Passive Side-Channel Interference Estimation for WiFi Networks

WiFi devices operating in the 2.4 GHz ISM band significantly suffer from the interference caused by overlapping side-channels. This paper proposes a metric to quantify the Quality of Experience (QoE) impact of side-channel interference based on passive, granular wireless driver parameter samples from consumer grade WiFi Access Points (APs). The variables of the metric are BadPLCP, NoPkt, and Glitch which are driver parameters determined in home environment experiments to be highly correlated to interference. Filed analysis of the proposed metric in 802.11ax and 802.11n deployments indicates that side-channel interference prevents medium access and causes frame losses, which are the main drivers of lowered QoE. Further, this paper proposes a Deep Neural Network (DNN) based algorithm to detect the interfering side-channel neighbour, where the wireless driver parameters are the input features. For 6 classes representing interference from different side-channels, we achieve 78% accuracy, 83% precision, 78% recall, and 79% f1-score.