A data analysis method, a noise construction method, an apparatus, and a storage medium.

Abstract

This application provides a data analysis method, a noise construction method, an apparatus, and a storage medium. In this application embodiment, noise parameters are adaptively selected based on the original analysis values ​​under a target analysis label. Then, the noise to be added to the original analysis values ​​under the target analysis label is determined based on the selected noise parameters, thereby obtaining the analysis result under the target analysis label. Accordingly, in the differential privacy protection process, the original analysis values ​​of different analysis labels can be used as a basis to adaptively determine the noise parameters used for different analysis labels, thereby reasonably controlling the range of noise added under different analysis labels, making the range of added noise compatible with the original analysis values. This not only ensures data security but also guarantees the usability of the analysis results.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, and in particular to a data analysis method, a noise construction method, a device and a storage medium. BACKGROUND

[0002] A Laplace mechanism with differential privacy protection can provide privacy protection for a data analysis process. Under the Laplace mechanism, Laplace noise can be added to an attribute analysis value to achieve differential privacy protection.

[0003] Currently, the attribute analysis value is usually added with noise within a unified Laplace noise range, which leads to the following problems: when the attribute analysis value is relatively small, the added noise can be too large, and the analysis result can be unavailable; or when the attribute analysis value is relatively large, the added noise can be too small, and the analysis result can be infinitely close to the attribute analysis value, causing privacy leakage. SUMMARY

[0004] Aspects of the present application provide a data analysis method, a noise construction method, a device and a storage medium to improve the security and usability of data.

[0005] The present application provides a data analysis method, comprising:

[0006] obtaining an original analysis value under a target analysis label;

[0007] selecting a target noise parameter matching the original analysis value from at least one candidate noise parameter corresponding to the target analysis label;

[0008] determining noise under the target analysis label based on the target noise parameter;

[0009] adding the noise to the original analysis value under the target analysis label to obtain an analysis result under the target analysis label.

[0010] The present application also provides a noise construction method, comprising:

[0011] in response to a noise construction instruction, obtaining a number of objects contained in a target group;

[0012] determining a target noise parameter adapted to the number of objects contained in the target group based on a mapping relationship between the number of objects and noise parameters;

[0013] constructing noise for the target group according to the target noise parameter;

[0014] wherein the noise parameter affects the range of the noise constructed for the target group.

[0015] The embodiment of the present application further provides a computing device comprising a memory and a processor;

[0016] The memory is used for storing one or more computer instructions;

[0017] The processor is coupled with the memory and is used for executing the one or more computer instructions, so as to:

[0018] Obtaining an original analysis value under a target analysis label;

[0019] From at least one candidate noise parameter corresponding to the target analysis label, a target noise parameter matching the original analysis value is selected;

[0020] Based on the target noise parameter, noise under the target analysis label is determined;

[0021] The noise is added to the original analysis value under the target analysis label, so as to obtain an analysis result under the target analysis label.

[0022] The embodiment of the present application further provides a computing device comprising a memory and a processor;

[0023] The memory is used for storing one or more computer instructions;

[0024] The processor is coupled with the memory and is used for executing the one or more computer instructions, so as to:

[0025] In response to a noise construction instruction, the number of objects contained in a target group is obtained;

[0026] Based on a mapping relationship between the number of objects and noise parameters, a target noise parameter adapted to the number of objects contained in the target group is determined;

[0027] According to the target noise parameter, noise is constructed for the target group;

[0028] The noise parameter affects the range of the noise constructed for the target group.

[0029] The embodiment of the present application further provides a computer readable storage medium storing computer instructions, when the computer instructions are executed by one or more processors, the one or more processors are caused to execute the data analysis method.

[0030] In the embodiments of the present application, the noise parameter can be adaptively selected according to the original analysis value under the target analysis label, and then the noise to be added to the original analysis value under the target analysis label is determined according to the selected noise parameter, so as to obtain the analysis result under the target analysis label. Accordingly, in the differential privacy protection process, the original analysis values of different analysis labels can be used as the basis to adaptively determine the noise parameters used by different analysis labels, so as to reasonably control the range of the added noise under different analysis labels, so that the range of the added noise is adapted to the original analysis value. This not only ensures the security of the data, but also ensures the usability of the analysis result. BRIEF DESCRIPTION OF DRAWINGS

[0031] The accompanying drawings, which are included to provide a further understanding of the present application, constitute a part of the present application and illustrate embodiments of the present application and the description thereof, which serve to explain the present application and do not constitute improper limitations on the present application. In the drawings:

[0032] Figure 1 A flowchart of a data analysis method provided for an exemplary embodiment of the present application is shown in FIG. 2;

[0033] Figure 2 A logic diagram of a data analysis scheme provided for an exemplary embodiment of the present application is shown in FIG. 3;

[0034] Figure 3 An example diagram of an analysis label and an analysis attribute provided for an exemplary embodiment of the present application is shown in FIG. 4;

[0035] Figure 4 A structural diagram of a computing device provided for an exemplary embodiment of the present application is shown in FIG. 5;

[0036] Figure 5 A flowchart of a noise construction method provided for an exemplary embodiment of the present application is shown in FIG. 6;

[0037] Figure 6 A structural diagram of another computing device provided for an exemplary embodiment of the present application is shown in FIG. 7. DETAILED DESCRIPTION

[0038] In order to make the purposes, technical solutions and advantages of the present application clearer, the technical solutions of the present application will be described clearly and completely below in conjunction with the embodiments of the present application and the corresponding drawings. Obviously, the described embodiments are only some of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0039] In existing differential privacy protection processes, unreasonable addition of noise often leads to unusable analysis results or data leakage. To improve these technical problems, in some embodiments of the present application, noise parameters can be adaptively selected according to original analysis values under a target analysis label, and then noise to be added to the original analysis values under the target analysis label is determined according to the selected noise parameters, so as to obtain analysis results under the target analysis label. Accordingly, in the differential privacy protection process in the embodiments of the present application, noise parameters used for different analysis labels can be adaptively determined based on original analysis values of the different analysis labels, so as to reasonably control the range of noise added under different analysis labels, so that the range of the added noise is adapted to the original analysis values. This not only ensures the security of data, but also ensures the usability of analysis results.

