Research On Compensation Scheme For Continuous Variable Quantum Key Distribution System Based On Machine Learning

Quantum key distribution(QKD)can establish a common secure key between two authenticated users over a public channel,and its unconditional security is guaranteed by the fundamental laws of quantum mechanics.Continuous variable Quantum key distribution(CV-QKD)loads the key information on the quadrature of the optical field,which has the technical advantages of high key rate,convenient source preparation,high detection efficiency,and compatibility with existing fiber optic communication networks.Therefore,it has attracted wide interest in the field of quantum information.In the practical CV-QKD system,due to the nonideal characteristics of the actual device,the measured quadrature values of the receiver’s side are subject to different degrees of interference,which limits the secure key rate and maximum transmission distance.In order to deal with the nonlinear distortion effects and phase drift problems,the corresponding compensation methods based on machine learning are proposed in this thesis.The main work is as follows:(1)An autoencoder-based nonlinear compensation framework is proposed to cope with the nonlinear distortion effects of high-speed CV-QKD systems.Usually,the pulse repetition rate is required to be much lower than the bandwidth of the balanced homodyne detector and the sampling frequency of the ADC,otherwise the measured quadrature values will suffer from severe nonlinear distortion.Considering that the secure key bit rate is proportional to the pulse repetition rate,it is proposed a method to compensate for the distorted measured quadrature values by training the autoencoder network.The numerical simulation show that the excess noise caused by the nonlinear distortion effect can be reduced to below 0.01 after effective compensation procedure.As a result,the CV-QKD system can still operate at a higher repetition rate when the detector’s bandwidth and ADC’s sampling frequency are limited,thus substantially improving the secure key bit rate.(2)A convolutional neural network(CNN)-based phase compensation approach is proposed for the phase drift problem of the local local oscillation(LLO)CV-QKD systems.For the LLO-CVQKD system,since the LO and the signal light are generated by two separate lasers,there is a relative phase difference between them.In the proposed phase compensation method,the phase drift value of the CV-QKD system is estimated by the CNN network,and then the phase compensation is performed on the signal according to the estimated value before the coherent detection process.This eliminates the impact of phase noise on the practical performance of the CV-QKD system to the maximum extent possible.In the numerical simulation,the performance of Kalman filter-based and CNN-based phase compensation methods are compared,and the simulation results show that the CNN-based phase compensation method has a higher accuracy and stability.