Survivability Against Intelligent Adversary in Next-Generation Wireless Networks
Computer Science and Engineering
AltmetricsView Usage Statistics
The conventional static spectrum allocation policy has resulted in a suboptimal use of spectrum resources, leading to over-utilization in some bands and under-utilization in others. As a solution, dynamic spectrum access-based Cognitive Radio Network (CRN) has been proposed. CRN allows secondary users (SUs) to use an unused licensed spectrum while the proprietary primary user (PU) is not transmitting. CRN being a next generation wireless network inherits all the challenges of wireless and brings some critical issues due to the dynamic spectrum sensing and acquirement. Multiple SUs compete for spectrum and create conflicts and collisions in spectrum acquirement. An adversary can intelligently exploit these vulnerabilities to disrupt the communications of legitimate SUs. In this research, we address these unique challenges in the battle for coexistence. We first present a framework for dynamic spectrum allocation with aggregation and fragmentation. Fitting the spectrum requirement of multiple SUs is an NP-hard problem. We propose three different techniques for spectrum allocation optimization: centralized, decentralized, and another hybrid solution with leader election. In this battle for coexistence, the broadcasting and open nature of transmission leaves a CRN open to jamming based Denial of Service (DoS) attacks. Since SUs use different channels for communication and attacker is also capable of attacking one channel, an intelligent attacker has to choose the DoS target by sensing all possible channels. We propose CR-Honeynet, a framework that exploits the intelligence of an attacker and lures it to a decoy transmission, while other legitimate communications bypass attacks. However, selecting a node to act as decoy degrades its performance. We propose state-based decoy selection strategies that select decoys dynamically based on optimization criteria that deals with queue length, service type and arrival rate. and optimizes the overall end-to-end system performance. We then model the battle of defender and attacker from a game-theoretic point of view, where both parties are intelligent and learn about heterogeneous channel utilities from history. The theoretical model showes that there exists a Nash equilibrium for learning period of both players and if they deviate, the other party wins. We extend our study to attackers with the capability of moving in all three directions where adaptive beamnulling is useful to filter the signal coming from a jammer spatially. We propose a tracking based framework to optimize the beamnull that minimizes the risk of attack while minimizing link failure. Last but not that leastFinally, we develop a state-of-the-art testbed with off-the-shelf devices to evaluate the performance of the proposed framework.