[0040] The technical solutions provided by the embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0041] Figure 1 A flowchart of a data analysis method provided by an exemplary embodiment of the present application. Figure 2 A logic diagram of a data analysis scheme provided by an exemplary embodiment of the present application. The data analysis method provided by the embodiment can be executed by a data analysis device, which can be implemented as software or as a combination of software and hardware, and can be integrated into a computing device. As shown in the figure, the method comprises: Figure 1

[0042] Step 100, obtaining original analysis values under a target analysis label;

[0043] Step 101, selecting a target noise parameter matching the original analysis values from at least one candidate noise parameter corresponding to the target analysis label;

[0044] Step 102, determining noise under the target analysis label based on the target noise parameter;

[0045] Step 103, adding the noise to the original analysis values under the target analysis label to obtain analysis results under the target analysis label.

[0046] The data analysis method provided by the embodiment can be applied to various scenarios requiring differential privacy protection, such as e-commerce scenarios, financial scenarios, government scenarios, etc. In different application scenarios, the objects of differential privacy protection can not be completely the same. From the perspective of data, the data types of differential privacy protection in the embodiment include but are not limited to attribute information of individuals or enterprises, transaction information, etc. From the perspective of objects, the objects of differential privacy protection in the embodiment include but are not limited to individuals, enterprises, orders or transactions, etc. The application scenarios, protected objects, data types, etc. are not limited in the embodiment.​

[0047] In this embodiment, data resources of different users can be managed respectively, wherein the data resources can include one or more initial group ranges. For example, if the object of differential privacy protection is an individual, the data resources of a certain user can include multiple initial group ranges such as a marketing activity deposited group, a custom group, an uploaded group, a full-link group, a link flow group, and the like. The generation manners of different initial group ranges can not be completely the same. For example, the marketing activity deposited group can be a group composed of participants collected in a marketing activity, the custom group can be a group filtered from a data warehouse by a user through a custom condition, the uploaded group can be a group collected by the user offline, and the like. In this way, the group features under different initial group ranges can not be completely the same.

[0048] In this embodiment, at least one noise parameter can be configured for a single initial group range, and different noise parameters can correspond to different group quantities. The noise parameter can be a factor affecting the noise boundary corresponding to the initial group range in the noise calculation process. The at least one noise parameter corresponding to different initial group ranges can not be completely the same, and the specific configuration manner will be described below.

[0049] For example, under the Laplace mechanism, an inverse cumulative function is usually used to represent noise, and the corresponding formula is as follows: The standard deviation b affects the noise boundary, that is, the range of noise added under the initial group range. Therefore, the standard deviation b can be taken as the noise parameter in this embodiment. In this embodiment, at least one standard deviation b can be configured for a single initial group range. In addition, in the above inverse cumulative function, other parameters are also included, such as expectation μ and noise factor The expectation μ can usually be set to 0, and each analysis attribute corresponds to a noise factor Each noise factor Usually, it is a uniformly distributed random value.

[0050] Based on this, a user can initiate a data analysis request and specify an initial group range as the analysis basis. For example, the user can initiate a data analysis request for the uploaded group described above. In this embodiment, the data analysis request can include at least one analysis tag. Since the processing manners for different analysis tags are similar, for the convenience of description, the data analysis scheme in this embodiment will be described below by taking a target analysis tag as an example.

[0051] In step 100, an original analysis value under the target analysis tag can be obtained. The original analysis value refers to an actual analysis value generated by data analysis under the target analysis tag, that is, a true value before noise is added.

[0052] In step 101, a target noise parameter matching the original analysis value can be selected from at least one noise parameter configured for the initial population range.

[0053] In this embodiment, the population quantity conforming to the target analysis label can generally be used as the original analysis value under the target analysis label. As mentioned above, at least one noise parameter can be configured for the initial population range in this embodiment, and different noise parameters can correspond to different population quantities. In this way, in step 101, a noise parameter matching the population quantity under the target analysis label can be found for the target analysis label as the target noise parameter.

[0054] For other analysis labels in the data analysis request, a target noise parameter matching the corresponding original analysis value can be selected for the other analysis labels in a polling manner. Thus, in this embodiment, different noise parameters matching the original analysis values can be selected for different analysis labels in the data analysis request. The noise parameters selected under different analysis labels can not be completely the same on the basis of the original analysis values. That is, in this embodiment, noise parameters can be adaptively selected for different analysis labels on the basis of the original analysis values.

[0055] As mentioned above, a noise parameter affects the noise boundary, that is, the upper and lower bounds of the noise range. Therefore, adaptively selecting a noise parameter on the basis of the original analysis value can make the noise parameter used under the target analysis label more reasonable, and thus can ensure that the generated noise is more reasonable, and avoid problems such as that the added noise is too large to cause the analysis result to be unusable or too small to cause the original analysis value to be leaked.

[0056] Reference Figure 1 and Figure 2 After the target noise parameter is determined, in step 102, noise under the target analysis label can be determined on the basis of the target noise parameter. For example, in the Laplace mechanism, the inverse cumulative function mentioned above can be used to calculate the noise. For other analysis labels in the data analysis request, noise can also be calculated on the basis of the corresponding target noise parameters.

[0057] On this basis, in step 103, the noise can be added to the original analysis value under the target analysis label to obtain an analysis result under the target analysis label. Similarly, the analysis results under other analysis labels in the data analysis request can also be determined in the same manner.

