High-speed Implementation of Privacy Amplification in Continuous-variable Quantum Key Distribution

Privacy amplification (PA) is an essential process for high-speed and real-time implementation of a continuous-variable quantum key distribution (CV-QKD) system. This work focuses on the improvement of the performance of PA, and we realize PA with Toeplitz matrix and accelerate it using fast Fourier transform (FFT) on graphic processing unit (GPU). Based on the architectural feature of FFT, we adjust its form of input length and obtained an average speed of PA about 2Gbps with input length ranges from 1Mbits to 128Mbits, which is length-adaptable to satisfy various requirements of CV-QKD systems at different transmission distances. Furthermore, we test this work with different compress ratios of PA, which can also achieve a high implementation speed around 2Gbps. With the method used in this paper, the requirements of PA for the high-speed and real-time CV-QKD system can be entirely satisfied.