Secrecy Anti-jamming Game Learning in D2D Underlay Cellular Networks with an Active Eavesdropper

In this paper, we study the physical layer security and transmission reliability problem where there is an active eavesdropper (AE) in the D2D underlaying cellular networks. We formulate the cooperation between the cellular user equipment (CUE) and the D2D user equipment (DUE), the completion between legitimate users and the AE to be a secrecy anti-jamming game. In the proposed game framework, DUE launches the cooperative relaying or the friendly jamming mode to help CUE to improve its anti-eavesdropping and anti-jamming performance. CUE gives different-level rewards for the assistance of the DUE. And AE shifts its attacking modes between actively jamming and passively eavesdropping to maximize the destruction for the D2D underlaying cellular networks. Under the perfect information, we prove the existence of the pure-strategy equilibrium of the proposed game. Under the imperfect information, we analyze the existence of the mixed-strategy equilibrium of the proposed game and propose a distributed Q-Iearning-based algorithm to converge to a mixed-strategy equilibrium. Simulation results show that the proposed algorithm is convergent and verify that average utilities of legitimate users are improved by the cooperation between CUE and DUE.