[0058] According to the embodiment, the noise parameter can be adaptively selected according to the original analysis value under the target analysis label, and then the noise to be added to the original analysis value under the target analysis label is determined according to the selected noise parameter, so as to obtain the analysis result under the target analysis label. According to the embodiment, in the differential privacy protection process, the noise parameters used by different analysis labels can be adaptively determined according to the original analysis values of different analysis labels, so as to reasonably control the range of the added noise under different analysis labels, so that the range of the added noise is adapted to the original analysis value, thereby effectively protecting the original analysis value. This not only ensures the security of the data, but also ensures the usability of the analysis result.

[0059] In the above or the following embodiments, at least one noise parameter can be configured for the initial population range in advance. Since the processing logic of configuring noise parameters for different initial population ranges is similar, in order to facilitate description, the configuration process is described by taking the initial population range specified by the data analysis request in the above embodiment as an example, but it should be understood that other initial population ranges can also be configured with at least one noise parameter in the same way.

[0060] In the embodiment, the order of magnitude of the initial population range can be obtained, and the order of magnitude of the initial population range can be progressively reduced to determine a plurality of segmentation values.

[0061] In a possible design, the order of magnitude of the initial population range can be progressively reduced according to a plurality of preset reduction conditions. The reduction conditions can be diversified and can be flexibly set according to actual needs. For example, in an e-commerce scenario, a plurality of reduction conditions such as advertisement delivery, coupon issuance, shopping behavior, shopping amount reaching one thousand yuan, etc. can be used. If the order of magnitude of the initial population range is 1000W, after the advertisement delivery level reduction, a segmentation value of 100W can be generated, after the coupon issuance level reduction, a segmentation value of 10W can be generated, after the shopping behavior level reduction, a segmentation value of 1W can be generated, and after the shopping amount reaching one thousand yuan level reduction, a segmentation value of 1000 can be generated.

[0062] On this basis, in the embodiment, a plurality of noise parameters can be configured for the plurality of segmentation values corresponding to the initial population range as at least one noise parameter corresponding to the initial population range. Here, the noise parameters can be configured for the plurality of segmentation values according to actual needs. For example, the corresponding noise parameter can be determined by Z(segmentation value) * noise ratio according to the required noise ratio of the user, wherein the noise ratios corresponding to different segmentation values can not be completely the same. Of course, other implementation manners for configuring noise parameters for the plurality of segmentation values corresponding to the initial population range can also be used, and the embodiment is not limited thereto.

[0063] On this basis, in the embodiment, a budget can also be configured for at least one noise parameter in the initial population range respectively. The budgets corresponding to different noise parameters can not be completely the same.

[0064] In the differential privacy protection process, each data analysis on a data set (such as the initial population range or the target group) consumes a budget, and after the budget is consumed, subsequent data analysis will not be allowed, thereby avoiding attackers from inferring the original analysis value through a large number of analysis times.

[0065] In this way, in the embodiment, after the target noise parameter corresponding to the target analysis tag is determined, the budget corresponding to the target noise parameter can be used as the budget under the target analysis tag. As mentioned above, in the data analysis process, the target noise parameters corresponding to different analysis tags can not be completely the same, and accordingly, the budgets corresponding to different analysis tags can also not be completely the same. Therefore, in the embodiment, different analysis tags no longer use the same budget, but use a budget that is adapted to the original analysis value of the analysis tag, which can reduce the budget corresponding to each analysis tag as much as possible while ensuring data security.

[0066] In summary, in the embodiment, a set of noise parameters can be configured for different initial population ranges as candidate noise parameters in data analysis. Different noise parameters correspond to different orders of magnitude, so in the data analysis process, the target noise parameter corresponding to the order of magnitude of the original analysis value under the target analysis tag can be selected, so that the noise added under the target analysis tag is more reasonable, and the problem of overlarge or oversmall noise can be avoided.

[0067] In the above or the following embodiments, if the target analysis tag includes multiple analysis attributes, the attribute analysis values corresponding to the multiple analysis attributes respectively can be obtained; and the original analysis value under the target analysis tag can be determined according to the attribute analysis values corresponding to the multiple analysis attributes respectively.

[0068] Figure 3 An example diagram of analysis tags and analysis attributes provided by an example embodiment of the present application is shown. Figure 3 In the example, the analysis tags include “predicted occupation” and the like, and each analysis tag includes one or more analysis attributes, for example, the analysis attributes included in “predicted occupation” include “company employee”, “student”, “faculty”, “civil servant” and the like.

[0069] It should be noted that, Figure 3 The analysis tags and analysis attributes in the example are only exemplary, and in the embodiment, the analysis tags and / or analysis attributes required by a user can be flexibly configured in a data analysis request according to actual needs, and the specific types of the analysis tags and the analysis attributes are not limited in the embodiment.

[0070] In this embodiment, the user can be provided with optional analysis tags and optional analysis attributes under the analysis tags. The user can select one or more analysis tags according to the analysis requirement, in which case the analysis attributes under the analysis tags can be selected by default, so that the selected analysis tags and analysis attributes are configured in the data analysis request. Of course, the user can also select one or more analysis attributes under the relevant analysis tags according to the analysis requirement, so that the relevant analysis tags and selected analysis attributes can be configured in the data analysis request. In this way, at least one analysis tag and the analysis attributes under the at least one analysis tag can be included in the data analysis request in this embodiment.

[0071] Based on this, in this embodiment, data analysis can be performed according to at least one analysis tag and the analysis attributes under the at least one analysis tag included in the data analysis request to obtain attribute analysis values under each analysis attribute. In actual application, the user can also configure circle selection condition information in the data analysis request to filter out a target group meeting the circle selection condition information from the initial population range, and perform data analysis on the target group according to at least one analysis tag and the analysis attributes under the at least one analysis tag included in the data analysis request to obtain attribute analysis values under each analysis attribute. The circle selection condition information includes but is not limited to time information, distance information, gender information or other object description information. Based on the circle selection condition information, object circle selection can be performed in the initial population range to determine the target group meeting the circle selection condition information. For example, in the e-commerce scenario, the circle selection condition information can be a male who appeared within a 5-kilometer range of store A on August 3.

