Construction method of industrial trusted data space and data processing method
By building a view layer and data network interface locally, a categorized data view is generated, which solves the problems of data ownership and privacy leakage in the centralized data collection mode, realizes the trusted sharing and trading of data, and supports plug-and-play and trusted circulation of multi-source data.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- WUHAN METROPOLITAN RING ROAD NORTHERN SECTION INVESTMENT MANAGEMENT CO LTD
- Filing Date
- 2025-12-28
- Publication Date
- 2026-05-19
AI Technical Summary
In existing technologies, the construction of industry data spaces mainly adopts a centralized data collection and over-the-counter trading model, which makes it difficult to define data ownership, increases the risk of privacy leakage, causes large differences in data formats and access protocols, and results in high transformation costs, making it difficult to achieve rapid access and reliable circulation of multi-source heterogeneous data.
Under the premise of local data retention, a view layer and data network interface are built to generate categorized data views. Through data openness, sharing, exchange and transaction interfaces, the trusted sharing and transaction of data are realized, ensuring that the data can be called by external systems through standardized interfaces while being retained locally.
It enables data to be stored locally while being shared, exchanged, and traded reliably through standardized interfaces, avoiding risks of unclear ownership and privacy leaks, and supporting plug-and-play and reliable circulation of multi-source data.
Smart Images
Figure CN122064751A_ABST
Abstract
Description
Technical Field
[0001] This application belongs to the field of trusted data space construction, and more specifically, relates to a method for constructing an industry trusted data space and a data processing method. Background Technology
[0002] The current construction of industry data spaces primarily adopts a centralized data collection and off-exchange trading model. This involves aggregating data from various parties to a central server or trading platform, followed by unified cleaning, processing, and distribution. Under this model, data must be removed from its local environment and rely on third-party platforms for sharing, exchange, or trading, leading to difficulties in defining data ownership, increased risks of privacy breaches, and a loss of continuous control over the data by data providers. Furthermore, existing technologies lack unified interface standards, resulting in significant differences in data formats and access protocols between different systems, high transformation costs, and low data circulation efficiency, making it difficult to meet the industry's needs for rapid access and reliable circulation of multi-source, heterogeneous data.
[0003] Therefore, how to build a unified data network interface while retaining data locally, so as to realize the trusted sharing, exchange and trading of industry data, is an urgent problem to be solved. Summary of the Invention
[0004] In view of the shortcomings of the existing technology, the purpose of this application is to provide a method for constructing an industry trusted data space and a data processing method, which can build a unified data network interface under the premise of local data retention, and realize the trusted sharing, exchange and transaction of industry data.
[0005] To achieve the above objectives, firstly, this application provides a method for constructing an industry-trusted data space, comprising the following steps: A view layer is built on top of the database, which is the metadata layer obtained by secondary processing of the internal data. The database includes a business database, a data center database, or a basic dataset. The view layer processes the data in the database based on the data attributes and openness requirements to generate multiple data views. The data views are classified as public welfare data, open data that does not require privacy protection, open data that requires privacy protection, and limited open data. A data network interface is constructed, which is a function layer that processes data in the view layer and can be called internally or externally. The data network interface is connected to the view layer and is used to realize data sharing, exchange and trading through data open interface, data sharing interface, data exchange interface, data transaction interface and data model interface.
[0006] The method for constructing a trusted data space in the industry provided in this application has the following effects: (1) By constructing a view layer on top of the database, a classified data view is generated based on data attributes and open requirements. Local data can be processed without transferring the original data to form a data view that conforms to different open levels, ensuring that the data is retained locally; (2) By constructing a data network interface connected to the view layer, differentiated services can be provided to the outside world through data open interface, data sharing interface, data exchange interface, data transaction interface and data model interface according to the different classification attributes of the data view. This allows the data to be retained locally while still being called, integrated or traded by external systems through standardized interfaces, avoiding the risks of unclear ownership and privacy leakage brought about by the traditional "first aggregation, then transaction" model; (3) The view layer and the data network interface work together to form a unified, plug-and-play interface system under the condition that the data does not leave the domain, supporting the trusted circulation of multi-source data in the industry. Specifically, it can realize the sharing, exchange and transaction capabilities of data that are "usable but not visible, controllable and measurable", thereby solving the problems of centralized data upload, inconsistent interfaces and uncontrollable circulation paths in the existing technology.
