Abstract
Low-altitude wireless networks (LAWNs) have the potential to revolutionizecommunications by supporting a range of applications, including urban parceldelivery, aerial inspections and air taxis. However, compared with traditionalwireless networks, LAWNs face unique security challenges due to low-altitudeoperations, frequent mobility and reliance on unlicensed spectrum, making itmore vulnerable to some malicious attacks. In this paper, we investigate somelarge artificial intelligence model (LAM)-enabled solutions for securecommunications in LAWNs. Specifically, we first explore the amplified securityrisks and important limitations of traditional AI methods in LAWNs. Then, weintroduce the basic concepts of LAMs and delve into the role of LAMs inaddressing these challenges. To demonstrate the practical benefits of LAMs forsecure communications in LAWNs, we propose a novel LAM-based optimizationframework that leverages large language models (LLMs) to generate enhancedstate features on top of handcrafted representations, and to design intrinsicrewards accordingly, thereby improving reinforcement learning performance forsecure communication tasks. Through a typical case study, simulation resultsvalidate the effectiveness of the proposed framework. Finally, we outlinefuture directions for integrating LAMs into secure LAWN applications.