Article Preview
Top1. Introduction
In recent years, wireless physical layer security has received more and more attention. The model of physical layer security was first established by Shannon (1949), it mainly includes two branches: security schemes without key leaded by Wyner (1975) and secret key based secrecy schemes leaded by Maurer (1993). The basic idea of secret key based secrecy schemes is to use the uniqueness and randomness of wireless communication link to convert channel mutual information of between authorized users into a shared key, encrypt the data and achieve the secure communication. As eavesdropper cannot obtain the same channel information as authorized user, it can hardly generate the identical key. How to effectively extract secret key from the common information has become the focus of physical layer security.
Orthogonal frequency division multiplexing (OFDM) technology has been widely used in wireless communication systems. Due to the diversity of channel information, key generation methods in OFDM systems have received wide attention. The generalized features, such as received signal strength (RSS) and bit error rate (BER), can be used to extract secret key. For example, the work in (Kitano et al., 2007) generates keys from BER of received signal, and uses an intermediate value of BER as a quantization threshold. A key generation method for wireless local area network systems is studied in (Yasukawa et al., 2008), the quantization threshold is determined by sorting RSS value of a period of time, which makes the probability of quantization value is basically the same. Because RSS and BER are both generalized channel information, the key generation rate (KGR) obtained by these methods is relatively low.
OFDM technology uses multi-carrier transmission, which can conveniently obtain the channel state information (CSI) on each subcarrier in channel estimation process. These large numbers of common random information provide favorable conditions for obtaining higher KGR (Liu et al., 2012; Liu et al., 2013). However, in order to effectively use this advantage, it still needs to solve some problems. Firstly, CSI contains a large amount of redundant information, which will reduce the randomness of key. Secondly, due to the influence of RF device asymmetry, time division duplex (TDD) transmission characteristics and environmental noise, channel reciprocity will be destroyed. A discrete cosine transform method has been studied in (Yasukawa et al., 2008) to reduce the redundancy. In addition, an adaptive multi-bit quantization method has been adopted to generate the secret key. In order to improve channel reciprocity, a low-pass filter has been designed to reduce the noise in (Zhang et al., 2015). A channel gain compensation method has been proposed to reduce the effect of the asymmetry of radio frequency circuit in (Liu et al., 2013).
In all of the secret key generation methods, it is necessary to quantize channel information. The selection of quantization method is related to the key generation efficiency. A suitable quantization method can be used to maximize the use of these channels information. At present, the primary quantization methods mainly include single threshold quantization (Aono et al., 2005), double threshold quantization (Mathur et al., 2008), multi bit quantization (Jana et al., 2009) and cumulative distribution function based quantization (Patwari, et al., 2010), etc. The key sequence generated in the quantization process can be called a candidate key. In order to generate an available key, it still needs to carry out information reconciliation and privacy amplification.