Stealthy edge decoy attacks against dynamic channel assignment in wireless networks

Abstract

Dynamic channel assignment algorithms allow wire- less nodes to select their communication channels based on the state of the network to reduce interference between the nodes and improve the overall network performance. They have been shown to outperform static channel assignment policies and thus are playing a critical role in Software Defined Networks (SDNs). In this paper, we examine the security of dynamic channel assignment algorithms against stealthy decoy attacks in which the attacker induces channel conflicts on a selective set of edges. When an edge is attacked, the victim nodes search for and switch to a different channel to use, possibly causing their neighbors to switch as well. As the effect of the attack propagates through the network, the performance of the network is severely degraded. The decisions of which edges to attack and what channels to use in creating conflicts are the solutions of various Markov Decision Processes (MDP) problems in which the attacker's goal is to maximize the number of conflicts in the network subject to an attack cost. We apply approximate policy iteration methods to identify suboptimal, albeit efficient, attack policies. Our results show that our exposed attack policy outperforms other attack policies and adapts to the cost of the attack