Data grading method and device

Abstract

The invention discloses a data grading method and device. When the method is executed, firstly, a to-be-graded data set is obtained; combining the attributes according to the association degree between the attributes to obtain a plurality of combined attributes; performing hierarchical calculation on each combined attribute by using a pre-constructed hierarchical algorithm to obtain a joint hierarchical vector of each combined attribute; and finally, determining a sensitivity level corresponding to each combined attribute according to the joint grading vector of each combined attribute, and taking the sensitivity level of each combined attribute as a grading result of the data set to be graded. Through the above mode, not only is the sensitivity degree of a single attribute considered, but also the privacy of inference information after multi-attribute association is considered, so that more comprehensive classification is performed. Moreover, compared with a qualitative analysis model, the method has the advantages that the combination attributes are subjected to hierarchical calculation through a pre-constructed hierarchical algorithm, and the hierarchical result of the multiple combination attributes can be quantitatively described, so that the accuracy of the data hierarchical result is improved.

Description

Technical Field

[0001] The present application relates to the field of data processing, and in particular to a data classification method and device. Background Art

[0002] Many institutions or organizations have a large amount of raw data that needs to be published or shared. However, before the data is published, it is necessary to analyze and classify the data in order to better conduct privacy assessment.

[0003] The current attribute classification algorithms are often qualitative analyses, that is, they only provide a model for attribute association analysis without involving a quantitative process. The data classification results are highly subjective, resulting in low accuracy of the data classification results. Summary of the invention

[0004] In view of this, the present application provides a data classification method and device, aiming to improve the accuracy of data classification results.

[0005] In a first aspect, the present application provides a data classification method, the method comprising:

[0006] Acquire a data set to be classified, the data set comprising a plurality of data and an attribute corresponding to each data;

[0007] Combine according to the correlation degree between each attribute to obtain multiple combined attributes;

[0008] Using a grading algorithm to perform grading calculation on each of the multiple combined attributes to obtain a joint grading vector of each combined attribute;

[0009] According to the joint classification vector of each combined attribute, the sensitivity level corresponding to each combined attribute is determined, and the sensitivity level of each combined attribute is used as the classification result of the data set to be classified.

[0010] Optionally, the construction process of the grading algorithm includes:

[0011] Acquire a sample data set, where the sample data set includes a plurality of sample data and an attribute corresponding to each sample data;

[0012] Determining a sensitivity level of each sample data attribute;

[0013] Quantitatively mapping the sensitivity level of each sample data attribute to obtain an index value corresponding to each sample data attribute;

[0014] Using a single attribute classification algorithm, according to the index value corresponding to each sample data attribute, a classification probability vector of each sample data attribute is calculated;

[0015] According to the contribution value of each sample data attribute to the privacy amount, set the weight value corresponding to each sample data attribute;

[0016] The classification probability vector of each sample data attribute and the weight value corresponding to each sample data attribute are jointly classified to obtain a joint classification vector of each sample data combination attribute.

[0017] Optionally, using a grading algorithm to perform grading calculation on each of the multiple combined attributes to obtain a joint grading vector of each combined attribute includes:

[0018] Determining a sensitivity level of each attribute in each of the combined attributes;

[0019] Quantitatively mapping the sensitivity level of each attribute in each combined attribute to obtain an index value corresponding to each attribute in each combined attribute;

[0020] Using a single attribute grading algorithm, according to the index value corresponding to each attribute in each combination of attributes, a grading probability vector of each attribute in each combination of attributes is calculated;

[0021] According to the contribution value of each attribute in each combined attribute to the privacy amount, a weight value corresponding to each attribute in each combined attribute is set;

[0022] The classification probability vector of each attribute in each combination of attributes and the weight value corresponding to each attribute in each combination of attributes are jointly classified to obtain a joint classification vector of each combination of attributes.

[0023] Optionally, determining the sensitivity level of each sample data attribute includes:

[0024] Determining an attribute value of each of the sample data attributes;

[0025] According to the distance between the attribute values ​​of each sample data attribute, each sample data attribute is graded to obtain the sensitivity level of each sample data attribute.

[0026] Optionally, the calculating, according to the index value corresponding to each sample data attribute, a classification probability vector of each sample data attribute includes:

[0027] Determining a grading index vector for each of the sample data attributes according to the index value corresponding to each of the sample data attributes;

[0028] According to the classification index vector of each sample data attribute, a classification probability vector of each sample data attribute is calculated.

