In this paper, we aim to secure the D2D communication of the D2D-underlaid cellular network by leveraging covert communication to hide its presence from the vigilant adversary. In particular, there are adversaries aiming to detect D2D communications based on their received signal powers. To avoid being detected, the legitimate entity, i.e., D2D-underlaid cellular network, performs power control so as to hide the presence of the D2D communication. We model the combat between the adversaries and the legitimate entity as a two-stage Stackelberg game. Therein, the adversaries are the followers and aim to minimize their detection errors at the lower stage while the legitimate entity is the leader and aims to maximize its utility constrained by the D2D communication covertness and the cellular quality of service (QoS) at the upper stage. Different from the conventional works, the study of the combat is conducted from the system-level perspective, where the scenario that a large-scale D2D-underlaid cellular network threatened by massive spatially distributed adversaries is considered and the network spatial configuration is modeled by stochastic geometry. We obtain the adversary's optimal strategy as the best response from the lower stage and also both analytically and numerically verify its optimality. Taking into account the best response from the lower stage, we design a bi-level algorithm based on the successive convex approximation (SCA) method to search for the optimal strategy of the legitimate entity, which together with the best response from the lower stage constitute the Stackelberg equilibrium. Numerical results are presented to evaluate the network performance and reveal practical insights that instead of improving the legitimate utility by strengthening the D2D link reliability, increasing D2D transmission power will degrade it due to the security concern.

翻译：本文旨在通过利用隐蔽通信来隐藏D2D通信的存在，从而保障D2D底层蜂窝网络中D2D通信的安全性。具体而言，存在试图根据接收信号功率检测D2D通信的对手。为避免被检测，合法实体（即D2D底层蜂窝网络）实施功率控制以隐藏D2D通信的存在。我们将对手与合法实体之间的对抗建模为两阶段Stackelberg博弈。其中，对手作为跟随者，其目标是在下层阶段最小化检测误差；而合法实体作为领导者，其目标是在满足D2D通信隐蔽性和蜂窝服务质量（QoS）约束的条件下，最大化自身效用。与传统工作不同，本研究从系统级视角展开对抗分析，考虑大规模D2D底层蜂窝网络受到海量空间分布对手威胁的场景，并利用随机几何对网络空间配置进行建模。我们推导出对手在下层阶段的最优策略作为最佳响应，并通过解析和数值方法验证其最优性。考虑下层阶段的最佳响应，我们设计了一种基于逐次凸近似（SCA）方法的两层算法来搜索合法实体的最优策略，该策略与下层阶段的最佳响应共同构成Stackelberg均衡。数值结果用于评估网络性能，并揭示了实用洞见：提升D2D链路可靠性并非通过增强D2D发射功率来改善合法效用，相反，出于安全考虑，增加D2D发射功率会降低合法效用。