A multi-standard data rights confirmation method based on multi-classifier fusion

Abstract

The present invention provides a multi-standard data rights confirmation method based on multi-classifier fusion, comprising: determining the rights confirmation classification standard according to the "three-part principle" and determining the data attribute description of the data to be confirmed; representing the data attribute description in the form of a vector, segmenting and integrating the vector according to the rights confirmation classification standard, and using it as the input vector of the rights confirmation classification model; constructing a rights confirmation classification model including a two-level SVM classifier, and using the primary SVM classifier to classify the power subject, competitiveness, and exclusivity of the data; integrating the classification information obtained by the primary SVM classifier and inputting it into the secondary SVM classifier to obtain the data rights confirmation result. The present invention overcomes the problem of insufficient hypothesis space expression capability existing in machine learning algorithms by constructing a data attribute description of the rights confirmation data and then using the rights confirmation classification model and multi-classifier fusion technology to confirm the data, thereby improving the accuracy of classification and ensuring the security and value of the data.

Description

Technical Field

[0001] The present invention relates to the technical field of data security management, and in particular to a multi-standard data rights confirmation method based on multi-classifier fusion. Background Art

[0002] With the advent of the digital age, the volume of various data has increased dramatically. This data comes from diverse sources, complex formats, and varying quality. Furthermore, with the continuous advancement of data technology, the application of technologies such as artificial intelligence and machine learning is becoming increasingly widespread, placing ever-higher demands on data accuracy and reliability. In this context, data ownership confirmation, as a crucial data management tool, can help users identify and differentiate the value of data, ensure its integrity, credibility, and reliability, and support its application and analysis, thereby better meeting user needs. Currently, with data security issues becoming increasingly prominent, data ownership confirmation has become a crucial means of protecting data security. It can mitigate the risk of data attacks and misuse by controlling data sources, collection, transmission, storage, and processing. In recent years, much research has been conducted in the field of data ownership confirmation. However, due to issues such as unclear data sources and difficulties in data classification, data ownership confirmation remains controversial and faces multiple challenges. Summary of the Invention

[0003] In order to solve the above technical problems, the purpose of the present invention is to provide a multi-standard data rights confirmation method based on multi-classifier fusion. By constructing a data attribute description of the rights confirmation data, and then using the rights confirmation classification model and multi-classifier fusion technology to confirm the data, the problem of insufficient hypothesis space expression ability existing in machine learning algorithms is overcome, the accuracy of classification is improved, and the security and value of the data are guaranteed.

[0004] The present invention provides a multi-standard data rights confirmation method based on multi-classifier fusion, comprising:

[0005] Step 1: Determine the classification standards for data rights confirmation based on the "three-part principle" of data rights confirmation, and determine the data attribute description of the data to be confirmed based on the classification standards;

[0006] Step 2: Represent the data attribute description in the form of a vector, split and integrate the vector according to different property rights classification standards, and use it as the input vector of multiple classifiers in the property rights classification model;

[0007] Step 3: Construct a property rights classification model consisting of three primary SVM classifiers and one secondary SVM classifier, and use the trained primary SVM classifier to classify the power subject, competitiveness, and exclusivity of the data;

[0008] Step 4: Through multi-classifier fusion technology, the classification information obtained by the three primary SVM classifiers is integrated, and the integrated classification information is input into the secondary SVM classifier to obtain the types of property rights owned by the data right subject and realize multi-standard classification.

[0009] Furthermore, the step 1 is specifically as follows:

[0010] Step 1.1: Determine the classification criteria for data ownership confirmation based on the "three-part principle" of segmentation, classification, and grading;

[0011] Step 1.2: Determine the data attribute description of the property rights confirmation data according to the property rights confirmation classification standards.

