Network Traffic Classification Using WiFi Sensing

Abstract

With the ubiquity of WiFi-enabled devices, WiFi Channel State Information (CSI) based sensing of the physical environment has been researched broadly, and for network management and monitoring, advanced measures for Network Traffic Classification (NTC) have been called. This paper proposes a novel CSI-based NTC model using off-the-shelf WiFi sensing tools. We conducted experiments in both controlled environment and real-world environment. Experiment results have shown that the frequency-selective CSI signatures can be used to distinguish four common NTC classes: ping, music streaming, buffered video streaming, and live video streaming. CSI features for NTC include the number of prominent CSI amplitude bins, locations of bins and relevant prominence of bins on the amplitude histogram over time for different subcarriers. We conclude with a clear WiFi sensing-based distinction of different network types where it is observed that ping and music streaming have similarities in their features, while buffered and live video streaming resemble each other in their CSI amplitude features.