Fuzz testing is a crucial technique for uncovering system vulnerabilities, particularly focusing on security. However, its effectiveness is challenged when testers are unfamiliar with communication protocols. This research addresses these challenges, taking the Bluetooth Low Energy (BLE) protocol as a case study. We introduce an approach based on Generative Adversarial Networks (GANs) to create diverse data across varied protocol categories, significantly improving the quality of inherently imbalanced protocol testing datasets. Our methodology allows for application to different testing targets, enabling efficient generation of fuzz testing data without prerequisite knowledge of target specifications, thereby simplifying the process for testers and advancing fuzz testing methodologies in security testing domains.