[0012] Furthermore, the classification criteria for title confirmation in step 1.1 are:

[0013] (1) According to the principle of segmentation of data rights confirmation, data property rights are divided into three types of property rights: basic data property rights, derived data property rights and data public property rights;

[0014] (2) According to the classification principle of data rights confirmation, data rights subjects are divided into: personal data, enterprise data and social data according to the different data subjects;

[0015] (3) Based on the hierarchical principle of data rights confirmation, data are divided into different levels according to competitiveness and exclusivity, and data are divided into private goods, quasi-public goods and public goods.

[0016] Further:

[0017] Basic data property rights include: management rights, security rights, transfer rights, modification and correction rights, right to be forgotten, right to informed consent, portability rights, income rights, and control rights;

[0018] Derivative data property rights include: use rights, residual control rights, and residual claim rights;

[0019] Public ownership of data includes: control rights, open rights, and management rights.

[0020] Furthermore, the data attribute description of the ownership confirmation data in step 1.2 includes the following conditional attributes:

[0021] The direct provider of the data; whether private data is involved, including but not limited to personal name, telephone number, address, occupation, education, preferences, habits, travel history, shopping history, and browsing history; whether the use of the data is legal or authorized; how the data is generated; whether the data involves natural resource data, economic and social data; whether the use of the data will generate significant external benefits; whether the dissemination of the data will reduce the effectiveness of the data or increase costs; whether the data involves relevant core interests; whether there are different permissions for internal and external personnel in the use of the data; and whether there is an agreement that the data content cannot be disclosed or is disclosed within a limited time.

[0022] Furthermore, the step 2 is specifically as follows:

[0023] Step 2.1: Map the data attribute descriptions entered by the user into discrete data and construct a vector. Set three options for the conditional attribute "Who is the direct provider of the data?": "Individual," "Enterprise," and "Government." Map "Individual" to 1, "Enterprise" to 2, and "Government" to 3.

[0024] Step 2.2: Set three options for the conditional attribute "Data Generation Method": "Natural Person Behavior", "Production and Operation", and "Collection According to Law". Map "Natural Person Behavior" to 1, "Production and Operation" to 2, and "Collection According to Law" to 3.

[0025] Step 2.3: Set three options for other condition attributes: "Yes", "No", and "Unclear". "Yes" is mapped to 1, "No" is mapped to 2, and "Unclear" is mapped to 3.

[0026] Step 2.4: Split and integrate the vectors according to different property rights classification standards, and use them as input vectors for multiple classifiers of the property rights classification model.

[0027] Furthermore, step 3 is specifically as follows:

[0028] Step 3.1: Construct a data ownership classification model consisting of three primary SVM classifiers and one secondary SVM classifier, and then use the three primary SVM classifiers to classify the data's ownership, competitiveness, and exclusivity.

[0029] Step 3.2: Train multiple SVM classifiers. Use the Gaussian kernel function to map the data in the low-dimensional space to the high-dimensional space, making the originally linearly inseparable data linearly separable and finding the optimal hyperplane in the high-dimensional space for classification. Obtain the parameters of each SVM classifier by solving the following convex optimization problem;

[0030]

[0031] Among them, xi is the conditional attribute vector of the i-th training data, y i is x i The corresponding category labels, w and b are the parameters of the SVM classifier;

[0032] Step 3.3: Train the SVM classifiers using a large number of labeled datasets until the accuracy of each SVM classifier reaches the required standard for the classification task.

[0033] Step 3.4: Input the input vectors obtained by segmenting and integrating the vectors according to different rights classification standards into the corresponding primary SVM classifier to classify the power subject, competitiveness, and exclusivity of the data.

[0034] Furthermore, the step 4 includes:

[0035] Step 4.1: Integrate the classification information of the three primary SVM classifiers as the input of the secondary SVM classifier;

[0036] Step 4.2: The data rights confirmation classification model outputs the results of data rights confirmation, that is, which property rights of the three types of property rights the data rights subject owns. The results of data rights confirmation include eight categories.

