Interdependent Security with Strategic Agents and Cascades of Infection

Abstract: We investigate cascades in networks consisting of strategic agents with interdependent security. We assume that the strategic agents have choices between i) investing in protecting themselves, ii) purchasing insurance to transfer (some) risks, and iii) taking no actions. Using a population game model, we study how various system parameters, such as node degrees, infection propagation rate, and the probability with which infected nodes transmit infection to neighbors, affect nodes' choices at Nash equilibria and the resultant price of anarchy/stability. In addition, we examine how the probability that a single infected node can spread the infection to a significant portion of the entire network, called cascade probability, behaves with respect to system parameters. In particular, we demonstrate that, at least for some parameter regimes, the cascade probability increases with the average degree of nodes.