[0007] As a further preferred embodiment, the method for constructing the view layer includes at least one of table reconstruction, table import, view creation, data dump, stored procedure, data statistics table, and data index.
[0008] As a further preferred embodiment, the table reconstruction includes extracting partial data from the table using the Create command, or hiding identifiers using anonymization technology, or changing the identifiers into k types using k-anonymization technology, or processing location information using fuzzification technology; The table import includes importing data from different tables in different databases using the import command, and merging raw data from different sources; The creation of the view includes using the Create View from Tables command to create a real-time view of a portion of the data from a portion of the tables; The data dump includes moving or backing up data using the dump command; The stored procedures include function libraries of the database management system or data processing programs developed by software developers; The data statistics table includes data statistics information or data calculation results; The data index includes an index for quick data location.
[0009] As a further preferred embodiment, a function layer is constructed on top of the view layer, the function layer including multiple processing functions for processing the data view.
[0010] As a further preferred embodiment, the processing functions include data statistics functions F1, data privacy protection functions F2, data machine learning functions F3, data transaction functions F4, data access control functions F5, federated learning functions F6, data processing functions F7, and model processing functions F8.
[0011] As a further preferred embodiment, F1 includes one or more of summation, mean calculation, and variance calculation; F2 includes one or more of desensitization function, decryption function, hierarchical function, classification function, privacy intersection function, and fully homomorphic encryption calculation function; F3 includes one or more of linear regression, logistic regression, support vector machine, random forest, and Naive Bayes; F4 includes one or more of payment, data verification, and withdrawal; F5 includes one or more of access control list filling, access control query, and access control policy definition; F6 includes submodel submission function and / or model parameter acquisition function; F7 includes one or more of image processing function, video processing function, and model generation function; F8 includes CNN model function and / or RNN model function, wherein the CNN model function and RNN model function take sample data as input and output the model structure and parameters.
[0012] Secondly, this application provides an application of the method for constructing an industry trusted data space as described in any one of the above statements, which can be applied to the intelligent transportation industry or the natural resources industry.
[0013] Thirdly, this application provides a fair data processing method based on a data space, wherein the data space is constructed using the industry trusted data space construction method described above, and the data processing method includes the following steps: S11, the data provider divides the data to be processed into multiple parts, and opens one of them to multiple data processors through the data view of the view layer; S12, each data processor uses the open data view to conduct model learning or algorithm design, obtains its own processing method, and provides the processing result and the hash value of the corresponding processing method to the data provider; S13, the data provider selects the data processor that provides the best processing result based on the multiple processing results received, and pays a fee to the selected data processor; S14, the selected data processor encrypts the processing method using the data provider's public key and sends it to the data provider. The data provider decrypts the processing method using its private key and verifies whether the hash value of the decrypted processing method is consistent with the hash value received in step S12. S15, when the hash value verification is consistent, the data provider uses the processing method to process the data to be processed to obtain the processed data.
[0014] As a further preferred embodiment, the processed data is stored as a limited open static data in the data space, and provides external access control based on payment, scope control based on access control lists, and scope control based on access control policies.
[0015] Fourthly, this application provides a model invocation method based on a data space, wherein the data space is constructed using the industry trusted data space construction method as described in claim 1, and the model invocation method includes the following steps: S21, Data processors train models on public datasets provided by data providers through the view layer to obtain models or algorithms; S22, The data processor discloses the hash value of the model or algorithm; S23, The model user requests to invoke the model or algorithm through the data network interface and pays the invocation fee; S24, the model user sends the data to be processed and the hash value to a trusted third party, and the data processor sends the model or algorithm to the trusted third party; S25, a trusted third party verifies whether the hash value of the received model or algorithm is consistent with the hash value disclosed in step S22. If they are consistent, the model or algorithm is used to calculate the data to be processed, and the calculation result is delivered to the model user.
