Identifying signal source using channel state information in wireless LANs

In this paper, we study the feasibility of identifying signal source in wireless LANs using channel state information obtained from preambles. In current standard, the signal source can be identified by reading the MAC header which requires high SNR and takes more time than receiving the preamble. For each packet, CSI is obtained for each (group of) subcarriers and for each TX-RX path, which makes the information rich in features to be used for machine learning-based classification. Experiments in a typical office environment show that with simple kNN or neural network models, we can classify tens of signal sources with over 90% accuracy. Moreover, confidence levels differ significantly between correct and incorrect samples, which can be exploited to avoid false positives with little sacrifice of correct samples. The CSI-based source identification method can be used to improve spectral efficiency of WLANs, and can be used along with schemes such as BSS coloring included in the IEEE 802.11ax standard for highly efficiency WLANs.