[0029] Optionally, determining a corresponding sensitivity level according to the joint classification vector of each combined attribute includes:

[0030] Determining the maximum value in the joint ranking vector of each combined attribute respectively;

[0031] The sensitivity level to which the maximum value in the joint classification vector of each combined attribute belongs is used as the sensitivity level corresponding to each combined attribute.

[0032] Optionally, the sensitivity levels include: an insensitive level, a general sensitive level, a relatively sensitive level and a very sensitive level.

[0033] In a second aspect, the present application provides a data classification device, the device comprising:

[0034] An acquisition module, used to acquire a data set to be classified, wherein the data set includes a plurality of data and an attribute corresponding to each data;

[0035] A combination module is used to combine attributes according to the degree of association between each attribute to obtain multiple combined attributes;

[0036] A calculation module, configured to perform classification calculations on the multiple combined attributes respectively using a classification algorithm to obtain a joint classification vector of each combined attribute in the multiple combined attributes;

[0037] The grading module is used to determine the corresponding sensitivity level according to the joint grading vector of each combined attribute, and use the sensitivity level of each combined attribute as the grading result of the data set to be graded.

[0038] Optionally, the grading module is specifically used to:

[0039] Determining the maximum value in the joint ranking vector of each combined attribute respectively;

[0040] The sensitivity level to which the maximum value in the joint classification vector of each combined attribute belongs is used as the sensitivity level corresponding to each combined attribute.

[0041] Optionally, the sensitivity levels include: an insensitive level, a general sensitive level, a relatively sensitive level and a very sensitive level.

[0042] In a third aspect, an embodiment of the present application provides an electronic device, the electronic device comprising:

[0043] A memory for storing one or more programs;

[0044] Processor; when the one or more programs are executed by the processor, the data classification method described in any one of the first aspects above is implemented.

[0045] In a fourth aspect, an embodiment of the present application provides a computer storage medium, in which a program is stored. When the program is executed by a processor, the data classification method described in any one of the first aspects is implemented.

[0046] The above technical solution has the following beneficial effects:

[0047] The present application provides a data classification method and device. When executing the method, first obtain a data set to be classified, wherein the data set includes multiple data and attributes corresponding to each data; then combine the attributes according to the degree of association between each attribute to obtain multiple combined attributes; then use a pre-constructed classification algorithm to perform classification calculation on each combined attribute in the multiple combined attributes to obtain a joint classification vector of each combined attribute; finally, according to the joint classification vector of each combined attribute, determine the sensitivity level corresponding to each combined attribute, and use the sensitivity level of each combined attribute as the classification result of the data set to be classified. In the above manner, the present application not only considers the sensitivity of a single attribute, but also considers the privacy of information inferred after multiple attributes are associated, so as to perform a more comprehensive classification. Moreover, compared with the qualitative analysis model, the present application uses a pre-constructed classification algorithm to perform classification calculation on the combined attributes, which can quantitatively describe the classification results of multiple combined attributes, thereby improving the accuracy of the data classification results. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the technical solutions in this embodiment or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0049] Figure 1 A flow chart of a data classification method provided in an embodiment of the present application;

[0050] Figure 2 A flow chart of a classification algorithm for constructing a classification algorithm provided in an embodiment of the present application;

[0051] Figure 3 A schematic diagram of the structure of a data classification device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0052] The technical scheme in the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiment of the present invention. Obviously, the described embodiment is only a part of the embodiment of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0053] In order to facilitate further understanding of the technical solution provided by this application, the background technology involved in this application will be explained below.

[0054] The existing classification algorithm is based on the sensitivity of a single sensitive attribute. For example, if we only process the attribute "income" and use "income power" as the classification indicator, assuming that the higher the income, the more sensitive it is, the corresponding classification values ​​are "monthly income of 30,000" > "monthly income of 10,000" > "monthly income of 5,000".

[0055] In some scenarios, a single sensitive attribute may have a low sensitivity, but multiple sensitive attributes with low values ​​may be combined to expose more information through association analysis, making the data more sensitive.

[0056] However, existing attribute grading algorithms are often targeted at a single attribute, that is, they grade a certain sensitive attribute without considering the risk of combining and inferring information between attributes, the overall risk grading of data items, or the application scenarios of marketer attacks, etc., and thus have certain limitations.

