transportation system, the Internet of Vehicles (IoV) optimizes traffic management, improves road safety, and travel experience by connecting vehicles with the cloud, infrastructure, pedestrians, and other vehicles. However, with the rapid development and widespread application of IoV technology, its network security issues have become increasingly prominent. Its attack threats may not only cause the functional failure of individual vehicles, but also seriously impact the stability of the entire transportation system. With the expansion of scale and increase in complexity, new threats and large-scale communications have put forward higher requirements for the security protection capabilities of IoV. Therefore, the introduction of adaptive active defense mechanisms can further improve the security and adaptability of the system. Because of the diverse defense needs of different levels of IoV systems, this paper constructs an intelligent agent attack and defense game scenario, and proposes an optimal defense strategy for moving targets based on an incomplete information Bayesian Markov game and Q-learning algorithm from the perspectives of adaptive active defense, reducing resource overhead, and intelligent defense scheduling. This strategy targets different types of attack problems existing at the vehicle side, pipe side, and cloud side of the IoV system. Taking into account four complex attack scenarios, namely scanning attack, false message injection attack, denial of service attack (DoS), and malware attack, this strategy establishes an incomplete information Bayesian Markov game model and adopts the reinforcement learning Q-learning algorithm to dynamically adjust the game environment strategy by continuously interacting with the environment to obtain reward values. This strategy seeks the most suitable defense elements and the optimal mutation time of defense elements for network attacks at all levels of the IoV, solves the problem of intelligent scheduling of defense resources under active defense systems at different levels, and avoids the limitations of a single defense mechanism. The experimental results show that compared with the existing comparative algorithms, the proposed method exhibits better defense performance and limitations in dealing with multi-level compound attacks on the IoV, and provides a new technical approach for building an intelligent and active IoV security protection system. This paper uses OMNeT++, SUMO, and Veins network joint simulation tools to verify the effectiveness of the above two solutions through four performance indicators: attack success rate, data packet loss rate, communication delay, and system overhead. The simulation results show that the defense mechanism performs well in reducing attack success rate, data packet loss rate, communication delay, and system overhead, and the overall defense effect is significant, especially when dealing with complex and changing attack scenarios.