Data acquisition method and device based on privacy protection

Through random slicing and hidden query protocols, the privacy exposure of models and filtered values ​​during data acquisition is solved, ensuring the privacy protection of data holders and model holders, and realizing the security of data acquisition.

CN118916392APending Publication Date: 2024-11-08LANXIANG ZHILIAN (HANGZHOU) TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411069972.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-06
Publication Date
2024-11-08

AI Technical Summary

Technical Problem

The prior art cannot effectively protect the privacy of data holders and model holders during the data acquisition process, especially when filtering data, it is impossible to avoid the exposure of models and filtered values, and it is impossible to ensure that the data holders do not know the model holders. Filtered data.

Method used

Random slicing technology is used to slice the model and data, and the hidden query protocol is used to ensure that the data holder can only obtain the filtered data ID and random slices. The model holder maintains the privacy of the filtered data.

Benefits of technology

Privacy protection of the model and filtered values ​​is achieved, ensuring that the data holder cannot know the specific data filtered out by the model holder and the final data obtained, and protecting the privacy of both parties.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118916392A_ABST
    Figure CN118916392A_ABST
Patent Text Reader

Abstract

The invention discloses a privacy protection-based data acquisition method, which comprises the following steps that: S11, a model holder randomly slices an LR model and stores first data; s12, the model holder receives second data sent by the data holder, and performs calculation to obtain slice data of the first classification result; s13, the model holder processes the slice data of the first classification result to obtain a first processing result; s14, the model holder receives the second processing result sent by the data holder and carries out calculation to obtain a third processing result; s15, the model holder compares the third processing result with the rule symbol, and if the third processing result is equal to the rule symbol, the ID is put into a set; s16, the model holder receives the code database sent by the data holder; s17, the model holder generates a query value and a query state for the coding database according to the data in the set; s18, the model holder receives a response result generated by the data holder; and S19, the model holder obtains required data according to a response result.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a data acquisition method and device based on privacy protection. Background Art

[0002] Party A holds the data, and Party B holds the logistic regression (LR) model and screening rules. Party B classifies Party A's data according to the LR model, and filters out data based on the screening values ​​on the classified data to obtain the data. Party A's data is sensitive data and cannot be fully disclosed to Party B. Only data that meets specific values ​​can be given to Party B. Party B's LR model and screening values ​​are commercial secrets and cannot be disclosed to Party A.

[0003] Currently, there is no relevant technology that can completely solve the following problems:

[0004] 1. During the data screening process, Party B’s LR model and screening value cannot be exposed to Party A;

[0005] 2.B can only obtain the filtered data, not other data;

[0006] 3. Party A cannot know which data B has screened and ultimately obtained.

[0007] In view of the above technical problems, the present invention proposes a data acquisition method and device based on privacy protection. Summary of the invention

[0008] The purpose of the present invention is to provide a data acquisition method and device based on privacy protection in view of the defects of the prior art.

[0009] In order to achieve the above purpose, the present invention adopts the following technical solutions:

[0010] A data acquisition method based on privacy protection, comprising:

[0011] S11. The model holder randomly slices the LR model and stores the first data obtained after random slicing;

[0012] S12. The model holder receives the second data obtained by randomly slicing the database sent by the data holder, and the model holder calculates the first data and the second data to obtain slice data of the first classification result;

[0013] S13. The model holder processes the slice data of the first classification result based on the rule model to obtain a first processing result;

[0014] S14. The model holder receives the second processing result obtained based on the rule model sent by the data holder, and calculates the first processing result and the second processing result to obtain a third processing result;

[0015] S15. The model holder compares the third processing result with the rule symbol in the rule model. If the third processing result is equal to the rule symbol, the ID corresponding to the third processing result is placed in the set;

[0016] S16. The model holder receives the encoding database for encoding the database sent by the data holder;

[0017] S17. The model holder generates a query value and a query status for the encoded database based on the data in the collection and the hidden query protocol, and sends the query value and the query status to the data holder;

[0018] S18. The model holder receives the response result generated by the data holder according to the query value and query status;

[0019] S19. The model holder obtains the required data based on the response result.

[0020] Furthermore, the step S11 specifically includes:

[0021] The model holder has a LR model model1 = (a1, a2, ..., a m ) to perform random slicing and obtain the result a after random slicing j =[a j ]0+[a j ]1; the model holder stores itself [a j ]1, send [a j ]0 to the data holder; where j∈1,2,…,m; model1 represents the LR model; a1,a2,…,a m Represents the m data of the model; a j Represents the j-th data of the model.

