Passive Detection of Fat Users in WiFi Networks using Thompson Sampling

The user experience in large WiFi networks, such as those on university campuses, can be greatly enhanced by monitoring user activity. Early detection of high-bandwidth users, often called fat users, who consume substantial radio resources, can help prevent the formation of network hotspots. This enables both short-term measures, such as traffic shaping, and long-term solutions, like deploying additional access points. In this paper, we present a fully passive method for detecting fat users in WiFi networks that requires no modifications to existing WiFi infrastructure. Our approach employs multiple WiFi interfaces to passively capture WiFi traffic over the air, with channel hopping managed intelligently through a Multi-Armed Bandit framework using a modified Thompson Sampling algorithm. This technique allows effective monitoring of a broad radio spectrum, even with a limited number of interfaces. Simulation results and real-world experiments demonstrate the feasibility of this approach in both stationary and non-stationary network environments.