[0016] It is understandable that the beneficial effects of the second, third and fourth aspects mentioned above can be found in the relevant descriptions in the first aspect above, and will not be repeated here. Attached Figure Description
[0017] Figure 1 This is a flowchart of the method for constructing the industry trusted data space provided in this application; Figure 2 This is a flowchart of the model transaction method provided in this application after the separation of data providers and data processors; Figure 3 This is a flowchart of the data processor's model and algorithmic trading method provided in this application; Figure 4 This application provides a BIM model drawing provided by the design institute to the bridge construction unit. Figure 5 This is a hoisting data diagram shared by the construction unit and the supervision unit, as provided in the embodiments of this application. Detailed Implementation
[0018] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.
[0019] It should be understood that, in the description of this application, the term "multiple" means two or more, unless otherwise expressly and specifically defined; the term "and / or" includes any and all combinations of one or more of the associated listed items.
[0020] This application, based on research, finds that the National Data Administration has proposed accelerating the establishment of a secure data element circulation mechanism and the construction of a trustworthy data space, requiring that data be "supplied, circulated, used effectively, and securely." Existing technical solutions mainly focus on how data is provided, how it is collected to a central server, and how it is traded over-the-counter. Currently, no solution addresses the direct local collection and sharing of data. Therefore, this application proposes a local plugin, which is plug-and-play, for building a local interface to share and trade local data, thus offering a new method for achieving "supplied, circulated, used effectively, and securely."
[0021] Among these, "supplying" requires ensuring data security, including data compliance, data classification and categorization, and data anonymization and declassification; "flowing" requires ensuring the interests of data providers, including data rights confirmation and protection, profit sharing, data value assessment, data entry into tables, and data financing; and "using well" requires ensuring that data can be processed under privacy protection conditions, and that the interests of data processors (such as data mining algorithm providers, CNN model providers, large model providers, and machine learning algorithm providers) are protected, primarily through privacy-preserving machine learning.
[0022] The establishment of an industry data space will enable the online sharing, circulation, and even trading of all relevant unit data, office data, and business data within the industry, forming a unified data interface. This will be a plug-and-play method with low transformation costs and quick results.
[0023] like Figure 1 As shown, this application provides a method for constructing an industry-trusted data space, including steps S10 to S20, which are detailed below: Step S10: Build a view layer on top of the database, which is the metadata layer obtained by secondary processing of the internal data. The database includes a business database, a data center database, or a basic dataset. The view layer processes the data in the database based on the data attributes and openness requirements to generate multiple data views. The data views are classified as public welfare data, openable data without privacy protection, openable data with privacy protection, and limited open data.
[0024] This step involves building a view layer to process the raw data, specifically enabling data cleaning, data privacy protection, data classification and openness, secondary data refinement, and metadata construction. This ensures the credibility and privacy protection of data during circulation and avoids the risks of data misuse and leakage.
[0025] Step S20: Construct the data network interface, which is a function layer that processes the data in the view layer and can be called internally or externally. The data network interface is connected to the view layer and is used to realize data sharing, exchange and trading through data open interface, data sharing interface, data exchange interface, data transaction interface and data model interface.
[0026] This step, by building a data network interface, can provide a standardized and controllable data access mechanism, specifically enabling secure data flow and value utilization. For example, it can support multi-party data collaboration in a plug-and-play manner, ensuring that data is shared, exchanged, and traded while remaining visible.
[0027] The data network interface provided in this embodiment serves as a plug-and-play plugin for existing data systems, enabling data sharing, virtualization, and trusted data computing.
[0028] In this embodiment, the data open interface refers to certain data in the local data space that is open for access, such as public resource data.
[0029] Data sharing interface: Certain data in the local data space is accessible to authorized (paid) users, such as certain data products, metadata, and proprietary data. Data exchange interface: Certain data in the local data space is used to join a data space exchange alliance, where this data is freely exchanged among alliance members.
[0030] Data transaction interface: Some data in the local data space is used for external transactions, with pricing mechanisms, payment mechanisms, etc.
[0031] Data Model Interface: Certain data in the local data space is used for federated learning, distributed learning, transfer learning, ensemble learning, etc. After the model of this data is trained, it provides external (paid) calling function (e.g., charging by token) and provides integration of external model parameters (e.g., submission of federated learning sub-models).
[0032] Specifically, this includes: Data Open Interface Security: Compliance monitoring of publicly shared data. Data Sharing Interface Security: Compliance monitoring of publicly shared data and control of external data access requests. Data Exchange Interface Security: Compliance monitoring and access control of external data exchange requests. Data Transaction Interface Security: Ensuring the rationality, fairness, security, and sustainability of externally traded data. Data Model Interface Security: Controlling the risk of data privacy leaks from publicly shared models and the privacy risks of information returned by model calls.