[0037] The multi-standard data rights confirmation method based on multi-classifier fusion of the present invention has at least the following beneficial effects:

[0038] The present invention determines the classification standard for data rights confirmation based on the "three-part principle" of data rights confirmation, determines the data attribute description of the rights confirmation data based on the classification standard, and maps the data attribute description to discrete data to form the input vector of the rights confirmation classification model. The constructed rights confirmation classification model is used to classify the rights subject, the competitiveness and exclusivity of the data, and finally outputs the final rights confirmation result through multi-classifier fusion technology. Using the multi-standard data rights confirmation method of the present invention, users do not need to read a large amount of relevant documents on data rights confirmation standards. They only need to provide relevant data attribute descriptions to automatically confirm data rights. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 This is a flow chart of a multi-standard data rights confirmation method based on multi-classifier fusion of the present invention;

[0040] Figure 2 It is a data rights map. DETAILED DESCRIPTION

[0041] like Figure 1 As shown, the present invention provides a multi-standard data rights confirmation method based on multi-classifier fusion, comprising:

[0042] Step 1: Determine the data rights classification standard based on the "three-part principle" of data rights confirmation, and determine the data attribute description of the input data to be confirmed based on the data rights classification standard. Step 1 is specifically as follows:

[0043] Step 1.1: Determine the classification standards for data rights confirmation based on the "three-part principle" of segmentation, classification, and grading.

[0044] In specific implementation, the classification standards for property rights confirmation are:

[0045] (1) According to the principle of data ownership division, data ownership is divided into three types: basic data ownership, derivative data ownership and data public ownership. Figure 2 shown.

[0046] Basic data property rights include: management rights, security rights, transfer rights, modification and correction rights, right to be forgotten, right to informed consent, portability rights, income rights, and control rights.

[0047] Derivative data property rights include: right of use, residual control right, and residual claim right.

[0048] Public ownership of data includes: control rights, open rights, and management rights.

[0049] (2) According to the classification principle of data rights confirmation, data rights subjects are divided into: personal data, corporate data and social data according to the different data subjects.

[0050] (3) Based on the hierarchical principle of data rights confirmation, data are divided into different levels according to competitiveness and exclusivity, and data are divided into private goods, quasi-public goods and public goods.

[0051] Step 1.2: Determine the data attribute description of the property rights confirmation data according to the property rights confirmation classification standards.

[0052] In specific implementation, the data attribute description of the ownership confirmation data includes the following conditional attributes:

[0053] 1) Direct provider of data, including three options: "Individual", "Enterprise", and "Government";

[0054] 2) Whether private data is involved, including but not limited to personal name, phone number, address, occupation, education, preferences, habits, travel history, shopping history, browsing history, etc.; including three options: "Yes", "No", "Unclear";

[0055] 3) Whether the use of the data is legal or authorized, including three options: "Yes", "No", "Unclear";

[0056] 4) How the data is generated, including three options: "Natural person behavior", "Production and operation", and "Collected in accordance with the law";

[0057] 5) Whether the data involves natural resource data, economic and social data, including three options: "Yes", "No", "Unclear";

[0058] 6) Whether the use of the data will generate significant external benefits, including three options: "Yes", "No", "Unclear";

[0059] 7) Will the dissemination of data reduce the benefits of the data or increase costs? There are three options: "Yes", "No", and "Unclear";

[0060] 8) Whether the data involves relevant core interests, including three options: "Yes", "No", "Unclear";

[0061] 9) Whether internal and external personnel have different permissions to use data, including three options: "Yes", "No", "Unclear";

[0062] 10) Is there an agreement that the data content cannot be disclosed or is disclosed within a limited time? There are three options: "Yes", "No", and "Unclear".

[0063] Step 2: Represent the data attribute description in the form of a vector, segment and integrate the vector according to different property rights classification standards, and use it as the input vector of multiple classifiers of the property rights classification model. Step 2 is specifically as follows:

[0064] Step 2.1: Map the data attribute descriptions entered by the user into discrete data and construct a vector. Set three options for the conditional attribute "direct provider of data": "individual", "enterprise", and "government". Map "individual" to 1, "enterprise" to 2, and "government" to 3.

