Deep Learning for Secure Transmission in Wireless Communication Networks

Advancements in information and communication technologies have contributed significantly to the optimization of next generation wireless communications networks. Moreover, wireless communications play a huge role in electronic warfare and the emergence of new technologies continue to provoke a corresponding revolution in battlefield operations. Nevertheless, the widened attack surface that result from the increased wireless networking of battlefield devices lure attackers like jammers and eavesdroppers who seek to explore the vulnerability of the network through the physical layer. Consequently, it is imperative to direct research efforts not only towards advancing spectral and energy efficiency but also secure wireless communications. In order to address the challenges, we leverage on the recent developments in the use of machine learning (ML) for wireless communications to propose a novel approach for secure wave-form transmission in battlefield operations. Specifically, we use the learning-based end-to-end representation of communication systems where the transmitter and receiver are represented by two deep neural networks (DNN). The goal is to prevent the eavesdropper from accessing the communications between the transmitter and the receiver without necessarily a feedback mechanism as in the conventional communication system. In comparison to traditional optimization methods that are usually iterative in nature, the advantages of employing deep learning (DL) includes its ability to address the challenges of statistically inclined methods. Using simulations, we compare the proposed method with other methods to ascertain its credibility.