An Efficient Information Reconciliation Scheme via Partial Encoding for Physical Layer Secret Key Generation

This letter proposes an efficient information reconciliation coding scheme for physical layer secret key generation. We consider a high-rate key generation system, where multiple key bits are extracted from one Gaussian random variable source, each of which has different reliability depending on its bit level. Although the conventional bit-interleaved coded modulation (BICM)-type reconciliation protects all the key bits by forward error correction (FEC), the computational complexity associated with its encoding and decoding can be a bottleneck in achieving a high secret key rate especially for low-cost devices with limited computational capabilities. To alleviate this complexity issue, we propose a new reconciliation scheme that efficiently reduces the complexity by applying FEC only to the least reliable bits (LRBs). We demonstrate by simulations that the proposed information reconciliation based on partial encoding achieves a lower key bit mismatch rate (BMR) than the conventional BICM-type reconciliation even with lower decoding complexity. Furthermore, we provide a lower bound on the asymptotic BMR of the proposed scheme that is useful for estimating the performance in high signal-to-noise ratio (SNR) regimes.