[0057] Moreover, in the study of attribute association analysis, the existing methods are basically qualitative analysis, which provides an attribute association analysis model without involving the quantification process and lacking the corresponding quantification algorithm. The results are highly subjective, resulting in low accuracy of data classification results.

[0058] In order to overcome the above technical problems, the present application provides a data classification method and device. When executing the method, first obtain a data set to be classified, wherein the data set includes multiple data and attributes corresponding to each data; then combine them according to the degree of association between each attribute to obtain multiple combined attributes; then use a pre-constructed classification algorithm to perform classification calculation on each combined attribute in the multiple combined attributes to obtain a joint classification vector of each combined attribute; finally, according to the joint classification vector of each combined attribute, determine the sensitivity level corresponding to each combined attribute, and use the sensitivity level of each combined attribute as the classification result of the data set to be classified. In the above manner, the present application not only considers the sensitivity of a single attribute, but also considers the privacy of information inferred after multiple attributes are associated, so as to perform a more comprehensive classification. Moreover, compared with the qualitative analysis model, the present application uses a pre-constructed classification algorithm to perform classification calculation on the combined attributes, which can quantitatively describe the classification results of multiple combined attributes, thereby improving the accuracy of the data classification results.

[0059] Moreover, the classification results can reflect the sensitivity of the new information inferred after the association analysis between attributes, and can provide analytical guidance for the marketer attack scenario. For example: in the marketer attack process, the attacker will use the combination relationship between attributes to infer sensitive information, thereby leaking user privacy. For the results after the attribute association analysis, the more sensitive the information is, the more attention it needs to pay before publishing, the more generalized the processing is, and the higher the security threshold is set, such as a higher degree of anonymization, and sensitive attributes are processed as much as possible without affecting the availability of data; for less sensitive attributes, the threshold can be set lower, the degree of generalization is lower, and the availability of data is better preserved.

[0060] See also Figure 1 , Figure 1 A flow chart of a data classification method provided in an embodiment of the present application. The method of the embodiment of the present invention can be executed by a data classification device, which can be implemented by software and / or hardware and can generally be integrated in a server or terminal device.

[0061] The method may include:

[0062] Step S101: obtaining a data set to be classified, wherein the data set includes a plurality of data and attributes corresponding to each data.

[0063] Specifically, the data classification device first obtains a data set to be classified, wherein the data set includes a plurality of data and an attribute corresponding to each data.

[0064] Step S102: Combining attributes according to the degree of association between each attribute to obtain a plurality of combined attributes.

[0065] In this embodiment, after the data set to be classified is obtained in step S101, further, each attribute is combined according to the degree of association between them, so as to obtain a plurality of combined attributes.

[0066] It is understandable that the combination of attributes is based on the degree of association between the attributes. For example, in a credit card repayment data set, the credit limit, bill amount, payment history, and default status can be combined to analyze the user's financial status, such as capital turnover, loan credit status, etc.

[0067] It should be noted that combining attributes may include combining multiple attributes, for example, three or more sensitive attributes may be combined. Therefore, this application does not limit the number of single attributes in the combined attributes, which mainly depends on the attribute association analysis of the specific data set.

[0068] Step S103: using a grading algorithm to perform grading calculation on each of the multiple combined attributes to obtain a joint grading vector of each combined attribute.

[0069] In this step, after multiple combined attributes are obtained in step S102, a pre-built classification algorithm is used to perform classification calculation on each of the multiple combined attributes, thereby obtaining a joint classification vector of each combined attribute.

[0070] The construction process of the grading algorithm will be introduced in detail later, so I will not go into details here.

[0071] It can be understood that compared with the qualitative analysis model, the embodiment of the present application performs grading calculations on combined attributes through a pre-built grading algorithm, which can quantitatively describe the grading results of multiple combined attributes, thereby improving the accuracy of data grading results.

[0072] Step S104: Determine the sensitivity level corresponding to each combined attribute according to the joint classification vector of each combined attribute, and use the sensitivity level of each combined attribute as the classification result of the data set to be classified.

[0073] In this embodiment, after obtaining the joint grading vector of each combined attribute through step S103, the sensitivity level corresponding to each combined attribute is determined according to the joint grading vector of each combined attribute, and the sensitivity level of each combined attribute is further used as the grading result of the data set to be graded.