[0033] The method for constructing a trusted data space in the industry provided in this application has the following effects: (1) By constructing a view layer on top of the database, a classified data view is generated based on data attributes and open requirements. Local data can be processed without transferring the original data to form a data view that conforms to different open levels, ensuring that the data is retained locally; (2) By constructing a data network interface connected to the view layer, differentiated services can be provided to the outside world through data open interface, data sharing interface, data exchange interface, data transaction interface and data model interface according to the different classification attributes of the data view. This allows the data to be retained locally while still being called, integrated or traded by external systems through standardized interfaces, avoiding the risks of unclear ownership and privacy leakage brought about by the traditional "first aggregation, then transaction" model; (3) The view layer and the data network interface work together to form a unified, plug-and-play interface system under the condition that the data does not leave the domain, supporting the trusted circulation of multi-source data in the industry. Specifically, it can realize the sharing, exchange and transaction capabilities of data that are "usable but not visible, controllable and measurable", thereby solving the problems of centralized data upload, inconsistent interfaces and uncontrollable circulation paths in the existing technology.
[0034] In one embodiment, the technical solution to achieve the above objective can be as follows: This embodiment describes a method for designing and constructing a data network interface for the intelligent transportation industry data space, and the technical solution is as follows: (1) Build a view layer on top of the existing data layer—that is, build a data classification view based on the attributes and openness requirements of the data. The view layer is a data layer built on top of the existing database, and it is the data layer that is accessible to external users.
[0035] The existing database serves as the business layer of the current information system, the database management system of a data center, or the basic dataset. The basic dataset cannot be directly shared; a view layer needs to be built on top of it. This view layer has functions such as data cleaning, data privacy protection, data classification and openness, secondary data refinement, and metadata construction.
[0036] A view layer is built upon the existing database. The methods used include: Table reconstruction can be achieved by using the Create command to extract a portion of the data from a table; or by using anonymization techniques to hide identifiers; or by using k-anonymization techniques to transform identifiers into k types; or by using obfuscation techniques, such as transforming a location into a street name and removing house numbers. Table import uses the import command to import data from different tables in different databases, merging raw data from different sources; The Create View command, Create View from Tables, creates a view from a portion of the data in a table and generates it in real time. The view reflects the points of interest, such as a specific area of a map. The `dump` command can be used to move data to a new location or back up data. Stored procedure commands are functions provided by the database management system, or they can be data processing programs developed by software developers for raw data. Data statistics tables, including various statistical information or the results of data calculations; Data indexing, including various indexes, facilitates quick data location, etc.
[0037] Based on the data attributes and data access requirements, this embodiment divides the data in the view layer into the following 7 categories: Data that must be made available to the public is referred to here as public interest data D1.
[0038] Data that can be made publicly accessible, historical static data D2 that does not require privacy protection; Data that can be made publicly accessible, and historical static data (D3) that requires privacy protection; Data that can be made publicly accessible, current dynamic data D4 that does not require privacy protection; Data that can be made publicly accessible, and current dynamic data (D5) that requires privacy protection; Limited open static data D6; Limited open dynamic data D7.
[0039] It should be noted that in this embodiment, public welfare data is data that does not require privacy protection; limited open data is data that requires privacy protection. D1, D2, and D4 are directly open and do not require view construction, and can be regarded as transparent views. D3, D5, D6, and D7 require view construction or privacy protection processing. Methods for creating views include: selecting specific records to form a new table, selecting specific fields to form a new table, anonymizing records in specific fields, and obfuscating data in specific fields.
[0040] (2) Constructing the function layer—constructing a function library for view processing functions. After data classification and view creation, a view function layer needs to be constructed above the view layer. This layer includes the processing of the view. The function layer includes various algorithms and models (large language models, small language models, various convolutional neural network models, machine learning algorithms, video recognition algorithms, video generation algorithms, image recognition algorithms, image generation algorithms, etc.). The function layer is a logical layer, and its physical location can be on the same machine as the view layer. Algorithms include methods for processing data. The functions in the function layer can be divided into eight categories: data statistics functions (F1), data privacy protection functions (F2), data machine learning functions (F3), data transaction functions (F4), data access control functions (F5), federated learning functions (F6), data processing functions (F7), and model processing functions (F8).
