LSTM Network Based Spoofing Detection and Recognition in a GNSS Receiver

摘要

The spoofing interference is significantly harmful for the applications based on GNSS technology. Aiming to realize the high-speed and high-accuracy anti-spoofing, a joint spoofing detection, recognition and elimination scheme is proposed based on the Long-Short-Time Memory (LSTM) neural network in this paper. First, by analyzing and utilizing the time-sequence characteristics of the spoofing attack, a spoofing detection and attack-mode recognition algorithm is designed by using a LSTM neural network. Second, two algorithms to identify the correlation peak based on the comparison of the peak-mean with the historical-mean and a LSTM network are designed for the static attack and the dynamic attack, respectively. Finally, the parameters of the spoofing peak given by the correlation peak identification module are input to a sub-space projection based signal eliminator to cancel the interference. Test results show that a higher accuracy for spoofing detection and identification could be obtained by the proposed solution and the spoofing signal could be efficiently suppressed. In addition, our scheme is executed during the signal acquisition stage. Therefore, it is beneficial to identify and eliminate interference earlier and more quickly.