Generative Adversarial Network Assisted Power Allocation for Cooperative Cognitive Covert Communication System
Xiaomin Liao, Jiangbo Si, Jia Shi, Zan Li, Haiyang Ding
Abstract
This letter investigates a power allocation problem for a cooperative cognitive covert communication system, where the relay secondary transmitter (ST) covertly transmits private information under the supervision of the primary transmitter (PT). Aiming to achieve the tradeoff between the covert rate and the probability of detection errors, a novel generative adversarial network based power allocation algorithm (GAN-PA) is proposed to perform power allocation at the relay ST for covert communication. Under the proposed GAN-PA, the generator adaptively generates the power allocation solution for covert communication, while the discriminator determines whether transmitting covert message or not. In particular, by utilizing the proposed deep neural network (DNN), the discriminator and the generator are alternately trained in a competitive manner. Numerical results show that the proposed GAN-PA can attain near-optimal power allocation solution for the covert communication and achieve rapid convergence.