[0072] Accordingly, attribute analysis values corresponding to each analysis attribute under the target analysis tag can be obtained.

[0073] In an optional implementation, the median of the attribute analysis values corresponding to each of the plurality of analysis attributes can be taken as the original analysis value under the target analysis tag. For example, Figure 3 The attribute analysis values corresponding to each of the analysis attributes "company employee", "student", "faculty", and "civil servant" under the analysis tag "predicted occupation" are 50, 100, 30, and 70 respectively, and the attribute analysis values corresponding to other analysis attributes are not exemplified. In this case, the median, i.e., the middle value, can be determined from the attribute analysis values corresponding to each of the plurality of analysis attributes under the analysis tag "predicted occupation" as the basis for selecting the target noise parameter under the analysis tag "predicted occupation".

[0074] In another alternative implementation, the mean of the attribute analysis values corresponding to the plurality of analysis attributes can be calculated as the original analysis value under the target analysis label. In the above example, the mean of the attribute analysis values corresponding to the plurality of analysis attributes under the analysis label "predicted occupation" can be calculated as the basis for selecting the target noise parameter under the analysis label "predicted occupation".

[0075] Of course, in the present embodiment, other implementations can also be used to determine the original analysis value under the target analysis label according to the attribute analysis values corresponding to the plurality of analysis attributes, and the present embodiment is not limited thereto.

[0076] Accordingly, in the case where the target analysis label includes a plurality of analysis attributes, the attribute analysis values of the plurality of analysis attributes can be comprehensively considered to determine the original analysis value corresponding to the target analysis label, and the original analysis value thus determined can be used as the basis for adaptively selecting the target noise parameter.

[0077] On this basis, in the present embodiment, in the case where the target analysis label includes a plurality of analysis attributes, the noise under the plurality of analysis attributes can be calculated according to the selected target noise parameter. The plurality of analysis attributes under the target analysis label can use the same target noise parameter for noise calculation. For example, in the Laplace mechanism, the noise under the plurality of analysis attributes can be calculated based on the aforementioned inverse cumulative function, and in the same analysis label, the standard deviation b used in the inverse cumulative function is constant, i.e., the plurality of analysis attributes under the same analysis label can use the same standard deviation b for inverse cumulative function calculation.

[0078] It is because that in the present embodiment, the attribute analysis values corresponding to the plurality of analysis attributes under the target analysis label are comprehensively considered when determining the original analysis value corresponding to the target analysis label, which can effectively improve the adaptation degree between the selected target noise parameter and the attribute analysis values corresponding to the plurality of analysis attributes under the target analysis label, so as to ensure that the range of the noise added under the target analysis label conforms to the overall trend of the attribute analysis values corresponding to the plurality of analysis attributes, and thus the problem of excessive or insufficient noise added under some analysis attributes can be effectively avoided.

[0079] In the above or the following embodiments, if the target analysis label includes a plurality of analysis attributes, the noise factors corresponding to the plurality of analysis attributes are obtained; and the noise corresponding to the plurality of analysis attributes is respectively calculated based on the target noise parameter and the noise factors corresponding to the plurality of analysis attributes.

[0080] For example, in the Laplace mechanism, the noise for the plurality of analysis attributes can be calculated according to the inverse cumulative function formula mentioned above.

[0081] In the embodiment, the noise factor corresponding to each analysis attribute can be obtained in various manners. In an optional manner, the noise seed corresponding to the circle selection condition information associated with the target analysis label is obtained, and the noise corresponding to each analysis attribute is calculated based on the noise seed.

[0082] In the implementation, the circle selection condition information can be included in the data analysis request, and the noise seed can be configured in advance for different circle selection condition information. Based on the noise seed, the noise factor corresponding to each analysis attribute in the data analysis request can be derived. For example, the random algorithm can be used to generate the noise factor corresponding to each analysis attribute under each analysis label in the data analysis request based on the noise seed. The noise factor generated in this way is uniformly distributed and meets the characteristics of random distribution.

[0083] In the implementation, the noise factor does not need to be configured for each analysis attribute in advance, but is generated in real time. In this way, only the noise seed needs to be stored for the circle selection condition information, and the noise factor does not need to be stored for each analysis attribute, which can save a large amount of storage resources and avoid performance bottlenecks caused by high concurrency access to the noise factor.

[0084] Of course, in the embodiment, other manners of calculating the noise corresponding to each analysis attribute can also be used, for example, the noise factor is stored for each analysis attribute in the conventional manner, and the embodiment is not limited in this regard.

[0085] On this basis, in the embodiment, the operation of adding noise can be performed under each analysis attribute, that is, the noise corresponding to each analysis attribute is added to the attribute analysis value corresponding to the corresponding analysis attribute. The attribute analysis value can be the parameter of the previous stage of the final analysis result. For example, Figure 3 The analysis attribute "single ratio" in the table can use the number of single people in the target group as the attribute analysis value, and does not use the actual single ratio as the attribute analysis value. This can further improve the data security.

[0086] In the embodiment, different noise seeds can be configured for different circle selection condition information, and when the data analysis request is received, the corresponding noise seed can be obtained, and the noise corresponding to each analysis attribute in the data analysis request can be derived based on the noise seed. Accordingly, in the differential privacy protection process in the embodiment, only one unit of storage space is needed for each circle selection condition information to store the corresponding noise seed, and the noise meeting the requirements can be generated, without the need to provide one unit of storage space for each analysis attribute as in the conventional manner. Thus, the storage pressure of differential privacy protection can be greatly reduced.

