An Efficient and Secure Decentralizing Data Sharing Model
Xiang-Qian DONG,Bing GUO,Yan SHEN,Xu-Liang DUAN,Yun-Cheng SHEN,Hong ZHANG
DOI: https://doi.org/10.11897/SP.J.1016.2018.01021
2018-01-01
Abstract:Data opening and sharing is the source power for driving the development of data-related industries.However,the typical data sharing model available at present,e.g.,data market,in which data providers upload their data to a centralized repository and data demanders download their requested data to carry out analysis,has the following flaws:(1 )As only considering the frequency of keyword in each dataset (or dataset name),the keyword-based dataset retrieval method,which widely used nowadays,cannot efficiently find the linkable datasets.(2 )Being lack of transparency in the process of data transactions,the current data trading model does not take full account of detecting the transaction collusion or other frauds among the involved parties. (3)The data owners lose the power of controlling their own data,which causes no guarantee of data ownership and data security.We found out that these problems exposed in the process of data sharing could be attributed to three factors:linkable dataset discovery,data transaction management,computing security and output security.For the purpose of solving them efficiently and effectively,we proposed a novel blockchain-based decentralization data sharing model,which characterized by followings:(1)It was inspired by restoring data providers greater control over their own data by means of DataSpace (DS).(2 )The computation or analysis was completed confidentially among the data providers,instead of in the data demanders,or in the third parties, as the latter two needed to download data into their own spaces which become the source of privacy leak.(3)It obtained computing datasets or tasks through domain indexing and interface mechanisms,and controlled user behavior and data flow by the blockchain technology.Concretely, in this paper,we first introduced the basic conception of the decentralized data sharing model based on the analysis of the traditional data sharing model.Then,we showed the hierarchical structure diagram of decentralized data sharing model,which included interface,transaction, index,and data layers.Finally,we analyzed the related technologies and implementation details of each layer respectively:In the interface layer,we obtained computing datasets through domain searching mechanism,and compiled the computation contract according to the requirements of the data demanders.In the index layer,we extracted multi-aspect metadata information from the shared dataset,and had the consensus nodes set up domain index to search linkable datasets efficiently.In the transaction layer,with the help of consensus mechanism,we implemented data transaction based on blockchain to achieve transparency and to prevent conspiracy.In the data layer,we introduced the computation contract,which assembled the secure multi-party computation and differential privacy,to ensure the computation and output privacy of the data providers.The experimental results show that the domain index mechanism proposed in this paper increases the average precision by 22％ without substantially reducing the recall rate.And the modified consensus mechanism,which combines time and transaction block number,takes both low trading frequency and high trading frequency into account.At the same time,on the premise of ensuring data security,comparing with the encryption method,our method saves the processing time of nearly 6 s.