Enhancing spectrum sensing performance for the cognitive radio based internet of things
Hossain, Mohammad Amzad
Hossain, Mohammad Amzad
Loading...
Publication Date
2021-12-26
Type
Thesis
Downloads
Citation
Abstract
In cognitive radio based internet of things (CR-IoT) networks spectrum sensing and decision making processes to determine whether the primary user (PU) signal is present or absent in the network are very important and vital issues to the utilisation of the idle licensed spectrum. Spectrum sensing is at the heart of CR-IoT networks to minimize the interference between the PU signal and the CR-IoT user signal. Within that context, this thesis makes the following contributions: Firstly, I proposed the concept of multiple reporting channels (MRC) for cluster-based cooperative spectrum sensing (CSS) for CR-IoT networks to better utilize the reporting time slot by extending the sensing time of CR-IoT users. A multiple reporting channels concept is proposed based on frequency division multiple access to enhance the spectrum sensing performance and reduce the reporting time delay of all cluster heads (CHs). This approach significantly enhances the sensing time for all CR-IoT users than the non-sequential as well as minimize the reporting time delay of all CHs than sequential single channel reporting approach. These two features of our proposed approach increase the decision accuracy of the fusion centre (FC) more than the conventional approach. Simulation results prove that my proposed approach significantly enhances the sensing accuracy and mitigate the reporting time delay of CH compared to the conventional approach. Secondly, I proposed a novel energy efficient sequential energy detection (ED) spectrum sensing technique which enhances the sensing duration of each unlicensed CR-IoT user by utilizing the reporting time slot when compared to the non-sequential conventional ED spectrum sensing scheme. In addition, each unlicensed CR-IoT user calculates the weight factor based on the Kullback Leibler divergence (KLD) score, which enhances the detection performance and sum rate. The simulation results indicate that the my proposed sequential ED spectrum sensing scheme achieves a better sensing gain, an increased sum rate, an enhanced energy and spectral efficiency when compared to the non-sequential conventional ED spectrum sensing scheme with interference constraints. Thirdly, I introduced a novel multi-user multiple-input and multiple-output (MU-MIMO) antennas aided cluster based cooperative spectrum sensing (CB-CSS) scheme for cognitive radio based internet of vehicles (CR-IoV) networks. In this proposed scheme, each CR embedded vehicles (CRV) sends sensing data to the cluster head which makes a cluster decision by using the soft data fusion rule like the equal gain combining (EGC) fusion rule and the maximal ratio combining (MRC) fusion rule; whereas the FC makes a final global decision by using the K-out-of-N rule to identify the presence of the PU signal. Simulation results show that the proposed MU-MIMO antennas aided CB-CSS scheme achieves a better sensing gain, enhanced the sum rate and lower global error probability when compared to both the conventional single-input and single-output (SISO) antenna based CSS and non-cooperative spectrum sensing (NCSS) schemes. In addition, the proposed scheme achieves a lower traffic overhead when compared to the MU-MIMO based CSS scheme without the cluster. Finally, I introduced a machine learning (ML)-based secure cooperative spectrum sensing techniques for CR-IoT networks. In the proposed scheme, we use the Support Vector Machines (SVM), K-Nearest Neighbors (KNN) and Naive Bayes (NB) machine learning algorithms to classify the legitimate CR-IoT users and three types of malicious users (MUs), such as (i) Always Low Energy Malicious Users (ALEMUs), (i) Always High Energy Malicious Users (AHEMUs), and (i) Random Energy Malicious Users (REMUs). In this thesis, I use majority fusion rule at the intelligent fusion centre (IFC) on the sensing results of legitimate CR-IoT users to make the global decision about presence or absence of the primary user (PU) in the networks. My proposed scheme is significantly improved the sensing performance of knowing the activity of PUs for three kinds of malicious users (i.e., ALEMUs, AHEMUs, and REMUs) cases than the conventional(without security technique) scheme. In addition, it decreases the global error probability at the IFC over the conventional scheme.
Publisher
NUI Galway