[0022] Furthermore, the second data obtained by randomly slicing the database and sent by the data holder in step S12 is specifically:

[0023] The data holder has a database D = d1, d2, ..., d n Each piece of data d i =(ID i ,f i1 ,f i2 ,…,f im ) i1 ,f i2 ,…,f imPerform random slicing to obtain the result f after random slicing ij =[f ij ]0+[f ij ]1Data holder’s own storage ij ]0, send [f ij ]1 to the model holder; where i∈1,2,…,n, j∈1,2,…,m; D represents the database; d1,d2,…,d n Represents n pieces of data in the database; d i Indicates the i-th data; ID i Indicates the ID of the i-th data; f i1 ,f i2 ,…,f im represents the m contents of the i-th data; f ij Represents the jth content of the i-th data.

[0024] Furthermore, in step S12, the model holder calculates the first data and the second data to obtain slice data of the first classification result, specifically:

[0025] The model holder will ij ]1 and the LR model [a j ]1 Calculate the slice data of the first classification result, expressed as:

[0026]

[0027] in, represents the multiplication operation between random slices; [y i ]1 represents the slice data of the first classification result; [a j ]1 represents the first data after random slicing of the LR model stored by the model holder itself; [f ij ]1 represents the second data after random slicing of database D sent by the data holder to the model holder.

[0028] Furthermore, the step S13 is specifically as follows:

[0029] The model holder obtains the screening content content1 based on the rule model, and processes the slice data of the first classification result based on the screening content content1 to obtain the first processing result, which is expressed as:

[0030] LTS([f ij ]1,[value1]1)=r i1

[0031] Among them, r i1represents the first processing result; LTS() represents the LTS algorithm; value1 represents the rule value in the rule model.

[0032] Furthermore, the step S14 is specifically as follows:

[0033] The data holder shall ij ]0 and [a j ]0Calculate the slice data of the second classification result, expressed as:

[0034]

[0035] Among them, [y i ]0 represents the slice data of the second classification result; [a j ]0 represents the data after random slicing of the LR model sent by the model owner to the data owner; [f ij ]0 indicates data stored by the data holder itself;

[0036] The model holder performs random slicing on the rule value value1, which is expressed as:

[0037] value1=[value1]0+[value1]1

[0038] The model holder sends the filtered content content1, rule symbol sign1, and rule value [value1]0 to the data holder based on the rule model;

[0039] The data holder processes the sliced ​​data of the second classification result according to the screening content content1 to obtain the second processing result, which is expressed as:

[0040] LTS([f ij ]0,[value1]0)=r i0

[0041] Among them, r i0 Indicates the second processing result;

[0042] The data holder calculates the first processing result and the second processing result, which is expressed as:

[0043] r i =r i0 +r i1

[0044] Among them, r i Indicates the third processing result.

[0045] Furthermore, the encoding database for encoding the database sent by the data holder in step S16 is represented as:

[0046] pp D ←PIR.setup(1 λ ,D)

[0047] Among them, pp D represents the encoded database; PIR() represents the hidden query protocol; setup() encoding; D represents the database; λ represents the security parameter.

[0048] Furthermore, in step S19, the model holder obtains the required data according to the response result, which is expressed as:

[0049] In step S19, the model holder obtains the required data according to the response result, which is expressed as:

[0050] (q,st)←PIR.query(pp D ,ID)

[0051] r←PIR.respond(pp D ,D,q)

[0052] d←PIR.process(pp D ,st,r)

[0053] Among them, query represents the query algorithm, and its input is the encoded database pp D and the ID of the data to be queried, its output is the query q and the query status st; responde represents the response algorithm, its input is the encoded database pp D , database D and query q, the output is the response result r; process represents the processing algorithm, and its input is the encoded database pp D , query state st, response result r, its output is the required data d.

[0054] Accordingly, a privacy-preserving data acquisition method is also provided, including:

[0055] S21. The data holder randomly slices the database and stores the second data obtained after random slicing;

[0056] S22. The data holder receives the first data obtained by random slicing of the LR model from the model holder, and the data holder calculates the first data and the second data to obtain slice data of the second classification result;

[0057] S23. The data holder processes the slice data of the second classification result based on the rule model to obtain a second processing result;

[0058] S24. The data holder receives the first processing result obtained based on the rule model sent by the model holder, and the data holder receives the third processing result obtained by calculating the first processing result and the second processing result by the model holder;

[0059] S25. The data holder encodes the database based on the hidden query protocol to obtain an encoded database, and sends the encoded database to the model holder;

[0060] S26. The data holder receives the query value and query status generated by the model holder according to the third processing result and the encoding database;

[0061] S27. The data holder generates a response result based on the query value and query status, and sends the corresponding result to the model holder. The model holder obtains the required data based on the response result.

