Research On Resource Assignment Schemes In The Optical Network Of Quantum Key Distribution

The emergence of quantum key distribution(QKD)provides a promising defence against the information security problem in the future,which is brought about by the rapid development of computing power.In recent years,the theory of QKD has been constantly improved,and it is leaping out of the laboratory to practical applications.The next phase of development is the large-scale QKD networking,and in order to improve the scalability and reduce implementation costs,integrating QKD with optical networks,i.e.establishing optical networks of QKD is necessary.This thesis focuses on how to transmit quantum signals based on the existing optical network resources and try to solve the major chal-lenges in the integrated QKD networks such as noise interference and resource competition.On these bases,the resource assignment schemes to improve the reliability and the efficiency of resource sharing are proposed.The main contributions and innovations are as follows:(1)To suppress the various noises in the co-fiber transmission system of quantum signals and classical signals,a jointly optimized channel allocation(JOCA)scheme targetting on the four-wave mixing(FWM)noise and Raman scattering noise is proposed.In JOCA scheme,the influence of FWM noise and Raman scattering noise on QKD can be simultaneously reduced by the unequally-spaced allocation selection and the Raman scattering optimized spacing selection.The wavelength reconfigurable co-fiber transmission system is experimentally demonstrated,and based on which the noise photons are measured under different conditions.The results verify that the proposed JOCA scheme can eliminate almost all of the FWM noises and reduce the Raman scattering noise by 23%at most.Besides,the performance evaluation of the QKD system shows that the secure key rate is increased by two to three times compared with the conventional scheme.In order to improve the practicability of the proposed scheme,we also research the adaptation of JOCA scheme to the point-to-multipoint optical access network.Under the condition of the simulations in this thesis,the scale of quantum optical access can be increased to 20 km through the JOCA scheme,and the secure key rate can be at most five times higher than that of the conventional schemes.(2)To reduce the time-varying noise impairments in the dynamic network environment,a machine learning based noise-suppressing channel allocation(ML-NSCA)scheme is proposed.Targetting on the challenge that the data traffics cannot be known ahead in the dynamic networks,a LightGBM-based ML framework is designed in this scheme to predict the optimal quantum channel allocation.The ML model is trained based on the Monte Carlo simulation so that it can predict the noises of each channel in the case that the determinate information about data traffics is unknown.Based on the prediction,the quantum channels are periodically reconfigured to guarantee the reliability of QKD under time-varying noises.Additionally,to improve the performance of the ML framework,the method of feature derivation and extraction is optimized,through witch,the tested prediction accuracy can reach above 95%.The proposed ML-NSCA scheme provides a more efficient method to suppress the time-varying noise interference.Compared with the conventional fixed band channel allocation(FBCA)scheme and performance prediction channel allocation(PPCA)scheme,the secure key rate can be increased by up to 42%and 31%respectively.(3)To mitigate the resource competition problem in the coexistence of QKD and data services,a key-driven wavelength assignment(KSD-WA)scheme is proposed,and the heuristic optimization algorithm and the Deep reinforcement learning(Deep RL)based optimization algorithm are designed.Specifically,guaranteed by the quantum storage and management technology based on the quantum key pool,the propose KSD-WA enables reconfiguration of quantum channels to recycle the wavelengths fragment to transmit quantum signals.In the meanwhile,to satisfy the encryption requirements under the constraints of physical layer noise impairments,the wavelength assignment in the KSD-WA scheme is optimized,and the minimum gain guaranteed(MGG)algorithm is proposed.Furthermore,a more efficient and more intelligent Deep RL-based optimization algorithm is also designed,which can automatically learn the policy.The simulation results show that the proposed MGG algorithm and the Deep RL-based algorithm effectively improve the secure key rate,while the latter shows better adaptability under different environments.In addition,under the same encryption requirement,the conventional channel allocation scheme severely affects the provisioning of data services(the blocking rate in the highly loaded scenario can be increased by about 10%),while the KSD-WA scheme guarantees the quality of service(QoS)of the data services(the same as that without QKD)and significantly improves the compatibility of QKD with the real-life optical networks.In summary,this thesis focuses on resource assignments in the optical networks of QKD and proposes innovative solutions in terms of noise interference suppression and efficient resource sharing.The machine learning is also introduced to achieve effectiveness and intelligence.The research of this thesis improves the feasibility of the integration of QKD and optical network and provide technical supports for the development of large-scale optical networks of QKD.