[0074] It is understandable that the embodiments of the present application can flexibly specify the combination mode and weight setting of each sensitive attribute, and can modify the parameters after analyzing different application scenarios to obtain more accurate classification results.

[0075] In a possible implementation, step S104 specifically includes:

[0076] Determining the maximum value in the joint ranking vector of each combined attribute respectively;

[0077] The sensitivity level to which the maximum value in the joint classification vector of each combined attribute belongs is used as the sensitivity level corresponding to each combined attribute.

[0078] In this embodiment, firstly, the maximum value in the joint classification vector of each combined attribute is determined respectively, and the sensitivity level to which the maximum value in the joint classification vector of each combined attribute belongs is used as the sensitivity level corresponding to each combined attribute.

[0079] In one possible implementation, the sensitivity levels may include: an insensitive level, a general sensitive level, a relatively sensitive level, and a very sensitive level.

[0080] It should be noted that the sensitivity level can be set according to actual conditions and is not limited in this application.

[0081] For example: The pre-sensitivity levels include {insensitive, moderately sensitive, relatively sensitive, very sensitive}.

[0082] The joint grading vector of a certain combination of attributes is calculated to be pro_set = [0.2, 0.3, 0.4, 0.1]. The maximum value in the joint grading vector of the combination of attributes is 0.4, and its sensitivity level is [relatively sensitive]. The sensitivity level to which the maximum value in the joint grading vector of the combination of attributes belongs is [relatively sensitive], which is taken as the sensitivity level corresponding to the combination of attributes.

[0083] It can be seen from the above technical scheme that the embodiment of the present application first obtains a data set to be classified, wherein the data set includes multiple data and attributes corresponding to each data; then, the attributes are combined according to the degree of association between each attribute to obtain multiple combined attributes; then, a pre-constructed grading algorithm is used to perform grading calculations on each of the multiple combined attributes to obtain a joint grading vector of each combined attribute; finally, according to the joint grading vector of each combined attribute, the sensitivity level corresponding to each combined attribute is determined, and the sensitivity level of each combined attribute is used as the grading result of the data set to be graded. In the above manner, the present application not only considers the sensitivity of a single attribute, but also considers the privacy of the information inferred after multiple attributes are associated, so as to perform a more comprehensive grading. Moreover, compared with the qualitative analysis model, the present application can quantitatively describe the grading results of multiple combined attributes by performing grading calculations on the combined attributes through a pre-constructed grading algorithm, thereby improving the accuracy of the data grading results.

[0084] See also Figure 2, is a flow chart of constructing a grading algorithm provided in an embodiment of the present application. In a possible implementation, the construction process of the grading algorithm may include:

[0085] Step S201: Acquire a sample data set, where the sample data set includes a plurality of sample data and attributes corresponding to each sample data.

[0086] In this embodiment, when constructing a grading algorithm, a sample data set is first obtained, where the sample data set includes a plurality of sample data and an attribute corresponding to each sample data.

[0087] Step S202: Determine the sensitivity level of each sample data attribute.

[0088] It is understandable that the sensitivity level of sample data attributes can be divided through subjective analysis methods; the classification model can also be defined through information entropy to calculate the sensitivity level of sample data attributes, which is not limited in this application.

[0089] Step S203: quantize and map the sensitivity level of each sample data attribute to obtain an index value corresponding to each sample data attribute.

[0090] Step S204: using a single attribute classification algorithm, according to the index value corresponding to each sample data attribute, a classification probability vector of each sample data attribute is calculated.

[0091] In this step, after obtaining the index value corresponding to each sample data attribute in step S203, a single attribute grading algorithm is used to calculate the graded probability vector of each sample data attribute according to the index value corresponding to each sample data attribute, wherein the sum of the probability values ​​of each vector in the graded probability vector of the sample data attribute is equal to 1.

[0092] Step S205: according to the contribution value of each sample data attribute to the privacy amount, a weight value corresponding to each sample data attribute is set.

[0093] In this step, a corresponding weight value is assigned to each sample data attribute according to its contribution to the privacy amount.

[0094] Step S206: performing joint classification calculation on the classification probability vector of each sample data attribute and the weight value corresponding to each sample data attribute to obtain a joint classification vector of each sample data combination attribute.

