Research On Data Sharing Model Based On Blockchain And Federated Learning

In the field of modern medical and health care,with the continuous improvement of the level of medical information,a large amount of patient medical data is digitized and stored in the medical information system.However,the security and privacy protection of these data has always been one of the important challenges in the field of healthcare.For example,the past few years have seen numerous healthcare data breaches around the world,some of them large and widespread.These leaks not only threaten the privacy of individuals,but also have a serious impact on the stability and trust of the entire healthcare system.On the other hand,the sharing of medical and health big data is also one of the important challenges currently facing.The inconsistency of data and information structures caused by the incompatibility of information systems used by current medical institutions makes it difficult for hospitals to exchange information smoothly and achieve cross-institutional information sharing.This also makes doctors lack the necessary information support in the process of diagnosis and treatment,thus reducing the effectiveness of patients’ diagnosis and treatment.Therefore,this thesis combines block chain and federated learning technology to study how to share the efficiency of medical and health data under the premise of ensuring security and privacy.The main research contents are as follows:In the entire medical and health network,in order to solve the problems of imperfect privacy protection,uneven data source quality,and poor quality of sharing models in the data sharing process of multiple medical institutions,this thesis proposes a new data sharing and privacy Protection scheme,which organically combines federated learning and block chain technology.First,we use federated learning technology to model the data of multiple medical institutions,and store the reputation values of these models on the block chain to better support and manage these models.After in-depth research,it is found that the quality of data sources will directly affect the performance of the federated learning algorithm,and a subjective logic model calculation scheme based on the combination of direct reputation value and indirect reputation value is proposed to screen relevant medical institutions participating in federated learning,and finally use the classic Fed Avg Compared with Fed Prox aggregation algorithm and related literature,the results show that this scheme can select high-quality data sources in the scenario of health data exchange of multiple medical institutions,which not only stimulates the participation of relevant medical institutions,but also improves the Accuracy of federated learning training.For the medical and health network based on block chain and federated learning,when there are malicious institutions using malicious and false information data to attack the model,which makes the local model training incorrect,this thesis designs additional participating institutions in the block transaction content The model parameters and model evaluation,and on this basis,this thesis proposes a new algorithm,which is based on PBFT improvement,can better evaluate the performance of the model,and has a high-performance Byzantine consensus,referred to as MEPBFT.The consensus algorithm fully considers the training quality of the local model and the local reputation and trust of the participating nodes.Based on the comprehensive evaluation,a consensus group is established to complete the efficient consensus process.At the same time,a dynamic node information table is established to simplify the consensus process and realize the participating nodes.Real-time dynamic joining effectively improves consensus efficiency.These optimizations ensure the normal operation of the medical and health data sharing network under this architecture.At the same time,the implementation of the algorithm further enhances the efficiency and reliability of the network.