The Research And Evolutionary Analysis Of The Centrality Measurements For AS-Level Internet Topology

The Internet grows rapidly and becomes an indispensible part of human society. The whole Internet could face massive failure when a few "Central Nodes" went wrong or were paralyzed. The Research on the centralities of AS-level Internet macro topology is to find the Central Nodes in huge and complex network structure of Internet precisely and rapidly, and protect them effectively so that the deliberate attacks could be stopped. It is a huge issue and a great affecting factor to the design of next generation Internet and the Internet security defense.The data used in this paper is based on the massive AS-level Internet data provided by CAIDA project. Firstly, the measurement issue of AS-level Internet topology centrality is studied using three kinds of widely applied centrality indices:Degree Centrality, Closeness Centrality, and Betweenness centrality. Meanwhile, a Coreness Centrality for measuring the nodes with high Coreness is proposed. Specific to Internet deliberate attacks, a node deletion method is used to measure the importance of nodes in the network by deleting them to see how much damage the deletion done to the connectivity of the network. As a result of the research, the Degree Centrality and Betweenness Centrality are better in measuring AS-level Internet than the ClosenessCentrality indice; Degree Centrality and Betweenness Centrality heve strong similarity when attacking ratio is lower than0.5%; Coreness Centrality measurement is fit for finding the communities in Internet made up of the nodes with higher degrees and tighter connections.Secondly, An analysis on the evolution of the central nodes of AS-level Internet. The metabolism mechanism is found in the central nodes evolution process, and the process is fit in three main kinds:surging type, growing type, and stationary type, In the central nodes evolution process, the centrality indice increasing is mainly cause by the tight connections link to the existing nodes. In the edge creation process of central nodes evolution, the three types of central nodes dependency on neighbor nodes are different. Surging type central nodes have the mode dependency on med-core nodes, and then the high-core nodes, the last are the low-core nodes; while the dependency order of growing nodes is high-core nodes, med-core nodes and low-core nodes; and the dependency order of stationary nodes is low-core nodes, med-core nodes and high-core nodes. The edge connections of central nodes are different.At last, the distribution characteristics of central nodes in the network communities are analyzed. The central nodes are mainly appear in the communities with larger scale. And the central nodes play important roles in the information transferring among the communities.