Reliable Spectrum Sensing and Physical Layer Security for Cognitive Radio Networks

Abstract

The need for spectrum is ever-growing as the use of data services is becoming pervasive. With the onset of new services like smart cities and internet of things (IoTs), and infotainment services in next generation of vehicles, the demand for data and thus for the limited spectrum is increasing. Wireless networks and services supported by wireless technology have been around for ages but because of the revolution in mobile computing brought by smart phones and other such devices the need and desire to stay connected 24/7 has put a unique demand on wireless networks. The services and the devices which need to be served are raising exponentially but the wireless spectrum is a physically limited spectrum. So, to meet all the demands the onus comes on managing the spectrum efficiently. There is a focus for the last decade or so to come up with techniques which can exploit the loopholes in the present management of the spectrum and also in some ways to radically shift the ways in which the spectrum is accessed. One of such approaches is cognitive radio network (CRN) which aims to exploit the underutilization of the allocated spectrum to fixed and dedicated nodes and services. CRN is a secondary network which has an unlicensed access the spectrum under certain conditions. CRN is faced with architectural as well as management issues because it has to meet not only the constraints of the primary network and the primary user (PU) but also it has to meet the service demands of the CR users. This dissertation focuses on two of the architectural issues, reliable spectrum sensing and physical layer security. The CR user has to first ascertain that the PU is absent before it can access the channel. This process is known as spectrum sensing. There are myriad of issues in spectrum sensing including the reliability of the spectrum sensing data and learning the changing behavior of the PU. These are addressed in the first part of the dissertation. In the second part of the dissertation the need for secure communication among the CR users is considered. As the CR users are mobile devices so they have limited computing power. The channel codes and encryption techniques used in conventional wireless communications cannot be used in a CRN because of the ad-hoc nature of the CRN and also because of exhaustive demands for computing power of the encryption techniques. Physical layer security which employs digital signal processing techniques to ensure secrecy is suitable for CRN because rather than taking exhaustive computational power it uses the features of channels for providing information secrecy. In the first part of the dissertation applying bioinformatics inspired techniques to be spectrum sensing is studied. String matching algorithms used in bioinformatics can be applied to scenarios in cognitive radios where reports of cooperative spectrum sensing nodes need to be compared with each other. Cooperative spectrum sensing is susceptible to security risks where malicious users who participate in the process falsify the spectrum sensing data, thus affecting cognitive radio network performance. In this work, an efficient spectrum sensing system is developed where each CR user senses the spectrum multiple times within an allocated sensing period. Each CR user quantizes its decision to predefined levels so as to achieve a trade-off between bandwidth utilization and decision reporting accuracy. The reports for all the CR users are compared at the fusion center using Smith-Waterman algorithm (SWA), an optimal algorithm for aligning biological sequences used in bioinformatics, and similarity indices are computed. Robust mean and robust deviation of the similarity indices are calculated and a threshold is determined by these values. The CR users who have similarity index below the given threshold are declared malicious and their reports are discarded. The local decisions of the remaining CR users are combined using the modified rules of decision combination to take a global decision. Simulation results show that our proposed scheme performs better than conventional schemes with and without malicious users. The study is extended for investigating optimal quantization schemes for spectrum sensing next. Cooperative spectrum sensing can be made more reliable by excluding the reports of unreliable CR users from the final decision combination at the fusion center (FC). Hard decision combination provides bandwidth efficiency but the results produced are unreliable while on the other hand soft decision combination has better results but at the expense of much consumption of bandwidth. If instead of hard decision or soft decision combination, quantized information is sent by the CR users to the FC, an acceptable trade-off is achieved. In this paper an optimal quantization scheme is proposed in which the local sensing information is quantized in such a way which ensures that the maximum detection probability is met while the false alarm probability remains under a certain constraint. The proposed optimal scheme works on the basis of energy detection and the local quantization thresholds are found through iterative search. A method inspired from bioinformatics, Smith-Waterman algorithm (SWA) is used to compare the local sensing reports of the CR users and on the basis of comparison similarity indexes are found for the CR users. On the basis of robust mean and robust standard deviation a threshold is calculated for the cooperative spectrum sensing. The local sensing decisions of the CR users below the calculated threshold are rejected and are not included in the final decision combination at the FC. As quantized information is used so the conventionally used rules of decision combination are modified to work on the quantized information and the FC combines the local sensing decisions of rest of the CR users through the modified rules. For detailed analysis, SWA-based rules of decision combination with optimal quantization thresholds are compared with a scheme that employs SWA-based rules of decision combination with heuristically selected quantization thresholds and a conventional majority combination scheme based on heuristically selected quantization thresholds. Simulation results show that the proposed scheme performs better than the other two schemes. After that a reliable spectrum sensing scheme is proposed, which uses K-nearest neighbor, a machine learning algorithm. In the training phase, each CR user produces