Game-theoretic security optimization in UAV networks through AI-enabled digital twin intelligence

The widespread adoption of unmanned aerial vehicle (UAV) networks in mission-critical and intelligent applications has substantially heightened their exposure to sophisticated and evolving cyber threats. The convergence of artificial intelligence (AI), IoT-enabled sensing technologies, and digital twin-assisted network modeling offers a robust foundation for proactive security evaluation and strategic defense planning in UAV communication environments. This study proposes a game-theoretic decision-making framework for comprehensive security assessment in UAV networks. A novel Security Risk Index (SRI) is developed to measure risk-aware security levels across UAV nodes, communication links, and swarm components, which are subsequently integrated to assess overall network resilience, reliability, and robustness. To support adaptive defense mechanisms, a game-theoretic optimization approach is employed to capture and analyze strategic interactions among network entities. The effectiveness of the proposed framework is validated using a public UAV security dataset containing approximately 70,125 communication and attack instances. Experimental findings demonstrate improved performance. The attack classification component exhibits strong predictive performance, attaining 94.34% precision, 93.08% specificity, and 93.57% sensitivity. Furthermore, the game-theoretic inference module delivers a strategic decision accuracy of 97.41%, accompanied by a low prediction error (RMSE = 1.94%). Moreover, the proposed framework demonstrates high operational robustness, achieving an overall system reliability of 94.27% and maintaining strong predictive stability with a score of 0.81.