Abstract
Visible light communication (VLC) is a promising technology with a high data rate that can supplement radio frequency communication. Although VLC systems have a natural advantage of high security due to the line-of-sight light propagation characteristic, they are still vulnerable when facing an open environment. Device fingerprinting is a technique that is widely viewed to detect transmitter impersonation attack in radio frequency (RF) based wireless systems. In this paper, we introduce the fingerprinting technique to discriminate illegal transmitting devices in VLC systems. We first investigate the hardware imperfections of the VLC transmitter, which can provide a unique device ID. Then we implement a feature separation network for transmitter fingerprinting (TF-FSN) and design a two-stage training strategy to obtain a stable classifier. Finally, we experimentally demonstrate the feasibility and performance of the proposed method. The results show that the accuracy of identification and verification is 92.65% and 98%, respectively. Moreover, our method is robust over different distances and a wide range of signal-to-noise ratios.
Highlights
W ITH the rapid growth of mobile devices, new technologies are currently being studied to meet the increasing demand for high data rate and low-latency communication
We propose a new mechanism for the physical layer security in the Visible light communication (VLC) system
Considering that the performance can be improved by using multiple slices, we can make flexible configurations depending on the number of classes and the overall recognition rate in practice
Summary
W ITH the rapid growth of mobile devices, new technologies are currently being studied to meet the increasing demand for high data rate and low-latency communication. Radio frequency fingerprinting (RFF) is a new security mechanism that aims to identify transmitters by extracting hardware-based features present in the signals [17], [18]. These unique hardware characteristics are induced by nonlinearities and imperfections in the manufacturing process of emitter circuitry [19]. DL provides a more flexible and efficient method than hand-crafted feature-based process in RFF, which traditionally requires sufficient knowledge about the signal types and characteristics [27] These manually extracted features are not general, they are not robust to devices and components variation. We adopt a statistical method with random slicing to improve the identification performance of our proposed method when facing complex environments
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
