Cooperative spectrum sensing for cognitive radio networks

Radio spectrum is a very scarce and important resource for wireless communication systems. However, a recent study conducted by Federal Communications Commission (FCC) found that most of the currently allocated radio spectrum is not efficiently utilized by the licensed primary users. Granting opportunistic access of the spectrum to unlicensed secondary users has been suggested as a possible way to improve the utilization of the radio spectrum. Cognitive Radio (CR) is an emerging technology that would allow an unlicensed (cognitive) radio to sense and efficiently use any available spectrum at a given time. Reliable detection of the primary users is an important task for CR systems. Cooperation among a few sensors can offer significant gains in the performance of the CR spectrum sensing system by countering shadow-fading effects. In this thesis, we consider a parallel fusion based cooperative sensing network, in which the sensors send their sensing information to an access point, which makes the final decision regarding presence or absence of the primary signal. We assume that energy detection is used at each sensor. Presence of few malicious users sending false sensing data can severely degrade the performance of such a cooperative sensing system. In this thesis, we investigate schemes to identify malicious users based on outlier detection techniques. We take into consideration constraints imposed by the CR scenario, such as limited information about the primary signal propagation environment and small sensing data sample size. Considering partial knowledge of the primary user activity, we propose a novel method to identify malicious users. We further propose malicious user detection schemes that take into consideration the spatial location of the sensors. We then investigate efficient sensor allocation and quantization techniques for a CR network operating in multiple primary bands. We explore different methods to assign CR sensors to various primary bands. We then study efficient single-bit quantization schemes at the sensors. We show that the optimal quantization scheme is, in general, non-convex and propose a suboptimal solution based on a convex restriction of the original problem. We compare the performance of the proposed schemes using simulations.Applied Science, Faculty ofElectrical and Computer Engineering, Department ofGraduat