[0095] In this step, after the hierarchical probability vector of each sample data attribute is calculated through step S204 and the weight value corresponding to each sample data attribute is set through step S205, the hierarchical probability vector of each sample data attribute and the weight value corresponding to each sample data attribute are further jointly hierarchically calculated to obtain a joint hierarchical vector of each sample data combination attribute.

[0096] It can be understood that compared with the qualitative analysis model, the embodiment of the present application performs grading calculations on combined attributes through a pre-built grading algorithm, which can quantitatively describe the grading results of multiple combined attributes, thereby improving the accuracy of data grading results.

[0097] In practical applications, for each piece of data in the data set, the classification space S_sv of a single attribute sensitive value is set to be equal to {sensitivity of the sensitive value}.

[0098] It can be divided into 4 sensitivity levels: S_sv = {SV_1, SV_2, SV_3, SV_4} = {insensitive, moderately sensitive, relatively sensitive, very sensitive}. The sensitivity level obtained after analyzing the combination of multiple attributes also belongs to the classification space S_sv.

[0099] Sensitive attribute space att_set = {all sensitive attributes} = {att_1, att_2, att_3.....}. Each sensitive attribute will be quantitatively mapped to the corresponding indicator value val_ij corresponding to the sensitivity level. The relationship between them can be represented by a two-dimensional hierarchical matrix MAT, as shown below. Among them, val_ij satisfies monotonicity from left to right in rows.

[0100]

[0101] The grading index vector of the known attribute att_i is [val_i1, val_i2, val_i3, val_i4]. For a single attval_i, the calculated probability grading vector is pro_iset = [pro_i1, pro_i2, pro_i3, pro_i4].

[0102] The probability classification vector satisfies pro_i1+pro_i2+pro_i3+pro_i4=1.

[0103] For example, pro_1set=[0.2, 0.25, 0.45, 0.1] means that for attribute att_1, there is a 20% probability of being classified as insensitive, a 25% probability of being classified as moderately sensitive, a 45% probability of being classified as relatively sensitive, and a 10% probability of being classified as very sensitive.

[0104] For n attributes, according to the importance of the attribute (the contribution value to the privacy), it is assigned a weight w_set = {w_1, w_2.....w_n}, w_1+w_2+w_3+...+w_n=1.

[0105] Then, according to the weights and the classification vector of a single attribute, we perform joint classification of multiple sensitive attributes and obtain a probabilistic joint classification vector pro_sumset = [pro_sum1, pro_sum2, pro_sum3, pro_sum4] where

[0106] Finally, the SV_i corresponding to the largest pro_sumi is the sensitivity level obtained after the combination of multiple attributes. For example, if the final calculated pro_set = [0.2, 0.3, 0.4, 0.1], the classification result SV_i = SV_3 = {relatively sensitive}.

[0107] In a possible implementation, step S103 specifically includes:

[0108] Determining a sensitivity level of each attribute in each of the combined attributes;

[0109] Quantitatively mapping the sensitivity level of each attribute in each combined attribute to obtain an index value corresponding to each attribute in each combined attribute;

[0110] Using a single attribute grading algorithm, according to the index value corresponding to each attribute in each combination of attributes, a grading probability vector of each attribute in each combination of attributes is calculated;

[0111] According to the contribution value of each attribute in each combined attribute to the privacy amount, a weight value corresponding to each attribute in each combined attribute is set;

[0112] The classification probability vector of each attribute in each combination of attributes and the weight value corresponding to each attribute in each combination of attributes are jointly classified to obtain a joint classification vector of each combination of attributes.

[0113] In this embodiment, the sensitivity level of each attribute in the combined attribute is first determined, and then the sensitivity level of each attribute is quantitatively mapped to obtain an index value corresponding to each attribute.

[0114] Then, a single attribute grading algorithm is used to calculate the grading probability vector of each attribute in the combined attributes according to the index value corresponding to each attribute.

[0115] Then, according to the contribution value of each attribute in each combined attribute to the privacy amount, the weight value corresponding to each attribute in each combined attribute is set, and finally, the grading probability vector of each attribute in each combined attribute and the weight value corresponding to each attribute in each combined attribute are jointly graded to obtain the joint grading vector of each combined attribute.

