Design And Implementation Of SQL Query System Based On Secure Multiparty Computation

At present,data has become a fundamental and key strategic resource comparable to that of petroleum.The circulation and sharing of data elements and the mining of core values are the core contents of the cultivation of the data element market.Information sharing must be realized on the premise of ensuring data security.However,according to the current practice of data circulation,factors such as "data silos" and privacy data leakage restrict the circulation and collaboration of data.Therefore,it is necessary to maintain data security while applying data.How to balance efficiency and risk between development and security,and how to develop data value while ensuring security is an important topic at present.Based on the secure multi-party computation technology,this paper designs and implements a SQL query system based on secure multi-party computation according to the current problems in data privacy protection and collaboration;and then combines the smart medical scenarios to design experimental cases to test and verify the system;finally combining the system with federated learning,a scheme of a federated learning platform based on homomorphic encryption is proposed,and the performance of the scheme is analyzed through experiments.The main work of this paper includes:(1)Aiming at the security issues in data circulation and sharing,by analyzing the limitations of the traditional centralized data query system technical solutions in detail,the architecture of the SQL query system based on secure multi-party computation is designed,and the detailed design of the architecture model is introduced.System function modules and page implementation,use Paillier encryption algorithm to design and implement algorithms such as four rules and comparisons of secure multi-party computation modules.Compared with the security issues brought about by the existing SQL query system delivered to trusted third parties,the limitations of the TEE-based SQL query system and the operation of a single SQL query system based on full homomorphism,the SQL query system designed in this paper overcomes these limitations,is more secure and practical,and supports complex SQL queries.(2)This paper combines the SQL query system with the actual scene of smart medical care to ensure that the original data cannot leave the hospital’s own data platform,and at the same time set up an approval protection mechanism,only after the registration and certification of the privacy computing platform is passed,a data alliance is formed,and then the data circulation scenario between the medical statistics company and the medical data platform is simulated.Design experimental use cases based on medical data,and use the algorithm designed by the secure multi-party computation module to realize the statistical fusion analysis function under the ciphertext of medical data,including secure summation,privacy calculation intersection,secure sorting,and secure calculation average.The realization results prove that the SQL query system has a certain application value in the smart medical scene.(3)Introduce federated learning based on the implementation of SQL query system,analyze the security mechanism of existing federated learning,and then propose a federated learning scheme based on homomorphic encryption.The difference in model accuracy between federated learning under plaintext and federated learning based on homomorphic encryption is compared through experiments.The experimental results show that the accuracy of the training model under ciphertext is slightly lower than that under plaintext,but in view of better security,so a small amount of model loss is acceptable.In addition,this paper provides a privacy retrieval scheme for federated learning prediction.Compared with the existing federated learning prediction scheme,the prediction scheme designed in this paper realizes the security of the query party for the querying user.At the same time,this paper analyzes the correctness and security of the privacy retrieval scheme,and relevant experiments prove the feasibility of the scheme.