A Robust RF Fingerprint Extraction Scheme for GNSS Spoofing Detection

Global navigation satellite systems (GNSS) have played an important role in space stations, aviation, maritime and mass transit. One of the main disadvantages of GNSS is their vulnerability to spoofing. A successful spoofing attack can have serious consequences. In regards to this issue, our method of GNSS spoofing detection based on radio frequency fingerprint (RFF) is considered a promising technology. Due to manufacturing defects, even GNSS transmitters of the same model exhibit subtle differences known as RFF, which possess uniqueness and persistence, and can be considered as the DNA of GNSS transmitters. Our method autonomously extracts the RFF from the received signals by exploiting deep learning, which avoids the laborious manual feature selection process compared to other methods. The time-frequency representation of the signal is used as input to the deep learning. We evaluate Shorttime Fourier Transform (STFT) time-frequency representation method. We explore the possibility of using the Support Vector Data Description (SVDD) for GNSS spoofing detection. We evaluate two deep learning-based GNSS signal classification methods. One is RFF identification based on the original signal, namely IQ+CNN in this article, which preprocesses the collected IQ samples and directly inputs them into the deep learning model for training and classification. This method completely uses the deep learning model to learn the physical layer characteristics of wireless signal. The second is RFF identification based on two-dimensional representation of signals, namely STFT+RESNET50 in this article, which aims to extract RFF in the time-frequency domain. The experimental dataset is generated by software, and we compare the classification accuracy of the two methods at different SNRs. The experiments show that our method is reasonable for GNSS spoofing detection. In addition, the research of RFF-based GNSS spoofing detection is still in its infancy, and we promote the development of this field.