[0041] View handling functions can be applied to multiple views.
[0042] Functions in F1 include: summation, mean, variance, etc. The functions in F2 include: de-identification function, decryption function, hierarchical function, classification function, privacy intersection function, and fully homomorphic encryption computation function; The functions in F3 include: linear regression, logistic regression, SVM, random forest, and Naive Bayes; The functions in F4 include: payment, data verification, and withdrawal; The functions in F5 include: access control table population, access control query, and access control policy definition; The functions in F6 include: submodel submission function and model parameter retrieval function; The functions in F7 include: image processing functions, video processing functions, and model generation functions; Functions in F8 include CNN models and RNN models. These model functions take a large amount of sample data as input and output the model structure and its parameters. F8 functions can also be called pattern extraction functions or data processing functions.
[0043] (3) Establish a model trading method after the separation of data providers and data processors (i.e., a fair data processing method based on data space). The National Data Administration's plan is to separate data providers and data processors, introduce more data processing capabilities, and activate the value of data. However, it does not provide a method for protecting the rights and interests of data providers and data processors after the separation.
[0044] A common problem is that data providers lack processing capabilities, and data processors lack data. How can data processors carry out data processing while ensuring that data providers do not lose data, thereby obtaining processed data or processing methods (models, algorithms), and then providing these processing methods to data owners (or selling models or algorithms) so that data owners can use these processing methods (models or algorithms) to create data products?
[0045] Therefore, this embodiment proposes a transaction protocol for a model (or algorithm). This protocol can leverage the capabilities of data processors to provide data processing methods and offer high-quality processing algorithms to data providers, thereby helping them generate more high-quality data products, such as... Figure 2 As shown, the protocol is designed as follows: S11, the data provider divides the data to be processed into multiple parts, and opens one of them to multiple data processors through the data view of the view layer; S12, each data processor uses the open data view to conduct model learning or algorithm design, obtains its own processing method, and provides the processing result and the hash value of the corresponding processing method to the data provider; S13, the data provider selects the data processor that provides the best processing result based on the multiple processing results received, and pays a fee to the selected data processor; S14, the selected data processor encrypts the processing method using the data provider's public key and sends it to the data provider. The data provider decrypts the processing method using its private key and verifies whether the hash value of the decrypted processing method is consistent with the hash value received in step S12. S15, when the hash value verification is consistent, the data provider uses the processing method to process the data to be processed, and obtains the processed data.
[0046] Specifically: (1) Data provider T divides the data D to be processed into n parts, and opens one of the parts d to data processor P through a data view; (2) Data processor P uses the publicly available d of T to carry out model learning and algorithm design, and provides the result r after processing the data d. The result r is sent to the data provider T, but the processing method F is not publicly available. Here, the input of F is d and the output is r. That is, F is a certain model or algorithm. Although F is not publicly available, the H value of F is sent to T along with r. Here, H=Hash(F). (3) The data provider T compares the obtained r and thus determines which P provides the best r. Assuming that the r provided by P1 wins, that is, P1 is selected, T pays the fee to P1 and requires P1 to provide F (i.e. model or algorithm). (4) P1 encrypts F with T’s public key and sends it to T. T decrypts F with its private key and verifies whether the hash value of F is consistent with the H received in (2). If they are consistent, F is accepted; otherwise, P1 is asked to refund. (5) T uses F to process D to obtain the processed data D', and put D' into D6; (6) D' is a type of limited open static data that provides access control to the outside world, including: payment-based access control, and scope control based on access control lists and access control policies.
[0047] The difficulty (innovation) of the above protocol design lies in how to protect the rights and interests of data providers and data processors: 1) Data provider T needs to protect the copyright or privacy of D, so it cannot provide all of D. However, it also needs data processor P to train the model and algorithm. Therefore, it can only provide part of the data d and use the processing result r of the data to judge the processing capability of data processor P; 2) Data processor P needs to demonstrate its data processing capability while protecting its intellectual property rights, that is, it cannot expose the parameters of F. Therefore, it cannot deliver F directly, but delivers part of the result r. F is delivered only after T pays. However, P is worried that F is not easy to use or that F is not the F it originally wanted. Therefore, P needs to deliver H to T in advance so that T can verify whether F is the F it wants to buy; 3) The protocol also provides a competitive model for data processors to provide models. The data provider selects the best model and processes it locally to form a data product.
