Research On Security Of Continuous-Variable Quantum Key Distribution Systems

Continuous-variable quantum key distribution(CVQKD)system can establish the secure sharing key over common channels.Especially,the CVQKD system based on the Gaussian-modulated coherent state can produce and measure quantum signals with the standardized optical devices,and it is compatible with the existing optical networks.Therefore,CVQKD system will be widely applied in prospect.In recent years,in order to solve the various security loopholes caused by the transmitted local oscillator in CVQKD system,a new implementation scheme with the local local-oscillator(LLO)was proposed,of which the feasibility and performance had been investigated widely.In this scheme,the phase of quantum signal was compensated with the reference phase from a reference pulse,and then the quantum signal was measured with the local oscillator generated by receiver.Therefore,those security loopholes caused by the transmitted local oscillator could be solved effectively.However,due to the imperfections of optical devices,there are various kinds of noises in a LLO-CVQKD system,such as the source noise from lasers,the quantization noise from modulators and the photon leakage noise.Moreover,these noises will affect the practical security of a LLO-CVQKD system.Especially,when reference pulses are transmitting in the quantum channel,the eavesdropper will control the phase of reference pulses according to the intercept and capture attack.As a result,the characteristic of phase compensation noise is changed and the potential security loophole appears.To improve the practical security of a LLO-CVQKD system,some important issues are researched as follows:1)In the case that the relative channel drift between reference pulse and quantum signal is a constant state,the zero-mean characteristic of phase compensation noise will be changed by the estimation error of channel drift,and then the mean-drift problem arises.However,the mean-drift is ignored in the traditional model of phase compensation noise,so that the secure bound is overrated.To evaluate the secure bound precisely,the model of phase compensation noise is improved,and the relation of the mean-drift and the secure bound is analyzed theoretically.The results show that the improved noise model is able to describe the impact of mean-drift on the secure bound precisely,so that the secure bound can be evaluated accurately.2)The characteristic of channel drift will change due to the complexity of the actual environment,and it will brings significant estimation error to the existing channel drift estimation algorithms.To reduce the impact of channel drift on system security,the phase searching algorithm is united with the linear interpolation algorithm for enhancing the accuracy of channel drift estimation,and the auto-regressive model is utilized to predict the variation of channel drift,so that the search range is narrowed and the implementation efficiency is raised remarkably.The experiment results show that the improved algorithm of channel drift estimation is able to track the variation of channel drift more precisely,so that the phase compensation noise can be reduced and the system security can be enhanced.3)The accuracy of reference phase estimation will decrease due to the amplified phase noise of reference pulses by eavesdropper,and then the phase compensation noise will increase and the system security will decline.To protect system from the amplification attack of reference phase noise,some algorithms are used to improve the accuracy of reference phase estimation,including the sliding block-averaging algorithm,the sliding polynomial fitting algorithm and the vector Kalman filter algorithm,all of which are associated with the low-order correlation of the laser phase drift.The experiment results show that the improved algorithms are able to defend against the amplification attack of reference phase noise effectively,and the system security is enhanced significantly.4)The eavesdropper is able to reduce system security by amplifying the phase noise of quantum signals.However,since the intensity of quantum signal is too weak and the phase noise is difficult to be distinguished from other noises,so that the phase noise is hard to be suppressed directly.To detect the amplification attack of phase noise,the phase noises of reference pulses and quantum signals are monitored and the attack intensity is evaluated.The results show that the monitoring algorithms based on the inserted training data and the randomly-selected training data are able to evaluate the phase noise level and the attack intensity effectively.