Reseach On Propagation Law And Immunization Strategy In Temporal Networks

The study on the propagation law, at the essential core of temporal network, lays a foundation for community discovering, web searching, virus spreading and network robustness and serves as a precondition for network immune strategy researches. Real networks mostly keep changing with time. The connection between network nodes is repeatedly intermittent, presenting typical burstiness. The time dimension makes the basic relations such as the path, accessibility and connectivity between network nodes more complicated and brings the new challenges to the study of propagation law and its relevant methodologies. At the same time, the huge size and the dynamic structure of modern networks makes it difficult to collect and process information for existing immune strategies that are hardly implemented with great speed and efficiency. Based on the existing studies of the temporal network burstiness and considering their mentioned pitfalls, we practiced a research centering temporal network propagation and immune strategies, initiating the potential impact of network evolution speed on propagation, examining the influence of individual activity on propagation and designing a novel immune strategy applicable to temporal network.The contributions of this dissertation to the researches of the temporal network propagation law and immunity strategy are as follows:First, an evolution model of the temporal networks with controllable evolution speed is proposed and the conclusion that the fast evolution of network facilitate the propagation is proved. The existing correlation coefficients have been redefined so that they are enabled to be applied to real temporal network and then used to define the network evolution speed. Meanwhile, a temporal network evolution model with non-markovian property is proposed, in which at each time step, every active node randomly select a node in the network with probability r and a node from the neighbors of the active node in the neighboring pre-snapshot with probability 1?r, and link them up. Then, the evolution speed of the generated temporal network can be controlled by the parameter r. The simulation results show that the fast evolution of the temporal networks facilitate the propagations.Second, a virus propagation model based on asynchronous interaction schema is proposed and the conclusion that individual activities burstiness suppresses virus spread is also proved. For the case that time intervals of node activation are subject to power-law distribution, the propagation threshold value of the model is deduced by using the renewal theory. The propagation threshold analysis and the experiment show that the stronger the heterogeneity of the time interval is, the larger the propagation threshold value is and the smaller the transmission scale and speed are. Above researches confirm a common conclusion that the heterogeneity of the node activation time interval including the individual activities burstiness curbs virus spread.Third, the immune models based on the transmission mechanism are put forward which have a significant impact on the virus propagation threshold value and the virus infection density. Given the problems of traditional immune models arising in collecting, analyzing network topology information in temporal network, an immune strategy based on transmission mechanism is proposed, in which the immune body spread over the network(similar to the spread of the virus program mechanism), the immune node is protected from being infected by the virus with a certain probability and transmits the immune body to its neighbor nodes. Based on this strategy, the basic immune transmission model and mass conservation model of immune transmission have been established. Their advantages are that they can be implemented quickly and need not collect and analyze the network topology information. Theory analytical and experimental results show that the immune propagation has significant effects on virus propagation threshold and the steady-state density of virus infection and can effectively restrain or eliminate virus spreadLast, the immune models based on random walk mechanism are established which have low immune particle density threshold and high immune effectiveness. Given network overhead problems brought by immune body transmission, an immune strategy based on random walk mechanism is put forward. Considering whether the random walkers exert impact upon each other when they move, the dependent random walk immune model and the P_independent random walk immune model is established respectively, in which the network transmission overhead of the immune particle is limited by a given immune particle density. Experiments show the two random walk immune models are characterized by their better immune effects with lower immune particle density and the network overhead when compared with acquaintances immune model. In addition, the comparative result with the target immune model depends on heterogeneity of network topology.