[0048] (4) Model and algorithm trading method for data processors (i.e., a model invocation method based on data space) Currently, a crucial component of the secure circulation of data elements or the construction of a trusted data space is the value-added data processing services provided by data processors. How to monetize this data processing capability is a key link in accelerating the marketization of data elements, promoting transaction prosperity, and facilitating data utilization and circulation.
[0049] This embodiment proposes a transaction agreement for a model or algorithm. This agreement is a three-party agreement, including: a data provider T, a data processor P, and a model user U. Because the model user may not be the data provider T, the model user U becomes the buyer of the model, further incentivizing the data processor P to provide better models and algorithms. Figure 3 As shown, the protocol is designed as follows: S21, Data processors train models on public datasets provided by data providers through the view layer to obtain models or algorithms; S22, The data processor discloses the hash value of the model or algorithm; S23, The model user requests to invoke the model or algorithm through the data network interface and pays the invocation fee; S24, the model user sends the data to be processed and the hash value to a trusted third party, and the data processor sends the model or algorithm to the trusted third party; S25, a trusted third party verifies whether the hash value of the received model or algorithm is consistent with the hash value disclosed in step S22. If they are consistent, the model or algorithm is used to calculate the data to be processed, and the calculation result is delivered to the model user.
[0050] Specifically: (a) Data processor P trains a model on data D (usually a public dataset) provided by data provider T to obtain a model or algorithm F; (b) The data processor P does not disclose the parameters of F, but discloses H, where H = Hash(F); (c) Model user U owns its own dataset D', requests data processor P to provide a call to F, and pays the call fee; (d) Model user U sends the data D' and H to be processed to a trusted third party TTP; data processor P sends F to the trusted third party TTP; TTP calculates Hash(F) to see if it is equal to H. If it is equal, it means that F is valid. Otherwise, it means that F provided by P is invalid, and the calculation is rejected. TTP is required to refund U. Calculate F(D') to obtain the result r, i.e., r=F(D'), and deliver r to U. U obtains the result of data processing, i.e., the corresponding data product.
[0051] It's important to clarify that U doesn't receive F, thus protecting P's rights; conversely, P doesn't receive D', thus protecting U's rights. TTP stands for Trusted Party, such as a computing center with its own computing power, which can use its own computing power, along with F and D', to process the final result r. U typically won't provide an incorrect H, because doing so would inevitably lead to a refund, and U, as the prospective buyer and initiator, has no incentive to cancel the transaction.
[0052] The challenge (and innovation) of the above protocol design lies in how to protect the rights and interests of data providers and data processors. This protocol is a competitive model where data processors provide models, and data providers select the best model, which is then processed locally to create the data product.
[0053] The following is a specific implementation example of this application: F is a high-accuracy image recognition algorithm or model for bridge engineering (such as box girders, piers, supports, etc. to be hoisted), with functions including recognition, classification, statistics, and regression analysis. In the algorithm and model provision section, the algorithm is shared, and the construction unit shares the hoisting data using the aforementioned protocol.
[0054] like Figure 4 As shown, the design institute's 3D drawings, such as BIM models, are shared with the bridge construction unit. Due to the confidentiality of BIM drawings, the construction unit needs to be able to view the BIM model only for the specific construction section. Figure 5 As shown, the construction unit shares the real-time hoisting data with the supervision unit to ensure project quality.
[0055] Data products: 10-year national traffic flow statistics chart, 5-year provincial highway traffic construction analysis chart, etc.
[0056] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the protection scope of this application.
Claims
1. A method for constructing an industry-trusted data space, characterized in that, Includes the following steps: A view layer is built on top of the database, which is the metadata layer obtained by secondary processing of the internal data. The database includes a business database, a data center database, or a basic dataset. The view layer processes the data in the database based on the data attributes and openness requirements to generate multiple data views. The data views are classified as public welfare data, open data that does not require privacy protection, open data that requires privacy protection, and limited open data. A data network interface is constructed, which is a function layer that processes data in the view layer and can be called internally or externally. The data network interface is connected to the view layer and is used to realize data sharing, exchange and trading through data open interface, data sharing interface, data exchange interface, data transaction interface and data model interface.