[0062] Accordingly, a data acquisition device based on privacy protection is also provided, which is used to execute the data acquisition method based on privacy protection. The data acquisition device includes a processor, and the processor is provided with:

[0063] A slicing module, used by the model holder to randomly slice the LR model and store the first data obtained after random slicing;

[0064] A first calculation module is used for the model holder to receive second data sent by the data holder after random slicing of the database, and the model holder calculates the first data and the second data to obtain slice data of the first classification result;

[0065] A processing module, used by the model holder to process the slice data of the first classification result based on the rule model to obtain a first processing result;

[0066] A second calculation module is used for the model holder to receive the second processing result obtained based on the rule model sent by the data holder, and calculate the first processing result and the second processing result to obtain a third processing result;

[0067] A comparison module, used for the model holder to compare the third processing result with the rule symbol in the rule model, and if the third processing result is equal to the rule symbol, the ID corresponding to the third processing result is put into the set;

[0068] An encoding module, used for the model holder to receive an encoding database sent by the data holder for encoding the database;

[0069] The query module is used by the model holder to generate a query value and a query status for the encoded database according to the data in the collection and based on the hidden query protocol, and send the query value and the query status to the data holder;

[0070] The corresponding module is used for the model holder to receive the response result generated by the data holder according to the query value and query status;

[0071] The acquisition module is used by the model holder to obtain the required data based on the response results.

[0072] Compared with the prior art, the present invention has the following beneficial effects:

[0073] 1. In the process of filtering data, the data holder obtains random slices of the classification model, and random slices are random values ​​that have no meaning, thus protecting the privacy of the classification model of the model holder. In the filtering model, the data holder only obtains random slices of the filtering value, not the complete random value, so this solution also protects the random value.

[0074] 2. In the process of filtering data, the model holder only obtains the set S of IDs of the filtered data and random slices of all data, and random slices are a string of random numbers that have no meaning. In the data acquisition stage, the model holder only obtains the data corresponding to the ID in the set S, that is, the filtered data, and no other data.

[0075] 3. The sets S1, S2, …, S1 are always maintained by the model holder, and the data holder does not know any information about the set, thus ensuring that the data holder cannot know which data the model holder has filtered out. The hidden query protocol ensures that the data holder cannot know which data the model holder has finally obtained. BRIEF DESCRIPTION OF THE DRAWINGS

[0076] Figure 1 This is a flow chart of a data acquisition method based on privacy protection provided in Example 1. DETAILED DESCRIPTION

[0077] The following describes the embodiments of the present invention by specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the following embodiments and features in the embodiments can be combined with each other without conflict.

[0078] The purpose of the present invention is to provide a data acquisition method and device based on privacy protection in view of the defects of the prior art.

[0079] Embodiment 1

[0080] This embodiment provides a data acquisition method based on privacy protection, such as Figure 1 As shown, including:

[0081] S11. The model holder randomly slices the LR model and stores the first data obtained after random slicing;

[0082] S12. The model holder receives the second data obtained by randomly slicing the database sent by the data holder, and the model holder calculates the first data and the second data to obtain slice data of the first classification result;

[0083] S13. The model holder processes the slice data of the first classification result based on the rule model to obtain a first processing result;

[0084] S14. The model holder receives the second processing result obtained based on the rule model sent by the data holder, and calculates the first processing result and the second processing result to obtain a third processing result;

[0085] S15. The model holder compares the third processing result with the rule symbol in the rule model. If the third processing result is equal to the rule symbol, the ID corresponding to the third processing result is placed in the set;

[0086] S16. The model holder receives the encoding database for encoding the database sent by the data holder;

[0087] S17. The model holder generates a query value and a query status for the encoded database based on the data in the collection and the hidden query protocol, and sends the query value and the query status to the data holder;

[0088] S18. The model holder receives the response result generated by the data holder according to the query value and query status;

[0089] S19. The model holder obtains the required data based on the response result.

[0090] This embodiment is described by taking the model holder as an example.

[0091] This embodiment involves a model holder and a data holder, wherein the model holder holds an LR model and a rule model, and the data holder holds a database; this embodiment defines the data holder as party A, the model holder as party B, and party A holds database D, and party B holds the LR model model1 and the rule model model2 for illustration.

[0092] This embodiment includes four stages: a preparation stage, a data classification stage, a classification result screening stage, and a data acquisition stage.