[0065] Step 2.2: Set three options for the conditional attribute "Data Generation Method": "Natural Person Behavior", "Production and Operation", and "Collection According to Law". Map "Natural Person Behavior" to 1, "Production and Operation" to 2, and "Collection According to Law" to 3.

[0066] Step 2.3: Set three options for other condition attributes: "Yes", "No", and "Unclear". "Yes" is mapped to 1, "No" is mapped to 2, and "Unclear" is mapped to 3.

[0067] In specific implementation, if the 10 conditional attributes input by the user are: "individual", "no", "unclear", "natural person behavior", "no", "unclear", "yes", "yes", "no", "unclear", then an original vector such as [1,2,3,1,2,3,1,1,2,3] is generated according to the conditional attribute mapping input by the user.

[0068] Step 2.4: Split and integrate the vectors according to different property rights classification standards, and use them as input vectors for multiple classifiers of the property rights classification model.

[0069] In specific implementation, the present invention uses three primary SVM classifiers to classify the power subject, competitiveness, and exclusivity of the data. For the No. 1 primary SVM classifier, which is used to classify the power subject, the input conditional attributes required include: the direct provider of the data; whether it involves privacy data; whether the use of the data is legal or authorized; the way the data is generated; and whether the data involves natural resource data or economic and social data. The vector is segmented, and these conditional attributes are segmented and spliced ​​into a new input vector as the input of the No. 1 classifier. Specifically, the conditional attributes of the first 5 dimensions are segmented, and a vector such as [1,2,3,1,2] is segmented.

[0070] For the second primary SVM classifier, which is used to classify competitiveness, the required input attributes include: the data's direct provider; whether the data involves natural resource data or economic and social data; whether the dissemination of the data will reduce its benefits or increase costs; and whether the data involves relevant core interests. The original vector is then segmented, and these attributes are concatenated into a new vector as input to the second primary SVM classifier. Specifically, the 1-, 5-, 7-, and 8-dimensional conditional attributes are segmented, resulting in a vector such as [1, 2, 1, 1].

[0071] The primary SVM classifier #3 is used to classify exclusivity. The required input attributes include: whether private data is involved; whether the data involves relevant core interests; whether there are different permissions for internal and external personnel using the data; and whether there are agreements that prohibit public disclosure or limit public disclosure of data content. The original vector is then segmented, and these attributes are concatenated into a new vector as input to the primary SVM classifier #3. Specifically, the 1-, 8-, 9-, and 10-dimensional conditional attributes are segmented, resulting in a vector such as [1, 1, 2, 3].

[0072] Step 3: Construct a property rights classification model consisting of three primary SVM classifiers and one secondary SVM classifier, and use the trained primary SVM classifier to classify the power subject, competitiveness, and exclusivity of the data. Step 3 is as follows:

[0073] Step 3.1: Construct a data ownership classification model consisting of three primary SVM classifiers and one secondary SVM classifier, and then use the three primary SVM classifiers to classify the data's ownership, competitiveness, and exclusivity.

[0074] Step 3.2: Train multiple SVM classifiers. Use the Gaussian kernel function to map the data in the low-dimensional space to the high-dimensional space, making the originally linearly inseparable data linearly separable and finding the optimal hyperplane in the high-dimensional space for classification. Obtain the parameters of each SVM classifier by solving the following convex optimization problem;

[0075]

[0076] Among them, x i is the conditional attribute vector of the i-th training data, y i is x i The corresponding category labels, w and b are the parameters of the SVM classifier;

[0077] Step 3.3: Train the SVM classifiers using a large number of labeled datasets until the accuracy of each SVM classifier reaches the required standard for the classification task.

[0078] Step 3.4: Input the input vectors obtained by segmenting and integrating the vectors according to different rights classification standards into the corresponding primary SVM classifier to classify the power subject, competitiveness, and exclusivity of the data.

