Modeling and evaluating targeted attacks in large scale dynamic systems

In this paper we consider the problem of targeted attacks in large scale peer-to-peer overlays. These attacks aimed at exhausting key resources of targeted hosts to diminish their capacity to provide or receive services. To defend the system against such attacks, we rely on clustering and implement induced churn to preserve randomness of nodes identifiers so that adversarial predictions are impossible. We propose robust join, leave, merge and split operations to discourage brute force denial of services and pollution attacks. We show that combining a small amount of randomization in the operations, and adequately tuning the sojourn time of peers in the same region of the overlay allows first to decrease the effect of targeted attacks at cluster level, and second to prevent pollution propagation in the whole overlay.