2. The method for constructing an industry-trusted data space as described in claim 1, characterized in that, The methods for constructing the view layer include at least one of table reconstruction, table import, view creation, data dump, stored procedure, data statistics table, and data index.
3. The method for constructing an industry-trusted data space as described in claim 2, characterized in that, The table reconstruction includes using the Create command to extract partial data from the table, or using anonymization technology to hide identifiers, or using k-anonymization technology to transform identifiers into k types, or using fuzzing technology to process location information; The table import includes importing data from different tables in different databases using the import command, and merging raw data from different sources; The creation of the view includes using the Create View from Tables command to create a real-time view of a portion of the data from a portion of the tables; The data dump includes moving or backing up data using the dump command; The stored procedures include function libraries of the database management system or data processing programs developed by software developers; The data statistics table includes data statistics information or data calculation results; The data index includes an index for quick data location.
4. The method for constructing an industry-trusted data space as described in claim 1, characterized in that, A function layer is built on top of the view layer. The function layer includes multiple processing functions, which are used to process the data view.
5. The method for constructing an industry-trusted data space as described in claim 4, characterized in that, The processing functions include data statistics functions F1, data privacy protection functions F2, data machine learning functions F3, data transaction functions F4, data access control functions F5, federated learning functions F6, data processing functions F7, and model processing functions F8.
6. The method for constructing an industry-trusted data space as described in claim 5, characterized in that, F1 includes one or more of summation, mean calculation, and variance calculation; F2 includes one or more of de-identification functions, de-classification functions, hierarchical functions, classification functions, privacy intersection functions, and fully homomorphic encryption calculation functions; F3 includes one or more of linear regression, logistic regression, support vector machine, random forest, and Naive Bayes; F4 includes one or more of payment, data verification, and withdrawal; F5 includes one or more of access control list filling, access control query, and access control policy definition; F6 includes sub-model submission functions and / or model parameter acquisition functions; F7 includes one or more of image processing functions, video processing functions, and model generation functions; F8 includes CNN model functions and / or RNN model functions, wherein the CNN model functions and RNN model functions take sample data as input and output model structure and parameters.
7. An application of a method for constructing an industry-trusted data space as described in any one of claims 1 to 6, characterized in that, It can be applied to the intelligent transportation industry or the natural resources industry.
8. A data processing method based on data space fairness, characterized in that, The data space is constructed using the industry trusted data space construction method as described in claim 1, and the data processing method includes the following steps: S11, the data provider divides the data to be processed into multiple parts, and opens one of them to multiple data processors through the data view of the view layer; S12, each data processor uses the open data view to conduct model learning or algorithm design, obtains its own processing method, and provides the processing result and the hash value of the corresponding processing method to the data provider; S13, the data provider selects the data processor that provides the best processing result based on the multiple processing results received, and pays a fee to the selected data processor; S14, the selected data processor encrypts the processing method using the data provider's public key and sends it to the data provider. The data provider decrypts the processing method using its private key and verifies whether the hash value of the decrypted processing method is consistent with the hash value received in step S12. S15, when the hash value verification is consistent, the data provider uses the processing method to process the data to be processed to obtain the processed data.
9. The data processing method based on data space as described in claim 8, characterized in that, The processed data is stored as a limited open static data in the data space and provides access control based on payment, scope control based on access control lists, and scope control based on access control policies.
10. A model invocation method based on data space, characterized in that, The data space is constructed using the industry trusted data space construction method as described in claim 1, and the model invocation method includes the following steps: S21, Data processors train models on public datasets provided by data providers through the view layer to obtain models or algorithms; S22, The data processor discloses the hash value of the model or algorithm; S23, The model user requests to invoke the model or algorithm through the data network interface and pays the invocation fee; S24, the model user sends the data to be processed and the hash value to a trusted third party, and the data processor sends the model or algorithm to the trusted third party; S25, a trusted third party verifies whether the hash value of the received model or algorithm is consistent with the hash value disclosed in step S22. If they are consistent, the model or algorithm is used to calculate the data to be processed, and the calculation result is delivered to the model user.