[0116] It can be understood that compared with the qualitative analysis model, the embodiment of the present application performs grading calculations on combined attributes through a pre-built grading algorithm, which can quantitatively describe the grading results of multiple combined attributes, thereby improving the accuracy of data grading results.

[0117] In a possible implementation, step S202 may specifically include:

[0118] Determining an attribute value of each of the sample data attributes;

[0119] According to the distance between the attribute values ​​of each sample data attribute, each sample data attribute is graded to obtain the sensitivity level of each sample data attribute.

[0120] In this embodiment, the attribute value of each sample data attribute is first determined, and then each sample data attribute is graded according to the distance between the attribute values ​​of each sample data attribute to obtain the sensitivity level of each sample data attribute.

[0121] It can be understood that the embodiment of the present application grades each sample data attribute by the distance between the attribute values ​​of each sample data attribute to obtain the sensitivity level of each sample data attribute, which facilitates the subsequent quantitative mapping of the sensitivity level of each attribute, thereby obtaining the indicator value corresponding to each attribute in each combined attribute.

[0122] In a possible implementation, step S204 may specifically include:

[0123] Determining a grading index vector for each of the sample data attributes according to the index value corresponding to each of the sample data attributes;

[0124] According to the classification index vector of each sample data attribute, a classification probability vector of each sample data attribute is calculated.

[0125] Specifically, after obtaining the index value corresponding to each sample data attribute, a grading index vector of each sample data attribute is determined according to the index value corresponding to each sample data attribute.

[0126] Then, according to the classification index vector of each sample data attribute, the classification probability vector of each sample data attribute is calculated.

[0127] Next, the embodiment of the present application is described in detail using a specific scenario example.

[0128] As shown in Table 1, it is a credit card repayment data table provided in the embodiment of the present application. The credit card repayment data table shown in Table 1 is used as the analysis object to analyze the user's fund status. The specific analysis process is given below.

[0129]

[0130] Table 1

[0131] Each piece of data in the data set is graded according to the sensitivity of its sensitive attributes, and is divided into 4 sensitivity levels: {SV_1, SV_2, SV_3, SV_4} = {insensitive, moderately sensitive, relatively sensitive, very sensitive}. The sensitivity level refers to the privacy of sensitive information inferred by combining irrelevant attributes.

[0132] Perform a combined analysis on sensitive attributes {credit limit, bill amount, payment history}. Assume that the attacker's goal is to launch a marketing attack on the users in the dataset, such as selling loans to users who "need funds". The attacker can obtain private information by performing the following combined analysis on sensitive attributes:

[0133] 1) The larger the ratio of (bill amount / credit limit), the greater the user's capital demand. The larger the ratio of six consecutive months is, the greater the user's recent consumption demand is.

[0134] 2) Payment history reflects whether the user can repay on time. The payment history of 6 consecutive months is added up (or averaged). If the value is negative, it means that the user can basically repay in advance within a certain period of time and has good credit. If the value is positive and the larger it is, the poorer the user's repayment ability is.

[0135] 3) The ratio of "bill amount / credit limit" is used as sensitive attribute 1, and "the sum of payment history for 6 consecutive months" is used as sensitive attribute 2; the attacker uses the above two sensitive attributes to make inferences about the user to whom the attribute belongs. If the user's credit card "bill amount / credit limit" ratio is high and the payment history value is large within half a year, it means that the user borrows a lot but repays slowly, indicating that the user's recent financial situation is not good and is in a "fund-needing state"; thus, the sensitivity level of this data will be higher.

[0136] 4) According to the above analysis process, the parameter w_1=w_2=0.5 is selected, and the following grading standard matrix MAT is constructed according to the empirical threshold:

[0137]

[0138] Where att_1 represents the sum of the ratio of the bill amount / credit limit of the user to which the data belongs for 6 consecutive months; att_2 represents the sum of the payment history values ​​of the user to which the data belongs for 6 consecutive months.

[0139] Then, the combined attributes are graded and calculated by using a grading algorithm to obtain a joint grading vector of the combined attributes. Then, the sensitivity level to which the maximum value in the joint grading vector belongs is used as the sensitivity level corresponding to the combined attribute.

[0140] Through the above embodiment, each data in the repayment data set can be classified according to the sensitivity level {insensitive, generally sensitive, relatively sensitive, very sensitive} = {level 1, level 2, level 3, level 4} to obtain the corresponding classification value. Among them, users with a classification level of 4 are the objects that need to be focused on and protected in the privacy analysis and protection application scenario.

