Monitoring Multi-Hop Multi-Channel Wireless Networks: Online Sniffer Channel Assignment

Abstract

Data capture is important for some critical network applications, such as network diagnosis and criminal investigation. In multi-channel wireless networks, the fundamental challenge for data capture is how to assign operation channels to wireless sniffers. The existing approaches make some impractical assumptions, such as the prior knowledge on network traffic and the perfect conditions of data capture. In this paper, we relax these assumptions and investigate the sniffer-channel assignment problem in multi-hop scenarios. Especially, sniffer redundancy deployment is discussed, which enables multiple sniffers to monitor one traffic. This problem is formulated as a combinatorial multi-arm bandit (MAB) problem, and a cooperative distribute learning policy is proposed. We analyze the regret of our policy in theory, and validate its effectiveness through numerical simulations.