Abstract: Real-time fine-grained traffic monitoring for mobile user is more and more important in wireless networks. However, the high granularity sampling of wireless user’s network traffic degrades system performance, while the transmission and storage of sampled data increase system overhead. In this paper, it is observed that wireless network traffic is high time and space correlated by using information entropy to measure a large-scale real data in campus WiFi networks. This paper presents a new traffic sampling scheme based on wireless user’s traffic spatial-temporal correlation to reduce network sampling frequencies of wireless user’s network traffic and ensure high accuracy. The experiment result proves that this scheme can reach more than 80% accuracy of sampling wireless user’s network traffic by using only less than 20% of the sampling rate.

Key words: user-level traffic monitoring, wireless network, spatial-temporal correlation, information entropy