[0141] It is understandable that the classification results can reflect the sensitivity of the new information inferred after the association analysis between attributes, and can provide analytical guidance for the marketer's attack scenario. For example, in the marketer's attack process, the attacker will use the combination relationship between attributes to infer sensitive information, thereby leaking user privacy. For the results after the attribute association analysis, the more sensitive the information is, the more attention it needs to pay before publishing, the more generalized the processing is, and the higher the security threshold is set, such as a higher degree of anonymization, and sensitive attributes are processed as much as possible without affecting the availability of data; for less sensitive attributes, the threshold can be set lower, the degree of generalization is lower, and the availability of data is better preserved.

[0142] For the above-mentioned method embodiments, for the sake of simplicity, they are all described as a series of action combinations, but those skilled in the art should know that the present invention is not limited by the order of the actions described, because according to the present invention, some steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present invention.

[0143] The above are some specific implementations of the data classification method provided in the embodiment of the present application. Based on this, the present application also provides a corresponding device. The device provided in the embodiment of the present application will be introduced from the perspective of functional modularization.

[0144] See also Figure 3 , is a structural schematic diagram of a data grading device provided in an embodiment of the present application, the device includes an acquisition module 100, a combination module 200, a calculation module 300 and a grading module 400.

[0145] An acquisition module 100 is used to acquire a data set to be classified, wherein the data set includes a plurality of data and an attribute corresponding to each data;

[0146] A combination module 200 is used to combine attributes according to the degree of association between each attribute to obtain multiple combined attributes;

[0147] A calculation module 300 is used to perform classification calculations on the multiple combined attributes respectively using a classification algorithm to obtain a joint classification vector of each combined attribute in the multiple combined attributes;

[0148] The grading module 400 is used to determine the corresponding sensitivity level according to the joint grading vector of each combined attribute, and use the sensitivity level of each combined attribute as the grading result of the data set to be graded.

[0149] Optionally, the grading module is specifically used to:

[0150] Determining the maximum value in the joint ranking vector of each combined attribute respectively;

[0151] The sensitivity level to which the maximum value in the joint classification vector of each combined attribute belongs is used as the sensitivity level corresponding to each combined attribute.

[0152] Optionally, the sensitivity levels include: an insensitive level, a general sensitive level, a relatively sensitive level and a very sensitive level.

[0153] It can be seen from the above technical scheme that the embodiment of the present application first obtains a data set to be classified, wherein the data set includes multiple data and attributes corresponding to each data; then, the attributes are combined according to the degree of association between each attribute to obtain multiple combined attributes; then, a pre-constructed grading algorithm is used to perform grading calculations on each of the multiple combined attributes to obtain a joint grading vector of each combined attribute; finally, according to the joint grading vector of each combined attribute, the sensitivity level corresponding to each combined attribute is determined, and the sensitivity level of each combined attribute is used as the grading result of the data set to be graded. In the above manner, the present application not only considers the sensitivity of a single attribute, but also considers the privacy of the information inferred after multiple attributes are associated, so as to perform a more comprehensive grading. Moreover, compared with the qualitative analysis model, the present application can quantitatively describe the grading results of multiple combined attributes by performing grading calculations on the combined attributes through a pre-constructed grading algorithm, thereby improving the accuracy of the data grading results.

[0154] An embodiment of the present application further provides an electronic device, comprising: a memory for storing one or more programs;

[0155] Processor; when the one or more programs are executed by the processor, the data classification method described in the above embodiment is implemented.

[0156] An embodiment of the present application further provides a computer storage medium, in which a program is stored. When the program is executed by a processor, the data classification method described in the above embodiment is implemented.

[0157] The "first" and "second" in the names such as "first" and "second" (if any) mentioned in the embodiments of the present application are only used as name identifiers and do not represent the first or second in order.

[0158] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0159] Those skilled in the art can understand that the flowchart shown in the figure is only an example in which the embodiments of the present application can be implemented, and the scope of application of the embodiments of the present application is not limited by any aspect of the flowchart.

[0160] In the several embodiments provided in the present application, it should be understood that the disclosed methods, devices and equipment can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interfaces, and the indirect coupling or communication connection of the device or unit can be electrical, mechanical or other forms.

