IDTD-RL: An Intelligent Dual-Track Defense Risk-Sensitive Reinforcement Learning Framework for Anti-Spoofing in UAV Communications

Intelligent spoofing poses a critical threat to unmanned aerial vehicle (UAV) communications by corrupting synchronization and channel estimation while evading power-based detection. This paper presents IDTD-RL, a risk-sensitive reinforcement learning framework for adaptive anti-spoofing defense in OFDM-based UAV systems. The hierarchical controller, guided by a conditional value-at-risk-based objective, dynamically balances performance optimization and tail-risk mitigation in uncertain and evolving interference conditions. Once spoofing is detected, the framework initiates a dual-track transmission scheme that maintains pilot-level separation and constrains signal-level cross-track leakage. A phantom rate control(PRC) mechanism introduces temporal variability to disrupt adversarial prediction, while fast constraint projection ensures real-time feasibility. Simulations under the 3GPP TR 36.777 UMi-AV air-to-ground channel demonstrate that IDTD-RL achieves 22.37 Mbps covert throughput with 99.64% SINR compliance and 0.36% packet error rate at 0 dB interference-to-signal ratio, effectively redirecting 89.71% of spoofing energy to the decoy. These results confirm that the proposed framework substantially enhances the reliability and resilience of UAV communications under intelligent spoofing conditions.