Data-and-Knowledge Dual-Driven Radio Frequency Fingerprint Identification
Zitong Zhang, Lu Yuan, Fuhui Zhou, Qihui Wu
Abstract
Wireless network security can be improved by radio frequency fingerprinting due to its stability and uniqueness. Although many radio frequency fingerprint identification (RFFI) methods based on deep learning have been proposed, they have low identification accuracy, especially at low signal-to-noise ratio. In order to overcome this drawback, a data-and-knowledge dual-driven RFFI scheme is proposed by utilizing a knowledge-driven multiscale attention convolutional network (AttMsCN). The protocol knowledge is exploited to provide more advanced semantics. Moreover, the AttMsCN is utilized to capture higher level features. Simulation results demonstrate that our designed scheme has the best performance than the representative schemes in the matter of convergence speed and identification accuracy.