Radio Frequency (RF) fingerprinting enables device authentication by exploiting hardware-induced signal imperfections, but the construction of large-scale RF fingerprinting datasets requires significant effort and expertise, making them valuable intellectual assets. Protecting such datasets from unauthorized use is therefore an important challenge. In this paper, we propose a lightweight and robust watermarking framework for RF fingerprinting datasets under black-box verification settings, enabling reliable watermark verification while preserving benign accuracy and minimizing unintended effects on unrelated classes. Experimental results on different datasets demonstrate that the framework achieves high verification success, low harmful degree, and strong robustness to channel noise, making it suitable for practical RF dataset copyright protection.
