Study On Quantization Methods Of Physical Layer Secret Key Generation

Key generation is a novel security technology in wireless communication,the core idea is to generate secret key with reciprocity,time variation and spatial variation of wireless channels,and the generated keys are naturally random.The new technology overcomes the inherent limitations of traditional encryption methods,it does not require the existence of a trusted third party,and is not based on difficult mathematical problems in cryptography.Key generation can achieve one-time pad in theory,so that the absolute security in Shannon's secure communication theory can be obtained.In the era of 5G and Internet of Things,physical layer key generation technology can complement or replace traditional encryption methods,so that the technology has great prospects.This paper focuses on the quantization process in key generation.First,the background and significance of key generation technology are introduced,and the related research is summarized.Then the basic principle of key generation is explained in detail,including theoretical basis,specific steps and evaluation index.Finally,the vector quantization(VQ)algorithm and the quantization algorithms based on K-means clustering are proposed,which can significantly improve the performance of generated keys.The workload and innovation of this paper are as follows.(1)The VQ algorithm is proposed.Traditional quantization algorithms can be divided into amplitude quantization(AQ)and phase quantization(PQ).However,both of two traditional quantization can only use part of channel information,and cannot achieve the optimal performance.In this paper,both AQ and PQ are analyzed in detail,and their quantization error probabilities are deduced in theory.Then,based on AQ and PQ,VQ algorithm is proposed,which can utilize both amplitude information and phase information,and the quantization error probability of VQ is also deduced.We prove the fact that VQ can achieve better performance than traditional AQ and PQ through formula derivation.Then the computational complexity of VQ and two traditional quantization is analyzed and compared.Finally,the error probabilities of three algorithms are simulated,which proves the superiority of VQ.(2)The quantization algorithms based on K-means clustering are proposed.Inspired by VQ,this paper applies K-means clustering algorithm into key generation.We propose a basic K-means quantization(BKQ)algorithm at first,which clusters the real and imaginary parts of channel coefficients,and completes quantization by transmitting the obtained clustering centers.Then,two improved algorithms based on BKQ are proposed,including the lossy K-means quantization(LKQ)algorithm and the compensation K-means quantization(CKQ)algorithm.The LKQ applies the idea of protection interval,which discardes the samples near the quantization boundaries to reduce the quantization error rate.And the CKQ uses the compensation technology,which moves samples toward to clustering centers,it can achieve less error rate and utilize all the samples at the same time.This paper deduces the complexity of three K-means quantization algorithms,and verifies the performance of them.The simulation results show that the K-means based algorithms are much better than traditional quantization algorithms.