HARDENING CYBERSECURITY FOR 5G AND BEYOND WIRELESS NETWORKS AT PHYSICAL LAYER

Abstract

One of the main challenges in 5G networks is to ensure secure and reliable communication service. An efficient and lightweight security mechanism is desired in the design. In this dissertation, we focus on the physical layer secret key generation and primary user (PU) location privacy protection in 5G wireless networks. Physical layer key generation is a promising secure mechanism, which does not rely on the hardness of computational solutions but on the information-theoretical security to generate symmetric secret keys. It thus induces low computational complexity and fit for the resource-constrained IoT devices. For instance, two entities (Alice and Bob) can perform physical layer (PHY) key generation based on the reciprocal channels to extract identical secret bits. To improve the secrecy of pair-wise physical layer key generation, we propose the power delay profile (PDP) as the common randomness for secret key generation in this dissertation proposal. Unlike the channel statistic information (RSS) and received signal strength (RSS) adopted in existing works, PDP can resolve the path gain for each multipath and flatten the channel noise. Secret beampattern configuration at the legitimate devices can directly decorrelate attackers' channel observations. This strategy is game-changing for the security of PHY key generation because it establishes an extra shell to prevent the passive attacker. We propose to utilize low-correlated beampatterns to collect non-correlated channel measurements in channel coherence time in order to increase the channel probing rate. For instance, K channel probings with low-correlated beampatterns in coherence time can produce the channel measurements K times more than the existing works and thus enhance the key generation rate. To reduce temporal correlation among channel measurements, we propose a cosine similarity-based metric to evaluate the inter-correlation of beampatterns. Based on this metric, the channel measurements with the least inter-correlation can be selected using the proposed algorithm. In this dissertation, we also investigate the scheme to improve the group secret key generation efficiency in 5G mmWave Massive MIMO networks by enhancing the efficiency of channel probing for group key generation. A new channel probing strategy for star-topology networks group key generation is proposed, which focuses on multiplexing of downlink probing signals to perform the downlink channel probing concurrently. The hybrid precoder has been considered in this scenario to mitigate inter-group interference, which includes an analog precoder and baseband precoder. To further balance the group key rates, a genetic algorithm (GA) based power allocation algorithm is developed to allocate more power to the nodes with unfavorable channel conditions. What’s more, we propose a scheme to estimate group key rates based on the maximum likelihood estimator (MLE) so that we can estimate the group key rates based on the probing samples. Various numerical results are provided including the group key rates and bits disagreement ratio (BDR). The numerical results show that the GA-based downlink channel probing scheme can increase the efficiency of channel probing and have higher group key rates compared with the existing channel probing schemes. When the SNR is 25dB, the key rates of GA-based power allocation scheme are 20% higher than the scheme with the conventional channel probing strategy. To protect the PU location privacy, we investigate the benefits of secondary user (SU) network beamforming on improving primary user (PU) location privacy in spectrum sharing systems in this dissertation proposal, where the beamformer in the SU network is designed to suppress the aggregate interference to improve the location privacy of PUs. We consider two problems: improving SU network communication throughput subject to the specified PU location privacy requirements and enhancing PU location privacy given the quality of service (QoS) requirements of SU networks. In the first problem, we provide an algorithm to achieve high data rates with the constrained PU location privacy level. Numerical results show that for a given PU location privacy requirement, the proposed scheme is able to interfere/exclude only a few SU nodes from the PU band and the network throughput can be improved by over 50\%. In the second problem, to fully explore the potential of SU network beamforming for enhancing PU location privacy, we propose a two-step scheme to decouple the beamforming and privacy zone design so that the PU location privacy can be improved while satisfying the SU network throughput requirement. According to numerical evaluations, the proposed scheme can maintain/achieve higher PU location privacy than the benchmark beamforming schemes while satisfying a QoS requirement for the SU network.