Deep Learning for Robust and Secure Wireless Communications

The rapid development of mobile technologies has enabled billions of people worldwide to stay connected, facilitating a broad spectrum of activities ranging from information access to social networking. However, the emergence of numerous wireless applications is driving the demand for spectrum to unprecedented levels. Simultaneously, wireless systems are becoming increasingly software-driven, reducing the barrier for wireless threats such as intelligent jammers and malicious drones. Consequently, one of the most crucial requirements of wireless and mobile systems today is to ensure the robustness of communications against wireless attacks and coexisting interference. Recent advancements in Deep Learning have achieved widespread success in various fields such as natural language processing and computer vision. This success has prompted new approaches to enhance the security and robustness of wireless communication systems. In this chapter, we present three Deep Learning-based solutions for achieving robust and secure wireless communications. The first is a real-time system capable of detecting, classifying, and spectro-temporally localizing wireless collisions and emissions across a wide range of RF technologies. This system is essential in countering wireless threats such as jammers or malicious drones, while also facilitating spectrum understanding and management. Next, we propose JaX, an anti-jamming approach that can detect and cancel high-power jammers. JaX utilizes Convolutional Neural Network to predict the existence of a jammer, estimate the phase and amplitude of the jamming signal, and remove the jamming signal from the received signal. Finally, the chapter introduces DEFORM, a Deep Learning-based universal beamforming approach that leverages neural network and multi-antenna algorithms to improve the robustness of wireless communications. DEFORM is agnostic to various types of RF signals and can be deployed on any RF receivers.