[0161] The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, the functional units in the various embodiments of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0162] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, and other media that can store program codes.

[0163] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A data classification method, characterized in that: The method comprises: Acquire a data set to be classified, the data set comprising a plurality of data and an attribute corresponding to each data; Combine according to the correlation degree between each attribute to obtain multiple combined attributes; Using a grading algorithm to perform grading calculation on each of the multiple combined attributes to obtain a joint grading vector of each combined attribute; According to the joint classification vector of each combined attribute, the sensitivity level corresponding to each combined attribute is determined, and the sensitivity level of each combined attribute is used as the classification result of the data set to be classified.

2. The method according to claim 1, characterized in that The construction process of the grading algorithm includes: Acquire a sample data set, where the sample data set includes a plurality of sample data and an attribute corresponding to each sample data; Determine the sensitivity level of each sample data attribute; Quantitatively mapping the sensitivity level of each sample data attribute to obtain an index value corresponding to each sample data attribute; Using a single attribute classification algorithm, according to the index value corresponding to each sample data attribute, a classification probability vector of each sample data attribute is calculated; According to the contribution value of each sample data attribute to the privacy amount, set the weight value corresponding to each sample data attribute; The classification probability vector of each sample data attribute and the weight value corresponding to each sample data attribute are jointly classified to obtain a joint classification vector of each sample data combination attribute.

3. The method according to claim 2, characterized in that Using a grading algorithm to perform grading calculation on each of the multiple combined attributes to obtain a joint grading vector of each combined attribute, including: Determining a sensitivity level of each attribute in each of the combined attributes; Quantitatively mapping the sensitivity level of each attribute in each combined attribute to obtain an index value corresponding to each attribute in each combined attribute; Using a single attribute grading algorithm, according to the index value corresponding to each attribute in each combination of attributes, a grading probability vector of each attribute in each combination of attributes is calculated; According to the contribution value of each attribute in each combined attribute to the privacy amount, a weight value corresponding to each attribute in each combined attribute is set; The classification probability vector of each attribute in each combination of attributes and the weight value corresponding to each attribute in each combination of attributes are jointly classified to obtain a joint classification vector of each combination of attributes.

4. The method according to claim 2, characterized in that: Determining the sensitivity level of each sample data attribute includes: Determining an attribute value of each of the sample data attributes; According to the distance between the attribute values ​​of each sample data attribute, each sample data attribute is graded to obtain the sensitivity level of each sample data attribute.

5. The method according to claim 2, characterized in that: The step of calculating the classification probability vector of each sample data attribute according to the index value corresponding to each sample data attribute comprises: Determining a grading index vector for each of the sample data attributes according to the index value corresponding to each of the sample data attributes; According to the classification index vector of each sample data attribute, a classification probability vector of each sample data attribute is calculated.

6. The method according to claim 1, characterized in that The determining, according to the joint classification vector of each combined attribute, a corresponding sensitivity level comprises: Determining the maximum value in the joint ranking vector of each combined attribute respectively; The sensitivity level to which the maximum value in the joint classification vector of each combined attribute belongs is used as the sensitivity level corresponding to each combined attribute.

7. The method according to claim 1, characterized in that The sensitivity levels include: insensitive level, general sensitivity level, relatively sensitive level and very sensitive level.

8. A data classification device, characterized in that: The device comprises: An acquisition module, used to acquire a data set to be classified, wherein the data set includes a plurality of data and an attribute corresponding to each data; A combination module is used to combine attributes according to the degree of association between each attribute to obtain multiple combined attributes; A calculation module, configured to perform hierarchical calculations on the plurality of combined attributes respectively by using a hierarchical algorithm to obtain a joint hierarchical vector of each of the plurality of combined attributes; The grading module is used to determine the corresponding sensitivity level according to the joint grading vector of each combined attribute, and use the sensitivity level of each combined attribute as the grading result of the data set to be graded.

9. The device according to claim 8, characterized in that The grading module is specifically used for: Determining the maximum value in the joint ranking vector of each combined attribute respectively; The sensitivity level to which the maximum value in the joint classification vector of each combined attribute belongs is used as the sensitivity level corresponding to each combined attribute.

10. The device according to claim 8, characterized in that The sensitivity levels include: insensitive level, general sensitivity level, relatively sensitive level and very sensitive level.

Images