[0079] In specific implementation, the input vector [1,2,3,1,2] obtained from the segmentation in step 2.4 is input into the primary SVM classifier #1 to classify the power subject (including individuals, enterprises, and society). The input vector [1,2,1,1] is input into the primary SVM classifier #2 to classify the competitiveness of the data (including competition, non-competition, and partial competition). The input vector [1,1,2,3] is input into the primary SVM classifier #3 to classify the exclusivity of the data (including exclusive, non-exclusive, and limited exclusivity).

[0080] Step 4: Using multi-classifier fusion technology, the classification information obtained by the three primary SVM classifiers is integrated, and the integrated classification information is input into the secondary SVM classifier to obtain the types of property rights owned by the data rights holder, thereby achieving multi-standard classification. Step 4 includes:

[0081] Step 4.1: Integrate the classification information of the three primary SVM classifiers as the input of the secondary SVM classifier;

[0082] During the specific implementation, the classification information obtained by primary SVM classifiers No. 1, 2, and 3 is integrated. Primary SVM classifier No. 1 obtains the power subjects of the data, including: individuals, enterprises, and society. "Individual" is mapped to 1, "enterprise" is mapped to 2, and "society" is mapped to 3.

[0083] The No. 2 primary SVM classifier obtains the competitiveness of the data, including: competitive, non-competitive, and partially competitive, mapping "competitive" to 1, "non-competitive" to -1, and "partially competitive" to 0.

[0084] The primary SVM classifier #3 obtains the exclusivity of the data, including exclusive, non-exclusive, and partially exclusive, mapping "exclusive" to 1, "non-exclusive" to -1, and "partially exclusive" to 0. Finally, a new vector is generated based on the classification results as the input to the secondary SVM classifier #4.

[0085] Step 4.2: The data rights confirmation classification model outputs the results of data rights confirmation, that is, which property rights of the three types of property rights the data rights subject owns.

[0086] Use the secondary SVM classifier for multi-standard classification. The vector generated in step 4.1 is used as the input vector of the secondary SVM classifier, and then input into the secondary SVM classifier to obtain the final classification result. The final classification result is a result containing one or more labels. For example, when the power subject is an individual, the competitiveness of the data is competitive, and the exclusivity of the data is exclusive, the data is a private product, and the final classification result is: basic data property rights. When the power subject is an individual, the competitiveness of the data is non-competitive, and the exclusivity of the data is exclusive, the data is a quasi-private product, and the final classification result is basic data property rights and public property rights. The results of data property rights confirmation include eight categories. As shown in Table 1.

[0087] Table 1 Data ownership classification results

[0088]

[0089] The above description is only a preferred embodiment of the present invention and is not intended to limit the concept of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A multi-standard data rights confirmation method based on multi-classifier fusion, characterized by: include: Step 1: Determine the data rights classification standards based on the "three-part principle" of data rights confirmation, and determine the data attribute description of the data to be confirmed based on the data rights classification standards; Step 2: Represent the data attribute description in the form of a vector, split and integrate the vector according to different property rights classification standards, and use it as the input vector of multiple classifiers in the property rights classification model; Step 3: Construct a property rights classification model consisting of three primary SVM classifiers and one secondary SVM classifier, and use the trained primary SVM classifier to classify the power subject, competitiveness, and exclusivity of the data; Step 4: Through multi-classifier fusion technology, the classification information obtained by the three primary SVM classifiers is integrated, and the integrated classification information is input into the secondary SVM classifier to obtain the types of property rights owned by the data right subject and realize multi-standard classification.

2. The multi-standard data rights confirmation method based on multi-classifier fusion according to claim 1 is characterized in that: The step 1 is specifically as follows: Step 1.1: Determine the classification criteria for data ownership confirmation based on the "three-part principle" of segmentation, classification, and grading; Step 1.2: Determine the data attribute description of the property rights confirmation data according to the property rights confirmation classification standards.

