Research On The Network Application Protocol Identification

With the rapid development of mobile Internet, the rapid increase in the size of network applications bring convenience to network users and also brought problems, such as traffic congestion and increase the difficulty of monitoring network. Network operators and regulatory authorities the need to strengthen the regulatory capacity of network traffic, the ability to classify network application protocol accurately and efficiently is the key point to strengthen the regulatory capacity of network traffic.The paper analysis three protocol identification methods combined with the current situation, studies the problem of protocol feature extraction and proposes a feature extraction algorithm for application layer protocol based on prefixSpan algorithm. It adds the constraint conditions to reduces number of projection database. Experimental results show that compared with the original algorithm, the running time of this algorithm is shorter. In addition, this paper also studies protocol identification methods, proposes protocol recognition method based on DPI and machine learning, using the identification result of BP neural network algorithm to detect the identification by DPI. Experimental results show that this protocol identification method can improve the recognition accuracy.