Empowering Reconfigurable Intelligent Surfaces with Artificial Intelligence to Secure Air-To-Ground Internet-of-Things

Abstract

Reconfigurable intelligent surfaces (RISs) and unmanned aerial vehicles (UAVs) have the potential to play a significant role in enhancing the security of the Internet-of-Things (IoT). RISs can be deployed as intelligent reflectors to augment wireless coverage passively. UAVs offer flexible and dynamic IoT platforms for communication, sensing, and monitoring. In this article, a particular interest is given to RIS-assisted, anti-jamming, UAV communication and radio surveillance, which are generally nonconvex and difficult to solve using traditional optimization tools. New artificial intelligence (AI) tools, more specifically, deep reinforcement learning (DRL), are developed to tackle the problems of UAV and RIS design. The use of DRL allows a UAV to learn its trajectory and RIS configuration to diffuse jamming signals and maximize its communication rate based on its received data rate. It also allows the UAV to maximize its eavesdropping rate based on the transmit rate of a suspicious transmitter that the UAV observes when conducting radio surveillance. The UAVs no longer rely on explicit knowledge of the channel state information, and can learn through trial and error. Simulations confirm the effectiveness of using UAVs, RISs, and AI to enhance the security of air-to-ground IoT networks, compared to baseline schemes without RIS or with non-AI-based RIS configurations.