Resilient and multi-dimensional cooperative spectrum sensing on cognitive radio networks

While great strides have been made in spectrum sensing techniques in cognitive radio networks, these approaches are susceptible to unconventional attacks that may result in catastrophic performance degradation of the spectrum usage efficiency. For example, primary user emulation, intelligent jamming and denial of service for spectrum usage may impact the performance of classical spectrum sensing approaches. To address these challenges, this paper proposes a multi-dimensional cooperative sensing framework that can flexibly incorporate a variety of physical layer features to identify cases related to malicious behavior and genuine node failures. Though our approach is distributed, it is resilient in the sense that it does not simply rely on majority voting by a collection of nearby nodes. The key contributions of this paper are as follows: (i) A multiple criteria analysis technique and a non-parametric Bayesian inference method are formulated for identifying the spectrum holes that are least susceptible to malicious activity and failures, and (ii) Using real traces from the CRAWDAD data repository, we test our framework in a variety of practical settings, to prove the performance benefit of our approach.