Cellular phones are expanding their popularity in data service. However, we do not know much about the data traffic consumed in cellular networks. Traditional traffic classification studies have focused on grouping a set of related packets into the application categories. However, the proposed methods are hard to be applied to Korean cellular data traffic analysis because most data is from a single application: Wireless Application Protocol (WAP). As WAP conveys WML documents, we have to inspect the WML documents included in a traffic flow for more detailed classification. In this paper, we propose a method that finds Korean words carried in WML documents and classifies the documents into appropriate content categories according to the their meaning. We implemented and evaluated simple classification heuristics using packet traces from CDMA 1x EVDO/WCDMA HSDPA networks. From the result, we show that a simple heuristic is effective for classifying WAP flows to their content categories.
Download Full PDF Version (Non-Commercial Use)