[0087] In the above or below embodiments, a maximum noise value corresponding to the original analysis value can be calculated according to a preset noise upper limit; and a target noise parameter meeting a preset requirement in distance from the maximum noise value can be selected from at least one candidate noise parameter corresponding to the target analysis label. The noise upper limit can be understood as a maximum noise proportion that a user can tolerate. The user can set the same or not exactly the same noise upper limit for different analysis labels, which is not limited in this embodiment.

[0088] In this embodiment, various implementation manners can be used to calculate the maximum noise value corresponding to the original analysis value of the target analysis label.

[0089] In one implementation manner, the product of the original analysis value and the noise upper limit can be taken as the maximum noise value. For example, the original analysis value is 654321, and the noise upper limit is 1‰, and then the maximum noise value is about 654.

[0090] In another implementation manner, a shift number N corresponding to the noise upper limit can be determined, and the decimal point of the original analysis value can be right-shifted by N bits to obtain the maximum noise value.

[0091] In actual application, the user can directly use the shift number to represent the noise upper limit without directly setting the specific noise upper limit when setting the noise upper limit. For example, the user can set N corresponding to the noise upper limit as 4. Of course, the user can also set the specific noise upper limit, for example, 1‰ in the previous example. In this case, the fraction corresponding to the noise upper limit can be reduced, and the reduced simplified fraction can be converted into a decimal number. In the process of converting into the decimal number, the shift number of the decimal point of the numerator is taken as the shift number corresponding to the noise upper limit. For example, the noise upper limit is 1‰, and the determined shift number is 3 bits after the above process, and the noise upper limit is 10‰, and the determined shift number is 2.

[0092] In this implementation manner, the shift operation can be performed on the original analysis value based on the determined shift number to obtain the maximum noise value. For example, the original analysis value is 654321, and if the determined shift number N is 3, then the maximum noise value is about 654.

[0093] It is worth noting that in this implementation manner, the molecule after the reduction of the noise upper limit can not be 1, for example, the noise upper limit is 5‰, and the noise value obtained by only performing the shift operation on the original analysis value can have a deviation from the actual maximum noise value. However, from the perspective of actual application, these deviations can usually be ignored.

[0094] In addition, in this embodiment, other implementation manners can also be used to determine the maximum noise value corresponding to the original analysis value of the target analysis label, which is not limited in this embodiment.

[0095] Based on the maximum noise value, a target noise parameter that meets a preset requirement in distance from the maximum noise value can be selected from the at least one candidate noise parameter corresponding to the target analysis label.

[0096] In one possible solution, the preset requirement can be configured as: the noise parameter closest to the maximum noise value and smaller than the maximum noise value. For example, the noise parameters corresponding to the initial population range are: 50, 100, 500, 1000, and 1W. If the maximum noise parameter corresponding to the original analysis value of the target analysis label is 654, based on the preset requirement here, the target noise parameter can be determined as 500.

[0097] Of course, in this embodiment, the preset requirement is not limited to this, and the preset requirement can also be configured as any one of the noise parameters closest to the maximum noise value, the noise parameters with a distance from the maximum noise value smaller than a preset threshold, and the like.

[0098] Accordingly, in this embodiment, the maximum noise value can be calculated for the original analysis value of the target analysis label according to the preset noise upper bound, and based on the maximum noise value, a target noise parameter matching the original analysis value of the target analysis label can be selected from the at least one candidate noise parameter corresponding to the target analysis label. This can ensure that the noise range determined by the selected target noise parameter meets the preset noise upper bound, so that the noise range added to the target analysis label is more reasonable, thereby ensuring data security and usability of the analysis result.

[0099] It should be noted that the execution subject of each step of the method provided in the above embodiments can be the same device, or the method can also be executed by different devices as the execution subject. For example, the execution subject of steps 101 to 103 can be device A; for another example, the execution subject of steps 101 and 102 can be device A, and the execution subject of step 103 can be device B; and the like.

[0100] In addition, in some of the processes described in the above embodiments and the accompanying drawings, a plurality of operations appearing in a certain order are included, but it should be clear that these operations can be executed in the order appearing in this document or in parallel, and the serial numbers of the operations such as 101, 102, etc. are only used to distinguish different operations, and the serial numbers themselves do not represent any execution order. In addition, these processes can include more or fewer operations, and the operations can be executed in sequence or in parallel. It should be noted that the "first", "second", etc. described herein are used to distinguish different messages, devices, modules, etc., and do not represent the order of precedence. In addition, "first" and "second" are different types.

[0101] Figure 4A structural diagram of a computing device is provided for another exemplary embodiment of the present application. As shown in Figure 4 The computing device includes a memory 40 and a processor 41.

[0102] The memory 40 is configured to store computer programs and can be configured to store other various data to support operations on the computing platform. Examples of the data include instructions of any application program or method for operating on the computing platform, contact data, phonebook data, messages, pictures, videos, etc.

[0103] The memory can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0104] The processor 41 is coupled to the memory 40 and is configured to execute the computer programs in the memory 40 for:

[0105] obtaining an original analysis value under a target analysis tag;

[0106] selecting a target noise parameter matching the original analysis value from at least one candidate noise parameter corresponding to the target analysis tag;

[0107] determining noise under the target analysis tag based on the target noise parameter;

[0108] adding the noise to the original analysis value under the target analysis tag to obtain an analysis result under the target analysis tag.

[0109] In an optional embodiment, under a Laplace mechanism, the noise parameter is a standard deviation.

[0110] In an optional embodiment, when the processor 41 obtains the original analysis value under the target analysis tag, the processor 41 is configured to:

[0111] if the target analysis tag includes multiple analysis attributes, obtaining attribute analysis values respectively corresponding to the multiple analysis attributes;

[0112] determining the original analysis value under the target analysis tag according to the attribute analysis values respectively corresponding to the multiple analysis attributes.

[0113] In an optional embodiment, when the processor 41 determines the original analysis value under the target analysis tag according to the attribute analysis values respectively corresponding to the multiple analysis attributes, the processor 41 is configured to:

[0114] The median of the attribute analysis values corresponding to the multiple analysis attributes is taken as the original analysis value under the target analysis label.

