Impact of Habitual Behavior on Human Dynamics and Propagation Process

摘要

Computer viruses spread over network of contacts between computers. Current researches assume that the contact patterns between computers follow the Poisson distribution. But a series of recent measurements indicate that the inter-contact time of human actions follows power-law distributions with heavy tails. We propose a virus propagation model based on human dynamics by considering habitual behaviors, to study the fundamental propagation patterns characterizing a virus outbreak affected by human behaviors. We further find human behavior is the reason causing long prevalence time besides factors like network topologies and virus infection patterns, from the results that the virus spread process has a long stage of slow start due to the non-Poisson nature of human dynamics, and the length of the stage increases with extending of network scales. Moreover, the slow start of spread results in decay time several orders of magnitude larger than the Poisson-process-based predictions. Simulations show that the conclusions are consistent with the prediction by a university email record and the prevalence data from virus bulletins.