In response to the vulnerability of UAV-based air-to-ground communication links to eavesdropping attacks, this paper proposes a Covert Information Mapped Generalized Spatial and Direction Modulation (CIM-GSDM) system. The system hides information within the index of the activated receiver subset and their selection combinations, while introducing artificial noise that is orthogonal to the legitimate channel to interfere with the eavesdropper. To further enhance the transmission security of the system, a joint optimization framework for the precoding matrix and power allocation factors is proposed. This framework effectively manages multi-beam transmission and artificial noise power allocation, thereby improving the system"s secure rate. First, the physical layer security metric based on the system"s secure rate is derived, and this is used as the optimization objective for the joint optimization of the precoding matrix and artificial noise power allocation factors. To solve the non-convex joint optimization problem, alternating optimization of the two variables is considered, and a natural gradient descent algorithm based on Nesterov’s method is proposed. This method quickly iterates to update the precoding matrix and resolves the computational complexity issue caused by the large size of the CIM-GSDM symbol candidate set. Based on the criterion of maximizing the product of the legitimate receiver’s signal-to-noise ratio (SNR) and the eavesdropper’s interference-to-signal-plus-noise (ISNR), the suboptimal closed-form solution for the power allocation factor is derived. Simulation results show that the proposed optimization algorithm significantly reduces the eavesdropper’s interception rate while ensuring the legitimate receiver’s achievable rate, effectively securing the transmission in the CIM-GSDM system. Compared to traditional beamforming algorithms and methods with fixed power allocation factors, the proposed method demonstrates a significant advantage in security performance.