Research And Implementation Of Data Security Access Mechanism Based On Blockchain

In recent years,the development of some emerging fields,such as Internet of Things,cloud computing,and artificial intelligence,has been tightly bound to the computer information system and cyber security.Taking the data as the carrier and the decentralization as the optimization means,the Big Data era breaks the conventional isolation of the individual data to achieve highly intelligent service.Under such a trend,cross-organizational data access becomes an inevitable requirement for future applications.Nevertheless,the large-scale data exchange and massive information storage caused by cross-domain data access also bring an unprecedented challenge for data security and privacy.Access control is the crucial technique to protect data and service security,which recently attracts widespread attention from both the academic and industrial fields.However,when access control is faced with the above challenge,the differences of security model,description language,logic expression and data storage structure across different domains can lead to the differentiated requirements for the degrees of data access security.For the conventional access control mechanism,it is almost impossible to implement the adaptive adjustment of access strategies.By applying the cryptographic tools,some newly proposed schemes implement data access control under the multi-party setting.However,all these schemes have to introduce a third party,which is totally trusted by all involved data points and has to undertake most of the communication and computation overheads.Once the trusted third party is corrupted,the target network shall be crashed instantly,which is against the goal of distributed computation.Blockchain is a kind of methodology to store messages,execute event and build trust relationships in the open environment,which has the characteristics of unforgeability,traceability,transparency,collectivity and etc.The nature of blockchain provides a new direction for us to implement cross-domain data access.By combining the advantages of blockchain and the attribution based encryption,this paper proposes a novel method to implement secure cross-domain data access for federated learning.For the cross-organizational distributed data access,we build a blockchain network composed of data-owning nodes to solve that problem among different medical institutions.By binding user attributes and identities,our scheme guarantees the authenticity and integrity of user attributes.And then,smart contracts are used to implement automatic management of attributes and access strategies.Due to blockchain storage limitations,we design compression and sharding methods to process the uploaded data.Finally,the validity and feasibility of the scheme are demonstrated by safety and performance analysis.For federated learning,we proposes a scheme for data security access based on blockchain.Combined distributed architecture with ciphertext-based policy attribute encryption(CPABE),our scheme can effectively prevent privacy leakage.Moreover,it can implement model version management,auditing,and data tracking.This paper also designs a federal learning secure access system.We specifically elaborated the system from requirements analysis,overall and functional design,as well as the function implementation and demonstration.Our scheme utility and maneuverability are further exemplified by the system realization.