| Summary: | 碩士 === 國立交通大學 === 網路工程研究所 === 103 === In the past decade, Internet has been widely used in everyday life. Different types of mobile broadband applications are created and require ever-increasing network resources. However, Internet service providers (ISPs) must make good use of these limited resources to provide users with different levels of Quality-of-Service (QoS) services. The first step towards traffic engineering (TE), is to make a traffic classification.
In this thesis proposes a classification method to recognize what mobile application is executed at an early stage. We firstly collected traffic traces from the WiFi access point and then developed a Hidden Markov Model (HMM) based on packet size and transmission direction of the first 20 packets.
In a series of our experiments, we evaluated the number of hidden states by 10-fold cross validation, and classified six different types of mobile applications. The accuracy of our proposed method achieves 99.17%. In addition, we set different threshold values for different application models and identified 91.33% of flows in the testing dataset which is added unknown traffic flows. These experimental results demonstrate that our proposed method is effective to classify Internet flows as well as unknown traffic in real network.
|
|---|
