Abstract:

How to analyze user information and summarize user behavior from large-scale network traffic is now a key technology in the Internet era. This paper makes a full investigation on the existing research results, summarizes the advantages and disadvantages of the previous methods in user classification and traffic analysis and analyzes the potential security risks in the cloud system,finally proposes a multi-scale user classification technology based on cloud system. Through analyze user behavior from the IP session and some other information, our method puts forward the classification label accordingly, and extracting effective classification features, last, we classify the user by machine learning methods combined with statistic. This method can extract valuable information from network traffic, and use the information to distinguish users, generalize user behavior in traffic.

Key words: user identification, traffic analysis, user classification, data mining, cloud computing