Network Elicitation in Adversarial Environment

We study a problem of a defender who wants to protect a network against contagious attack by an intelligent adversary. The defender could only protect a fixed number of nodes and does not know the network. Each of the nodes in the network does not know the network either, but knows his/her neighbours only. We propose an incentive compatible mechanism allowing the defender to elicit information about the whole network. The mechanism is efficient in the sense that under truthful reports it assigns the protection optimally.