a sensing report under varying conditions, and based on a global decision, either transmits or stays silent. In the training phase the local decisions of CR users are combined through a majority voting at the fusion center and a global decision is returned to each CR user. A CR user transmits or stays silent according to the global decision and at each CR user the global decision is compared to the actual primary user activity, which is ascertained through an acknowledgment signal. Based on this comparison, the sensing report is assigned to a sensing class. In the training phase enough information about the surrounding environment i.e. the activity of PU and the behavior of each CR to that activity is gathered. In the classification phase, each CR user compares its current sensing report to existing sensing classes, which are formed in the training phase, and distance vectors are calculated. Based on quantitative variables, the posterior probability of each sensing class is calculated and the sensing report is declared to represent either the absence or presence of a primary user at the CR user level. The quantitative variables used for calculating the posterior probability are calculated through K-nearest neighbor algorithm. These local decisions are then combined at the fusion center using a novel decision combination scheme, which takes into account the reliability of each CR user. The CR users then transmit or stay silent according to the global decision. Simulation results show that our proposed scheme outperforms conventional spectrum sensing schemes, both in fading and non-fading environments, where performance is evaluated using metrics such as the probability of detection, total probability of error and the ability to exploit data transmission opportunities. The first part of the dissertation is concluded by investigating a joint spectrum sensing and transmission framework. Actor-Critic algorithm is employed to get the optimal policy and value function and the algorithm is trained through all possible actions and states. In transmission the CR user is constrained by the residual energy. So, energy harvesting is used to harvest energy and then to transmit with transmission energy which meets the long term requirement of the CR user. Given a state which is composed of the remaining energy, the belief and the local and global spectrum decisions an action is selected on the basis of optimal value function and optimal policy function after the training phase is done with. Simulation result show the occurrence of each action selected which points to the probability of the occurrence of the particular state and action combination. The average rate achieved is also shown and is compared with an exhaustive search scheme which acts as the upper bound for the scheme. The second part of the dissertation deals with physical layer security. First, a physical layer-security scheme for an underlay relay-based CRN that uses OFDM as the medium access technique is proposed. Resource allocation in relay-aided CRNs becomes a hard problem especially if it is under security threat. Different from conventional relay-based OFDM schemes, in the paper we consider the relay network which has two dedicated relay nodes; One relay which is capable of subcarrier mapping forwards the received signal to the destination and the other sends a jamming signal to add noise to the signal received by the eavesdropper. Optimization is performed under a unified framework where power allocation at the source node, power allocation and subcarrier mapping in the relay network are optimized to maximize the secrecy rate of the CRN while satisfying the maximum transmission power constraints and the interference threshold of the PU. The power allocation problem at the forwarding relaying node is a non-convex optimization problem. Therefore, at first the optimization problem is simplified and a closed form solution is obtained which satisfies the maximum PU interference constraint. Afterwards, the optimization problem is solved for satisfying the maximum transmission power constraint. An algorithm is also proposed for subcarrier mapping at the forwarding relaying node. The proposed power allocation method and subcarrier mapping scheme have low complexity, compared to the baseline schemes. Finally, simulation results are provided for different parameters to show the performance improvement of the proposed scheme in terms of secrecy rate. Physical layer security is furthered explored by proposing a physical layer security-based scheme for an underlay CRN that has energy-constrained relay nodes. In the scheme, the cooperative diversity of multiple relays is exploited to provide physical layer security against an eavesdropping attack. Different from conventional relay schemes, relay-based CRN faces other issues, such as the maximum interference–constraint with the PU, and takes into consideration leakage to eavesdroppers in case of an eavesdropping attack. For a CRN to be practical, the energy constraint should be taken into consideration because ad-hoc networks cannot have a fixed power supply all the time. If the nodes in a CRN are able to harvest energy and then spend less energy than the total energy available, we can ensure a perpetual lifetime for the network. In this paper, an energy-constrained CRN is considered where relay nodes are able to harvest energy. A cooperative, diversity-based relay and subchannel–selection algorithm is proposed, which selects a relay and a subchannel to achieve the maximum secrecy rate while keeping the energy consumed under a certain limit. A transmission power factor is also selected by the algorithm, which ensures long-term operation of the network. The power allocation problem at the selected relay and at the source also satisfies the maximum-interference constraint with the PU. The proposed scheme is compared with a variant of the proposed scheme where the relays are assumed to have an infinite battery capacity (so maximum transmission power is available in every time slot), and is compared with a scheme that uses jamming for physical layer security. The simulation results show that the proposed scheme closely follows the infinite battery capacity scheme, which works as the upper bound for the proposed scheme. The infinite battery–capacity scheme outperforms the jamming-based physical layer security scheme, thus validating that cooperative diversity–based schemes are suitable to use when channel conditions are better employed, instead of jamming for physical layer security.