[0115] In an optional embodiment, the processor 41 is configured to:

[0116] receive a data analysis request, the data analysis request comprising the circle selection condition information, the target analysis label, and the analysis attributes under the target analysis label;

[0117] determine the target group meeting the circle selection condition information from the initial population range;

[0118] perform data analysis on the target group according to the target analysis label and the analysis attributes under the target analysis label to obtain the attribute analysis values corresponding to the multiple analysis attributes.

[0119] In an optional embodiment, the processor 41 is further configured to:

[0120] determine the multiple segment values based on the order of magnitude of the initial population range;

[0121] configure noise parameters for the multiple segment values to obtain at least one noise parameter corresponding to the initial population range.

[0122] In an optional embodiment, the processor 41 is configured to:

[0123] if the target analysis label comprises multiple analysis attributes, obtain noise factors corresponding to the multiple analysis attributes;

[0124] calculate the noise corresponding to the multiple analysis attributes based on the target noise parameter and the noise factors corresponding to the multiple analysis attributes.

[0125] In an optional embodiment, the processor 41 is configured to:

[0126] obtain a noise seed corresponding to the circle selection condition information associated with the target analysis label;

[0127] generate noise factors for at least one analysis attribute based on the noise seed.

[0128] In an optional embodiment, the processor 41 is configured to:

[0129] add the noise corresponding to the multiple analysis attributes to the attribute analysis values corresponding to the respective analysis attributes.

[0130] In an optional embodiment, when the processor 41 selects the target noise parameter that matches the original analysis value from the at least one candidate noise parameter corresponding to the target analysis label, the processor 41 is configured to:

[0131] calculate a maximum noise value corresponding to the original analysis value according to a preset noise upper limit;

[0132] select the target noise parameter that meets a preset requirement in distance from the maximum noise value from the at least one candidate noise parameter corresponding to the target analysis label.

[0133] In an optional embodiment, the preset requirement includes:

[0134] the target noise parameter closest to the maximum noise value and smaller than the maximum noise value.

[0135] In an optional embodiment, when the processor 41 calculates the maximum noise value corresponding to the original analysis value according to the preset noise upper limit, the processor 41 is configured to:

[0136] determine a shift number N corresponding to the noise upper limit, N being a positive integer;

[0137] shift the decimal point of the original analysis value to the left by N bits to obtain the maximum noise value.

[0138] In an optional embodiment, the processor 41 is further configured to:

[0139] configure a budget for each of the at least one noise parameter.

[0140] It is worth noting that the technical details in the above embodiments of the computing device can refer to the descriptions in the above embodiments of the data analysis method. For the sake of brevity, they will not be repeated here, but this should not cause any loss of the scope of protection of the present application.

[0141] Further, as shown in Figure 4 , the computing device further includes a communication component 42, a power supply component 43, and other components. Figure 4 Some components are only schematically shown in Figure 4 , and it does not mean that the computing device only includes the components shown in .

[0142] Correspondingly, the embodiments of the present application also provide a computer readable storage medium storing a computer program, which is executed to implement each step that can be executed by the computing device in the above method embodiments.

[0143] Figure 5 A flowchart of a noise construction method provided by an exemplary embodiment of the present application is shown in Figure 5 . The method includes:

[0144] Step 500, in response to the noise construction instruction, obtaining the number of objects contained in the target group;

[0145] Step 501, determining a target noise parameter that is adapted to the number of objects contained in the target group based on a mapping relationship between the number of objects and the noise parameter;

[0146] Step 502, constructing noise for the target group according to the target noise parameter;

[0147] The noise parameter affects the range of the noise constructed for the target group.

[0148] The noise construction method provided in this embodiment can be applied to various scenes that need to construct noise, for example, data analysis and the like. The application scenarios are not limited in this embodiment. In different application scenarios, the specifications of the group for which noise needs to be constructed, the types or fields of the objects contained, the sources, and the like can not be completely the same, and this embodiment does not limit this. For example, in the data analysis scenario, the target group can be a group selected under a certain analysis label, such as Figure 3 the group selected under the "predicted occupation" in the middle.

[0149] In actual applications, under many privacy protection schemes, noise needs to be added to the original value to avoid data leakage. For example, in the differential privacy protection scheme, random noise can be added to each attribute analysis value in the original output result to obtain the output result after differential privacy protection.

[0150] The applicant found in the research process that in the process of calculating noise, the noise parameter has an impact on the range of the noise. For example, in the Laplace mechanism of differential privacy protection, the inverse cumulative function is used to calculate the noise, and the standard deviation can be used as the noise parameter in this embodiment. Of course, the noise parameter in this embodiment is not limited to this. In different noise calculation methods, the noise parameter can not be completely the same.

[0151] For the same original value, if the range of the added noise is too small, the output result after adding noise can be infinitely close to the original value, causing a data leakage risk. If the range of the added noise is too large, the output result after adding noise can not be used.

[0152] Therefore, in this embodiment, the number of objects contained in the target group can be obtained, and a target noise parameter that is adapted to the number of objects contained in the target group can be determined based on a mapping relationship between the number of objects and the noise parameter. In this embodiment, the number of objects contained in the target group can correspond to the original analysis value under the target analysis label in the data analysis method embodiment. Of course, in other application scenarios, the number of objects contained in the target group can also be described as other names, and this embodiment does not limit this.

[0153] Different target groups may contain different numbers of objects, and thus the determined target noise parameters for different target groups may be different. The determined target noise parameters are adapted to the number of objects contained in the target groups. In this way, in this embodiment, the number of objects contained in a target group can be used as a basis to adaptively determine the most suitable noise parameter.

