Abstract
Several email virus models have been proposed in previous papers. Usually, users in these models are homogeneous and symmetric which fail to model the real situation accurately. In this paper, a novel model with local group is presented. In real email system, users will usually group their contacts into various kinds of groups. Moreover, users have different levels of security awareness according to different closeness with certain group. To model the condition, we propose the propagation model with local group. In our model, a user is modeled as an autonomy agent who is able to deal with email and group its ego network. Based on the local group algorithm, open probability is carefully defined overcoming the disadvantage of homogeneous model. Applying our model on benchmark networks, we use the space-time diagrams for a small network and curve graph for real email network to verify our model. Simulations indicate that users will not be infected until his close friend group mostly being infected, showing that the group provides buffer for virus spreading. On the other hand, in accordance with the local group action in reality, the propagation is largely deferred compared with the homogeneous model.