中文摘要：

在基于信道特征的密钥生成方案中,用作随机信源的信道特征序列样值之间往往存在着相关性,这会使得最终提取的密钥强度较弱。为了提高生成密钥的强度,相关论文[4-5]引入Karhunen-Loève（K-L）变换对信道特征序列进行去相关处理。然而,通过大量仿真实验,我们发现K-L变换在其实现过程中存在敏感性问题,即通信两端独立地实现K-L变换而不通过公开信道交互协商信息时,两端生成密钥的不一致率非常高。针对这一问题,本文提出利用格型白化滤波器代替K-L变换实现去相关处理。仿真结果表明,采用格型白化滤波器用于去相关处理时无需通过公开信道交换任何信息。

英文摘要：

In the secret key establishment using channel characteristics,elements of the measured channel vector are always correlated,which will result in low entropy of the extracted key. Relevant researches has employed Karhunen-Loève transform to decorrelate the measured channel vector in order to furtherly generate secret key with a higher entropy. However,through extensive simulations,we find that the implementation of Karhunen-Loève transform is very sensitive. When Karhunen-Loève transform is independently implemented at two communicating nodes,the bit disagreement ratio of the generated key at two nodes is unusually high. In this paper,we propose to apply lattice whitening filter to do decorrelation instead of Karhunen-Loève transform. Simulation results shows that the decorrelation process with lattice whitening filter can be independently implemented at two nodes without transmission of any information through the public channel.