Agent incentives of strategic behavior in resource exchange

In this paper, we focus on the resource exchange over networks with autonomous participants (or agents), which goes beyond the peer-to-peer (P2P) bandwidth sharing idea. In such a resource exchange system, participants act as both suppliers and consumers of resources. Each agent obtains the utility by exchanging its resources with its neighbors according to the preset rules. However, agents may play strategically to improve their utilities by influencing the allocation, since the allocation find depends on what they submit. We consider a tit-for-tat popular proportional response dynamics and discuss the sybil attack strategy, which is a grave threat in P2P system. We are interested in the robustness of the proportional response dynamics in withstanding such a strategy. In this work, we prove that when the underlying network of a resource exchange system is a tree, any unilateral sybil attack could gain no more than twice as much utility.