[0154] As mentioned above, the noise parameter affects the noise range, and thus based on a suitable noise parameter, it can be ensured that the range of the constructed noise is adapted to the number of objects contained in the target group. Accordingly, the constructed noise can ensure that the output result after adding noise is neither infinitely close to the original value nor loses usability. The process of constructing noise based on the target noise parameter can refer to the scheme of "determining noise under the target analysis tag based on the target noise parameter" in the foregoing data analysis method embodiment, and will not be described in detail here.

[0155] It should be noted that in this embodiment, the mapping relationship between the number of objects and the noise parameter can be used as a basis to determine the noise parameter through the number of objects.

[0156] In an optional implementation, a noise boundary value satisfying a preset noise proportion requirement can be calculated according to the number of objects contained in the target group; and a noise parameter whose distance to the noise boundary value satisfies a preset condition is selected from at least one noise parameter contained in the mapping relationship between the number of objects and the noise parameter, as the target noise parameter.

[0157] In actual application, a maximum noise value corresponding to the target group can be calculated according to the preset noise upper limit and the number of objects contained in the target group; and a target noise parameter whose distance to the maximum noise value satisfies a preset condition is selected from at least one noise parameter contained in the mapping relationship between the number of objects and the noise parameter. For example, the noise parameter closest to the maximum noise value is selected as the target noise parameter.

[0158] In the process of calculating the maximum noise value corresponding to the target group, a shift number N corresponding to the noise upper limit can be determined, N is a positive integer; and the number of objects contained in the target group is shifted to the left of the decimal point by N bits to obtain the maximum noise value.

[0159] Of course, in this embodiment, other implementation manners can also be used to determine the target noise parameter corresponding to the target group, for example, a mapping relationship between the number of objects, the noise upper limit and the noise parameter is constructed in advance, so that the target parameter is directly determined according to the noise upper limit corresponding to the target group and the number of objects contained in the target group. This embodiment is not limited thereto.

[0160] In this embodiment, at least one noise parameter available for the target group can be determined in advance. In actual application, the mapping relationship between the object quantity and the noise parameter can include the at least one noise parameter available for the target object.

[0161] In some possible implementation manners, according to a scene where the target group is located, a plurality of sample objects can be selected, and a plurality of segmentation values can be determined based on the order of magnitude of the sample objects; and the plurality of segmentation values can be respectively configured with noise parameters to obtain at least one noise parameter corresponding to the target group. For example, in a case where the target group corresponds to the original analysis value under the target analysis label in the data analysis method embodiment, the initial population range in the data analysis method embodiment can be used as the sample object in this embodiment. Accordingly, the specific noise parameter configuration process can refer to the related description in the data analysis method embodiment, which will not be described in detail herein.

[0162] For the selection scheme of the sample object, in this embodiment, no limitation is performed, and in different application scenarios, different ways can be used to select the sample object. For example, the selection manner of the initial population range in the data analysis method embodiment can be used. In addition, the quantity of the sample object can be greater than the quantity of the target group, to support the function of adaptively determining the noise parameter under different groups covered by the sample object.

[0163] It should be noted that the technical details in the above noise construction method embodiments can refer to the description in the data analysis method embodiments, which will not be described in detail herein to save space, but this should not cause loss of the protection scope of the present application.

[0164] Figure 6 Another structural schematic diagram of a computing device provided by another exemplary embodiment of the present application is shown in FIG. 6. As shown in FIG. 6, the computing device includes a memory 60 and a processor 61. Figure 6 The processor 61 is coupled with the memory 60 and is configured to execute a computer program in the memory 60, so as to:

[0165] The processor 61 is coupled with the memory 60 and is configured to execute a computer program in the memory 60, so as to:

[0166] In response to the noise construction instruction, the quantity of the objects included in the target group is obtained;

[0167] Based on the mapping relationship between the object quantity and the noise parameter, a target noise parameter adapted to the quantity of the objects included in the target group is determined;

[0168] According to the target noise parameter, noise is constructed for the target group;

[0169] The noise parameter affects the range of the noise constructed for the target group.

[0170] In an optional embodiment, when determining the target noise parameter adapted to the target group based on the mapping relationship between the number of objects and the noise parameter, the processor 61 is configured to:

[0171] According to the number of objects contained in the target group, a noise boundary value satisfying a preset noise proportion is calculated;

[0172] From the at least one noise parameter, a noise parameter whose distance from the noise boundary value satisfies a preset requirement is selected as the target noise parameter.

[0173] For other technical contents in this embodiment, refer to the descriptions in the foregoing embodiments of the noise construction method and the data analysis method, which will not be repeated here to save space, but this should not cause any loss to the protection scope of the present application.

[0174] Further, as shown in Figure 6 , the computing device further includes a communication component 62, a power component 63, and other components. Figure 6 Only some components are shown in the figure, which does not mean that the computing device only includes the components shown in the figure. Figure 6 The components shown in the figure are only schematic.

[0175] Correspondingly, the embodiments of the present application also provide a computer readable storage medium storing a computer program, which can implement the steps that can be executed by the computing device in the above method embodiments when the computer program is executed.

[0176] The communication component in the above Figure 4 and 6 is configured to facilitate wired or wireless communication between the device where the communication component is located and other devices. The device where the communication component is located can access a wireless network based on a communication standard, such as a WiFi, 2G, 3G, 4G / LTE, 5G, or the like, or a combination thereof. In an example embodiment, the communication component receives broadcast signals or broadcast-related information from an external broadcast management system via a broadcast channel. In an example embodiment, the communication component also includes a near field communication (NFC) module to facilitate short-range communication. For example, the NFC module can be implemented based on radio frequency identification (RFID) technology, infrared data association (IrDA) technology, ultra wide band (UWB) technology, Bluetooth (BT) technology, and other technologies.

[0177] The power component in the above Figure 4 and 6 provides power for various components of the device where the power component is located. The power component can include a power management system, one or more power sources, and other components associated with generating, managing, and distributing power for the device where the power component is located.

