Generative AI Empowered Covert Communications: Autonomy, Efficiency, and Suitability

Given tremendously growing security and privacy concerns, there is a substantial surge in demand for covert communication techniques. Traditional methods, constrained by static configurations and an inability to leverage historical data, suffer from inefficiencies in resource utilization, vulnerability to evolving threats, and limited adaptability to interference. This article introduces a Generative Artificial Intelligence (GAI)-driven framework to overcome such limitations, integrating advanced models such as Variational Autoencoders (VAEs), Generative Adversarial Networks (GANs), and Generative Diffusion Models (GDMs) into covert communication systems. By employing unsupervised learning, the framework dynamically optimizes transmission parameters—including power allocation, modulation schemes, and frequency hopping—in real time, enabling adaptive responses to channel conditions and historical patterns. Through this optimization, the proposed approach enhances security via unpredictable transmission patterns, reduces bit error rates with optimized modulation and coding schemes, and strengthens resistance to interference and interception through noise-like signal generation. In addition to these advancements, the paper identifies key challenges and further outlines future directions. This work underscores GAI’s transformative potential in balancing concealment, efficiency, and adaptability for next-generation covert communication systems.