Research On Industrial Big Data Sharing Based On Blockchain

With the continuous expansion of the scale of industrial big data,big data intelligent analysis applications for industrial scenarios have gradually emerged.However,in practical applications,industrial data has the problems of large scale,uncentralized distribution,complex structure,and diverse types.In addition,due to factors such as uneven distribution of enterprise data and poor data management capabilities,enterprises are carrying out required data analysis has been hindered.Therefore,how to tap the value of industrial data in one step becomes the focus of attention.By building a data sharing platform,strengthening the cooperation and data openness of superior enterprises,it is possible to effectively solve the problem of data islands among enterprises,develop high-quality data analysis applications,and maximize the value of industrial enterprises' data.The mainstream data sharing system relies on the central server to store and manage data.However,when the central node fails,the centralized data sharing system will collapse on a large scale.Secondly,the cloud service provider of this type of system can obtain full access to all user data,and user data faces certain security threats.Finally,in existing data sharing researches,it is assumed that users are willing to participate in data sharing.However,in real life,in view of data privacy issues and economic benefits,users' willingness to participate is not strong,resulting in data sharing applications that cannot be effectively developed.Therefore,ensuring data privacy and security and effective incentive mechanisms have become an important part of data sharing research.This article proposes the following innovations based on the above two issues:In response to data privacy issues in data sharing,this paper proposes a data sharing privacy protection model based on Blockchain and Federated Learning,BFLDS.BFLDS is based on the blockchain architecture to build a decentralized data sharing trusted network.In response to specific data sharing requests,an on-chain data retrieval mechanism is proposed to select the node of the data providers.Participating nodes perform federated learning based on their local data sets to train the same machine learning model to achieve privacy protection of user data.At the same time,in order to avoid parameter derivation attacks,this paper is based on homomorphic encryption technology to protect the privacy of the participants' local parameters.In addition,in order to achieve the auditability of global parameters,based on BFLDS,this paper innovatively proposes a consensus mechanism based on contribution-based authorized Byzantine fault-tolerant algorithm.The consensus mechanism selects accounting nodes and provides the global parameters updated in each round to other participants for verification in the form of transactions,thereby reaching a consensus.Finally,the BFLDS proposed in this paper realizes data sharing in intelligent analysis scenarios by sharing machine learning models.The safety analysis and experimental results show that the BFLDS model is significantly better than the existing scheme.To solve the problem of insufficient incentives in data sharing,a based on Smart Contact and Evolutionary Game Theory dynamic Incentive is proposed in this paper.By establishing the evolutionary game model of user data sharing,the evolutionary stability strategy of usersAiming at the incentive problem in data sharing,this paper proposes a dynamic incentive model based on smart contract and evolutionary game theory,SC-EGTI.First,introduce "reputation coins" into the system as the encrypted currency for data sharing transactions.Then,aiming at the stability problem of user participation in data sharing has not been realized in the existing researches,this paper proposes to establish an evolutionary game model of user data sharing to analyze the evolutionary stability strategy of users under different conditions.Based on smart contract technology,to dynamically adjust the user's participation benefits under different conditions,so that the user's strategy gradually evolves to participation in decision-making.This paper proposes for the first time based on evolutionary game theory to solve the incentive problem in the federated learning process,and the experimental simulation results show that SC-EGTI can fully encourage users to participate in the collaborative task of data sharing and maximize the effect of the model.Finally,in order to apply the above research to practical scenarios,this paper implements the data sharing system Fabric DS based on Hyperledger Fabric based on the BFLDS and SC-EGTI models.The system can meet the needs of enterprise users for data sharing,and has high use value.