[0178] Those skilled in the art will appreciate that embodiments of the application can be readily used as software, hardware, or a combination of software and hardware. In a typical configuration, a computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0179] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks.

[0180] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks.

[0181] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks.

[0182] In a typical configuration, a computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0183] The memory can include non-persistent memory and / or persistent memory, such as flash memory, readonly memory (ROM), and / or the like, in a computer-readable medium. The memory is an example of computer-readable media.

[0184] Computer-readable media includes permanent and non-permanent, movable and non-movable media that can be implemented by any method or technology to store information. The information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information accessible to a computing device. According to the definition herein, computer-readable media does not include transitory media such as modulated data signals and carriers.

[0185] It should also be noted that the terms "comprising", "containing", or any other variant thereof are intended to cover non-exclusive inclusions, so that a process, method, article or apparatus that includes a list of elements does not only include those elements, but also includes other elements not explicitly listed, or further includes elements inherent in such a process, method, article or apparatus. Without more limitations, the element defined by the statement "comprising a" does not exclude the presence of additional identical elements in the process, method, article or apparatus that includes the element.

[0186] The above only describes the embodiments of the present application and is not intended to limit the present application. Those skilled in the art can make various changes and modifications to the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A data analysis method, characterized by, The method comprises the following steps: receiving a data analysis request, wherein the data analysis request comprises at least one analysis label; obtaining an original analysis value under a target analysis label, wherein the target analysis label is any one of the at least one analysis label; selecting a target noise parameter matching the original analysis value from at least one candidate noise parameter corresponding to the target analysis label; determining noise under the target analysis label based on the target noise parameter; adding the noise to the original analysis value under the target analysis label to obtain an analysis result under the target analysis label.

2. The method of claim 1, wherein, Under a Laplace mechanism, the noise parameter is a standard deviation.

3. The method of claim 1, wherein, The step of obtaining the original analysis value under the target analysis label comprises the following steps: if the target analysis label comprises a plurality of analysis attributes, obtaining attribute analysis values corresponding to the plurality of analysis attributes respectively; determining the original analysis value under the target analysis label according to the attribute analysis values corresponding to the plurality of analysis attributes respectively.

4. The method of claim 3, wherein, The step of determining the original analysis value under the target analysis label according to the attribute analysis values corresponding to the plurality of analysis attributes respectively comprises the following step: taking a median value in the attribute analysis values corresponding to the plurality of analysis attributes respectively as the original analysis value under the target analysis label.

5. The method of claim 3, wherein, The step of obtaining the attribute analysis values corresponding to the plurality of analysis attributes respectively comprises the following steps: the data analysis request further comprises circle selection condition information, a target analysis label, and analysis attributes under the target analysis label; determining a target group meeting the circle selection condition information from an initial group range; performing data analysis on the target group according to the target analysis label and the analysis attributes under the target analysis label to obtain the attribute analysis values corresponding to the plurality of analysis attributes respectively.

6. The method of claim 5, wherein, The method further comprises the following steps before the step of selecting the target noise parameter from the at least one candidate noise parameter corresponding to the target analysis label: determining a plurality of segmentation values based on the order of magnitude of the initial group range; configuring noise parameters for the plurality of segmentation values respectively to obtain the at least one noise parameter corresponding to the initial group range as the at least one candidate noise parameter corresponding to the target analysis label.

7. The method of claim 3, wherein, The step of determining the noise under the target analysis label based on the target noise parameter comprises the following steps: if the target analysis label comprises a plurality of analysis attributes, obtaining noise factors corresponding to the plurality of analysis attributes respectively; calculating the noise corresponding to the plurality of analysis attributes respectively based on the target noise parameter and the noise factors corresponding to the plurality of analysis attributes respectively.

8. The method of claim 7, wherein, The step of obtaining the noise factors corresponding to the plurality of analysis attributes respectively comprises the following steps: obtaining a noise seed corresponding to the circle selection condition information associated with the target analysis label; generating noise factors for the at least one analysis attribute respectively based on the noise seed.

9. The method of claim 7, wherein, The step of adding the noise to the original analysis value under the target analysis label comprises the following step: adding the noise corresponding to the plurality of analysis attributes respectively to the attribute analysis values corresponding to the corresponding analysis attributes respectively.

10. The method of claim 1, wherein, The selecting, from the at least one candidate noise parameter corresponding to the target analysis label, a target noise parameter matching the original analysis value comprises: calculating a maximum noise value corresponding to the original analysis value according to a preset noise ratio upper limit; selecting, from the at least one candidate noise parameter corresponding to the target analysis label, a target noise parameter whose distance from the maximum noise value meets a preset requirement.

11. The method of claim 10, wherein, The preset requirement comprises: being closest to the maximum noise value and being less than the maximum noise value.

12. The method of claim 10, wherein, The calculating a maximum noise value corresponding to the original analysis value according to a preset noise ratio upper limit comprises: determining a shift number N corresponding to the noise ratio upper limit, N being a positive integer; shifting the decimal point of the original analysis value to the left by N bits to obtain the maximum noise value.

13. The method of claim 1, wherein, Further comprising: configuring a budget for each of the at least one noise parameter.

14. A computing device, comprising: comprising a memory and a processor; the memory is used to store one or more computer instructions; the processor is coupled with the memory and is used to execute the one or more computer instructions for: receiving a data analysis request, the data analysis request containing at least one analysis label; obtaining an original analysis value under a target analysis label, the target analysis label being any one of the at least one analysis label; selecting, from at least one candidate noise parameter corresponding to the target analysis label, a target noise parameter matching the original analysis value; determining noise under the target analysis label based on the target noise parameter; adding the noise to the original analysis value under the target analysis label to obtain an analysis result under the target analysis label.

15. A computer readable storage medium storing computer instructions, wherein, When the computer instructions are executed by one or more processors, the one or more processors execute the data analysis method of any one of claims 1-13.

Images