Having the appropriate amount and quality of signaling data to test telecommunication
data is a constant challenge for vendors and operators alike. Hence, signaling message
generation is often a crucial step during testing. When generating signaling messages,
traditional approaches often fall short in meeting the dynamic and complex requirements
of 5G core and IoT environments. The current paper proposes to leverage Generative
AI, particularly Large Language Models (LLMs), to address these challenges, offering
a novel methodology for generating test data that is both diverse and representative
of real-world scenarios. Through a series of experiments involving various protocols
such as CoAP, HTTP, and data representation formats such as XML, JSON, and SenML,
we evaluated the efficacy of LLMs in generating accurate and high-quality test data.
Our findings demonstrate that Generative AI can enhance the test data generation process
with attention to further needs in improving accuracy.
