Yingwei Zhao

As critical national infrastructure, the reliability and security of wireless communication and environmental sensing under the smart grid system face significant challenges in complex electromagnetic environments. This paper proposes an Integrated Sensing and Communication (ISAC) architecture based on Time-Modulated Metasurface (TMM) to address these challenges. Leveraging TMM’s capability to simultaneously generate multiple controllable harmonics, the system enables highly reliable communication, sensing, and active electronic countermeasure functions on a single hardware platform, applicable to conventional data transmission, Uncrewed Aerial Vehicle (UAV) inspection, and other scenarios. Using a time-modulated reflection with angle-adaptive linear phase, the system achieves deception in both range and Doppler, effectively shielding critical power infrastructure from radar detection. For communication security, TMM is used for directional modulation, focusing signals toward legitimate receivers while radiating jamming signals toward potential eavesdropping directions, thereby achieving physical-layer security. Simulation results demonstrate excellent performance in electronic countermeasure scenarios, providing an innovative solution for the secure operation of critical power facilities in complex electromagnetic environments.

In Internet of Things (IoT) environments, malicious nodes can not only eavesdrop on legitimate communications but also emit jamming signals, which severely undermine secure and reliable communication among nodes. Reconfigurable Intelligent Surfaces (RIS) have emerged as a lightweight and promising technique to mitigate such threats. However, in typical scenarios that simultaneously involve RIS and multiple nodes, existing Deep Reinforcement Learning (DRL)-based methods often suffer from lengthy training procedures and are vulnerable to adversarial perturbations (AP) introduced by malicious nodes. This paper proposes an offline secure and anti-jamming communication decision framework based on a Temporal Convolutional Network (TCN)-based Twin-Efficiency Diffusion Model (TCN-TEDM). The framework effectively alleviates the tendency of traditional offline RL to yield degraded policies under strong jamming due to adversarial perturbations and state missingness. The model reconstructs perturbed states under a limited compute budget, thereby markedly improving decision stability under jamming. Extensive experimental results demonstrate that the proposed framework can efficiently generalize and transfer existing secure-communication and anti-jamming policies without online interaction, achieving superior stability and performance under state-perturbation and missingness scenarios.