Deconstructing Interference Relations in WiFi Networks

TLDR: This work model the 802.11 MAC as a Hidden Markov Model (HMM), and uses a machine learning approach to learn the state transition probabilities in this model using the observed trace, and coupled with an estimation of collision probabilities helps to deduce the interference relationships.

Abstract: Wireless interference is the major cause of degradation of capacity in 802.11 wireless networks. We present an approach to estimate the interference between nodes and links in a live wireless network by passive monitoring of wireless traffic. This does not require any controlled experiments, injection of probe traffic in the network, or even access to the network nodes. Our approach requires deploying multiple sniffers across the network to capture wireless traffic traces. These traces are then analyzed to infer the interference relations between nodes and links. We model the 802.11 MAC as a Hidden Markov Model (HMM), and use a machine learning approach to learn the state transition probabilities in this model using the observed trace. This coupled with an estimation of collision probabilities helps us to deduce the interference relationships. We show the effectiveness of this method against simpler heuristics, and also a profiling-based method that requires active measurements. Experimental results demonstrate that the proposed approach is significantly more accurate than heuristics and quite competitive with active measurements. We also validate the approach in a real WLAN environment.