In this paper, we expose a class of stealthy attacks, which we coin pinball attacks, that exploit the adaptive nature of Dynamic Channel Assignment algorithms in wireless networks. In a pinball attack, the attacker strategically induces interference in the neighborhood of certain nodes to cause a cascading channel switching behavior, thereby exploiting the topology of the network to his own advantage. The attacker’s problem is formulated as a Markov Decision Process in which the attacker seeks to maximize his long-term net rewards. To overcome the exponential size of the state space, we derive suboptimal attack policies through approximation methods. We expose the pinball attacks on two general setups that arise practically in wireless communication systems. In the first setup, each wireless node is an access point aiming to select a channel that has minimum interference with its neighbors. In the second setup, we consider wireless nodes with multiple antennas whereby every node should communicate with each of its neighbors on a separate channel. We formally establish the equivalence of both setups by showing that the identified pinball attack policies on one can be directly mapped to attack policies on the other. We present numerical results to assess the impact of the obtained pinball attack policies under various topologies, and show that they outperform other known attack policies over all topologies considered. We also propose a novel approach to scale up the attack to larger size networks with special structures.
Read full abstract