Research On Resource Allocation Technology Of Quantum Key Distribution Network Based On Machine Learning

At present,information network is facing increasingly serious security challenges.Modern cryptography is based on computational complexity,which may be easily cracked by quantum computer in the future.Quantum key distribution(QKD)is a kind of information encryption technology based on quantum mechanics.It can provide the absolute security key for two communication nodes in theory to ensure the security of data transmission.In order to realize the wide deployment of QKD system and meet the encryption needs of multi-users,the QKD network for multi-point interconnection has gradually entered into industrialization.But the key generation rate in QKD network is still low,and the key has high cost of generation and can not be reused.Therefore,it is of great scientific value and research significance to optimize the network resource allocation technology,so as to improve the utilization rate of key resources and build a high-security,extensible,low-cost,and flexibly applied QKD network.This thesis studies how to construct an efficient QKD network from the aspects of network architecture,secret-key storage,management scheme and resource allocation.A intelligent oriented QKD network architecture is designed.In view of low efficiency of secret-key resource utilization and the contradiction between secret-key storage and secret-key security in QKD network,two resource allocation algorithms are designed.The main work and innovation points completed are as follows:(1)In this thesis,an intelligent quantum key distribution network architecture is proposed.In the original network,the intelligent decision layer is added,combined with the ability of SDN global control,to provide better decision-making ability for SDN to allocate network resources.In order to manage key resources better,a quantum key pool(QKP)scheme based on asynchronous transceiver is proposed.The secret-key distribution and secret-key consumption are designed into two independent processes,which improves the efficiency of secret-key management and distribution.(2)In order to solve the problem of low efficiency of key resource utilization in QKD network,a quantum key routing and resource allocation algorithm based on service adaption is proposed.Reinforcement learning model is introduced in this algorithm so that each secure service can choose a route with higher long-term return according to the current resource usage.A heuristic quantum key routing and resource allocation method based on optimal fitting is designed as a comparative study.The simulation results show that,compared with the quantum key routing and resource allocation algorithms based on the best fitting and shortest path,the proposed method based on service adaptation can effectively improve the security traffic blocking and key resource utilization.The performance of the algorithm under different simulation conditions is verified.(3)In order to solve the problem of secret-key preservation and secret-key security,this thesis analyzes the factors influencing the secret-key lifetime in the QKP.Then,the influence of routing on the distribution of secret-key resources in QKD network is analyzed.In order to balance the use of secret-key,the concept of QKP health is proposed to evaluate the risk of the QKP in the state of insufficient resources or resource overflow.Based on this,a quantum key routing and resource allocation algorithm based on the double threshold of the QKP is proposed.Reinforcement learning model is introduced in this algorithm,so that the secret-key amount of the QKP in the network does not exceed the threshold value as much as possible.The simulation results show that,after setting reasonable threshold for the QKP,compared with the quantum key routing and resource allocation algorithm based on the best fitting and shortest path,the health of the QKP proposed in this thesis is improved by 5.5%and 12.6%respectively,the blocking rate of security service is reduced by 26.6%and 42.6%respectively,and the utilization rate of secret-key resources is improved 2.4%and 4.5%respectively.