Vector partitioning quantization utilizing K-means clustering for physical layer secret key generation

Abstract

Most existing key generation schemes cannot achieve the optimal performance because the quantization algorithm cannot make full use of channel information, some schemes use amplitude quantization and others use phase quantization. This paper proposes the concept of vector partitioning quantization (VPQ), and several new quantization algorithms are obtained further, which can utilize amplitude and phase at the same time. First, the traditional amplitude quantization (TAQ) and the traditional phase quantization (TPQ) are reviewed in detail. Second, the regular vector partitioning quantization (RVPQ) is proposed based on TAQ and TPQ, and the error probabilities of RVPQ is deduced, moreover, the concept of optimal RVPQ is presented. Then, three irregular vector partitioning quantization (IVPQ) algorithms are proposed which utilizes K-means clustering, including basic K-means quantization (BKQ), lossy K-means quantization (LKQ) and compensation K-means quantization (CKQ). The three K-means quantization algorithms can not only reduce the quantization error rate greatly, but also ensure the confidentiality of key generation. To overcome the problem of weak uniformity caused by K-means, we put forward the balance mechanism (BM) and apply it to three K-means quantization algorithms. Finally, the simulation results show the VPQ algorithms proposed are superior to previous methods.