Pivotal AI Paradigms for Satellite Communication Security: From Machine Learning to Large Language Models

As satellite communication (SATCOM) networks become increasingly integral to global connectivity, ensuring their security has emerged as a paramount concern. Among all potential techniques for protecting SATCOM security, AI-based approaches have emerged as particularly promising. However, effectively harnessing the diverse spectrum of AI technologies to construct a comprehensive and intelligent defense system for SATCOM networks remains a non-trivial task. Motivated by the scarcity of research that systematically analyzes different AI capabilities to address SATCOM security requirements, this paper aims to fill this gap by providing a holistic set of guidelines for leveraging AI to enhance security in SATCOM. To achieve this, we first establish a SATCOM threat model to identify practical security requirements. We then categorize diverse AI models, from machine learning to large language models, into four pivotal AI paradigms. Building on this, we conduct an extensive survey of existing AI-driven solutions with detailed analysis of the usage of different AI paradigms, to distill crucial lessons learned for applying them effectively. Finally, we summarize open issues and outline the urgent future efforts to advance AI-enhanced SATCOM security. This survey offers a valuable foundation for the future development of a comprehensive, defense-in-depth architecture for SATCOM networks.