Modeling Penetration and Defense in Covert Video Surveillance via Attack Path Generation

With the rapid expansion of the Transportation Internet of Things, covert video surveillance systems have emerged as critical nodes for information acquisition and transmission. However, their inherent network complexity also exposes them to significant security risks, particularly penetration attacks aimed at intelligence exfiltration. To address this challenge, this study proposes a systematic framework modeling approach for covert video surveillance systems, develops an attack path generation algorithm, and establishes an agent-based modeling framework to analyze attack success probability. Furthermore, three defense strategies are designed: static rule-based defense, graph reasoning-based path interception, and reinforcement learning-driven dynamic response. Simulation experiments comparing the performance of different attack-defense combinations demonstrate that the reinforcement learning strategy achieves optimal outcomes in reducing attack success rates, minimizing system losses, and improving resource utilization. This research provides methodological and theoretical foundations for securing Internet things, offering practical insights for protecting covert information in real-world scenarios.