3. The multi-standard data rights confirmation method based on multi-classifier fusion according to claim 2 is characterized in that: The classification criteria for title confirmation in step 1.1 are: (1) According to the principle of segmentation of data rights confirmation, data property rights are divided into three types of property rights: basic data property rights, derived data property rights and data public property rights; (2) According to the classification principle of data rights confirmation, data rights subjects are divided into: personal data, enterprise data and social data according to the different data subjects; (3) Based on the hierarchical principle of data rights confirmation, data are divided into different levels according to competitiveness and exclusivity, and data are divided into private goods, quasi-public goods and public goods.

4. The multi-standard data rights confirmation method based on multi-classifier fusion according to claim 3 is characterized by: Basic data property rights include: management rights, security rights, transfer rights, modification and correction rights, right to be forgotten, right to informed consent, portability rights, income rights, and control rights; Derivative data property rights include: use rights, residual control rights, and residual claim rights; Public ownership of data includes: control rights, open rights, and management rights.

5. The multi-standard data rights confirmation method based on multi-classifier fusion according to claim 2 is characterized in that: The data attribute description of the ownership confirmation data in step 1.2 includes the following conditional attributes: The direct provider of the data; whether private data is involved, including personal name, telephone number, address, occupation, education, preferences, habits, travel records, shopping records, and browsing history; whether the use of the data is legal or authorized; how the data is generated; whether the data involves natural resource data, economic and social data; whether the use of the data will generate external benefits; whether the dissemination of the data will reduce the effectiveness of the data or increase costs; whether the data involves relevant core interests; whether the use of the data has different permissions for internal and external personnel; and whether there is an agreement that the data content cannot be disclosed or is disclosed within a limited time.

6. The multi-standard data rights confirmation method based on multi-classifier fusion according to claim 5 is characterized in that: The step 2 is specifically as follows: Step 2.1: Map the user-entered data attribute description into discrete data and construct a vector. For the conditional attribute "Who is the direct provider of the data?", set three options: "Individual," "Enterprise," and "Government." Map "Individual" to 1, "Enterprise" to 2, and "Government" to 3. Step 2.2: Set three options for the conditional attribute "Data Generation Method": "Natural Person Behavior," "Production and Operation," and "Collection According to Law." Map "Natural Person Behavior" to 1, "Production and Operation" to 2, and "Collection According to Law" to 3. Step 2.3: Set three options for other condition attributes: "Yes", "No", and "Unclear". "Yes" is mapped to 1, "No" is mapped to 2, and "Unclear" is mapped to 3. Step 2.4: Split and integrate the vectors according to different property rights classification standards, and use them as input vectors for multiple classifiers of the property rights classification model.

7. The multi-standard data rights confirmation method based on multi-classifier fusion according to claim 1 is characterized in that: Step 3 is as follows: Step 3.1: Construct a data ownership classification model consisting of three primary SVM classifiers and one secondary SVM classifier, and then use the three primary SVM classifiers to classify the data's ownership, competitiveness, and exclusivity. Step 3.2: Train multiple SVM classifiers. Use the Gaussian kernel function to map the data in the low-dimensional space to the high-dimensional space, making the originally linearly inseparable data linearly separable and finding the optimal hyperplane in the high-dimensional space for classification. Obtain the parameters of each SVM classifier by solving the following convex optimization problem; Among them, x i is the conditional attribute vector of the i-th training data, y i is x i The corresponding category labels, w and b are the parameters of the SVM classifier; Step 3.3: Train the SVM classifiers using a large number of labeled datasets until the accuracy of each SVM classifier reaches the required standard for the classification task. Step 3.4: Input the input vectors obtained by segmenting and integrating the vectors according to different rights classification standards into the corresponding primary SVM classifier to classify the power subject, competitiveness, and exclusivity of the data.

8. The multi-standard data rights confirmation method based on multi-classifier fusion according to claim 1 is characterized in that: The step 4 comprises: Step 4.1: Integrate the classification information of the three primary SVM classifiers as the input of the secondary SVM classifier; Step 4.2: The data rights confirmation classification model outputs the results of data rights confirmation, that is, which property rights of the three types of property rights the data rights subject owns. The results of data rights confirmation include eight categories.

Images