Abstract

Every peer-to-peer system is based on some overlay network connecting its peers. Many of the overlay network concepts proposed in the scientific community are based on the concept of virtual space. These designs are usually highly scalable, but they do not guarantee robustness against adversarial attacks, especially when considering open peer-to-peer systems. In these systems, determined adversaries may start both insider and outsider attacks in order to harm the overlay network as much as this is possible. We will focus on insider attacks in which the adversarial peers in the network perform join-leave attacks, and we will consider outsider attacks in which an adversary can perform a denial-of-service attack against any of the honest peers in the network. Strategies have been proposed that can defend an overlay network against even massive join-leave attacks, and strategies are also known that can defend an overlay network against limited denial-of-service attacks. However, none of these can protect an overlay network against combinations of these attacks. We illustrate in this paper why it is not easy to design strategies against these attacks and propose join and leave protocols for overlay networks based on the concept of virtual space that can make them provably robust against these attacks without creating too much overhead.