Research And Application Of Randomness Extraction In Quantum Secure Communication

Quantum secure communication is a hot research topic at home and abroad,which can achieve provably and unconditionally secure key generation and distribution through the use of physical properties of quantum mechanics.Its representative technologies mainly include Quantum Key Distribution(QKD)and Quantum Random Number Generator(QRNG).QKD can generate and distribute a set of random numbers with information provable security for both parties of communication as the key for encryption and decryption,so as to ensure the security of communication.QRNG generates truly random numbers based on the observation of quantum states,with information provable security and a high speed.Quantum secure communication system usually includes quantum state preparation,transmission,detection and data processing.The randomness extraction is the key of data processing,and affects both the security and efficiency of the system.For example,Privacy Amplification(PA)in Continuous Variable Quantum Key Distribution(CV-QKD)system and the post-processing in QRNG are both belong to randomness extraction operation.An efficient method to implement randomness extraction is based on Toeplitz matrix.This paper focuses on the design of implementation scheme,tackle problems in key technologies under the two above application scenarios,which provide key technology support for the high-performance CV-QKD and QRNG system developed by project team.Specific research achievements are summarized as follows:1.The efficiency of PA is one of the decisive factors for the performance of CV-QKD system.In order to reduce the influence of finite size effect,CV-QKD system requires the input length of PA under the transmission distance of 50 km fiber channel to be at least10⁸.However,the limited computing resources and storage space will seriously affect the performance of PA under the large-scale input length.In this paper,we design and implement an algorithm of PA:block FFT algorithm,which effectively solves the implementation problem of PA under the large input length with limited resources.What is more,the block FFT algorithm is optimized to ensure the high performance of PA.This work achieves an average speed of 2G bps under the condition of different input length of PA,which is at the international advanced level among the related fields(the fastest speed that has publicly reported is1.35G bps based on the same computing platform).This algorithm can support input length of PA with 2²⁷,which can greatly reduce the finite size effect of CV-QKD system under 50 km communication distance,and effectively support the research of high-speed and stable CV-QKD system developed by the project team.2.Classical noise will be introduced in the process of quantum signal detection and data sampling of the QRNG system,so it is necessary to perform post-processing to remove the bias and redundancy in the initial random sequence,so as to obtain truly random numbers.Compared with XOR/von Neumann and other methods,the implementation of post-processing based on Toeplitz matrix is information provable secure.However,execution speed of post-processing based on Toeplitz matrix is the bottleneck to limit the high-speed QRNG system because of its complex implementation.Aiming at the problem that the speed of real-time post-processing does not match the practical application demand of high-speed QRNG,this paper proposes a multi-thread and streamed processing algorithm based on Toeplitz matrix.It mainly refines large-scale matrix multiplication into data processing among multiple threads,which work with pipelining mode to achieve the maximization of resource utilization with limited time.The experimental results show that the post-processing with an average speed up to 6G bps is achieved in this paper,which effectively improves the performance of post-processing with a large-scale Toeplitz matrix and ensures the stable operation of the real-time and high-speed QRNG system.At the same time,this work does three kinds of random test(ENT,NIST and Diehard)with the final random numbers.The results show that all can pass the random test,which means the performance of random extraction is excellent.