The structure analysis of user behaviors for web traffic

Understanding the structure and dynamics of the behavior networks for Web traffic that connect users with servers across the Internet is a key to modeling the network and designing future application. The Web-visited bipartite networks, called the user behavioral networks, display a natural bipartite structure: two kinds of nodes coexist with links only between nodes of different kinds. We obtained the result that the out-degree distribution of clients (the host initiating the connection), the in-degree distribution of servers (the host receiving the connection) and the strength distribution (the exchange bytes between clients and servers) is power-law, whose exponential is between 1.7 and 3.4. The clustering coefficient of clients and servers was larger than larger than in randomized, degree preserving versions of the same graph. At last, based on the algorithm of finding the community structure in bipartite network, we divided the clients into different interest communities. The structure analysis of user behavior for Web traffic is very helpful for the network management, resource allocation, traffic engineering and security.