CSI-Based NTC Using Ambient WiFi: Channel Selection, Topology Control and Traffic Interference

The ubiquity of WiFi-enabled devices raises the need for advanced network monitoring and management due to security and privacy issues associated with wireless networks. One method is Network Traffic Classification (NTC). However, robust and resilient NTC when traffic is encrypted and without compromising privacy is challenging. One possibility is to use WiFi Channel State Information (CSI), as it will change depending on the characteristics of the information being transmitted. In this paper, we show that it is possible to use CSI for NTC by extracting CSI amplitudes from different network traffic streams, and then using this information to create a feature set that can be used with machine learning classifiers to develop a novel NTC mechanism. We show the robustness of our CSI-based NTC by considering real-world scenarios under different wireless interference, namely overlapping frequencies, location-based interference, and interference generated by various network streams. Consequently, our proposed NTC scheme achieved 0.84 NTC F-score as a baseline, and we identify that spectrally overlapping interference reduces the overall F-score of the CSI-based NTC classifier by upto 0.6. For traffic classes with similar characteristics, the proposed framework achieved an NTC F-score above 0.95, corroborating the scalability of our non-intrusive sensing technology.