[0093] Preparatory stage:

[0094] Party A holds database D, which contains n pieces of m-dimensional data, D = d1, d2, …, d n , where the i-th data can be represented as d i =(ID i ,f i1 ,f i2 ,…,f im ), where ID i Indicates the ID of the i-th data, f i1 ,f i2 ,…,f im Represents the m contents of the i-th data; i∈1,2,…,n.

[0095] Party B holds the LR model model1 = (a1, a2, ..., a m ), a1, a2, …, a m Represents the m data of the model; the LR model can be used to analyze the data (f1, f2, ..., f m ) to classify: y is the classification result.

[0096] Party B holds the rule model model2 = (r1, r2, ..., r l ), r1, r2, …, r l Represents the l data of the model. The rules r1, r2, ..., r in the rule model l The details are as follows: For the rule r = (context, sign, value), it contains three pieces of information, namely: the content to be filtered (it can be based on a certain feature f i (i=1,…,m) for screening, or for classification result y), rule symbol sign, rule value value. The rule symbol sign has two values, one is 0, which means less than, and the other is 1, which means greater than or equal to. In essence, rule r is used to screen data that is less than the rule value, or data that is greater than the rule value.

[0097] Data classification stage:

[0098] In step S11, the model holder randomly slices the LR model and stores the first data obtained after random slicing.

[0099] Party B’s LR model model1=(a1,a2,…,a m ) for random slicing, i.e. a1,a2,…,a m Randomly split into a j =[a j ]0+[a j ]1, j∈1,2,…,m, Party B retains [aj ]1, send [a j ]0 to Party A.

[0100] Among them, a j Represents the jth data of the model, a j ∈a1,a2,…,a m , a1=[a1]0+[a1]1, a2=[a2]0+[a2]1,…,a m =[a m ]0+[a m ]1; that is, Party B reserves ([a1]1,[a2]1,…,[a m ]1), send ([a1]0,[a2]0,…,[a m ]0) to Party A.

[0101] In this embodiment, the data holder also randomly slices the database, and sends the second data obtained after random slicing to the data holder.

[0102] Party A has a certain amount of data in database D. i =(ID i ,f i1 ,f i2 ,…,f im ) do the following: i1 ,f i2 ,…,f im Perform random slicing, that is, f i1 ,f i2 ,…,f im Randomly split into f ij =[f ij ]0+[f ij ]1, j∈1,2,…,m, Party A retains [f ij ]0, send [f ij ]1 to Party B.

[0103] Among them, f ij Indicates the jth content of the i-th data; f ij ∈f i1 ,f i2 ,…,f im , f i1 =[f i1 ]0+[f i1 ]1,f i2 =[f i2 ]0+[f i2 ]1,…,f im =[f im ]0+[f im ]1; that is, Party A retains ([fi1 ]0,[f i2 ]0,...,[f im ]0), send ([f i1 ]1,[f i2 ]1,...,[f im ]1) To Party B.

[0104] In step S12, the model holder receives the second data obtained by randomly slicing the database from the data holder, and the model holder calculates the first data and the second data to obtain the slice data of the first classification result.

[0105] Party B based on each piece of data d i =(ID i ,f i1 ,f i2 ,…,f im ) of random slices [f ij ]1∈([f i1 ]1,[f i2 ]1,...,[f im ]1) and LR model model1=(a1,a2,…,a m ) of random slices [a j ]1∈([a1]1,[a2]1,…,[a m ]1) Calculate the slice of the classification result to obtain the slice data of the first classification result, which is expressed as:

[0106]

[0107] in, Represents the multiplication operation between random slices. For the specific principle, refer to Algorithm 3 in the paper Actively Secure Two-Party Computation: Efficient Beaver Triple Generation; [y i ]1 represents the slice data of the first classification result; [a j ]1 represents the first data after random slicing of the LR model stored by Party B itself; [f ij ]1 represents the second data sent by Party A to Party B after random slicing of database D.

[0108] In this embodiment, the data holder also calculates its own data and the data sent by the model holder to obtain slice data of the second classification result.

[0109] Party A based on each piece of data d i =(ID i ,f i1, f i2 , …, f im )'s random slice [f ij 0 ∈ ([f i1 0, [f i2 0, ..., [f im 0) and the LR model model1 = (a1, a2, …, a m )'s random slice [a j 0 ∈ ([a1]0, [a2]0, …, [a m 0) calculates the slice of the classification result to obtain the slice data of the second classification result, denoted as:

[0110]

[0111] Among them, [y i 0 represents the slice data of the second classification result; [a j 0 represents the data after the random slice of the LR model sent by Party B to Party A; [f ij 0 represents the data stored by Party A itself.

[0112] Classification result screening stage:

[0113] For the rule model model2 = (r1, r2, …, r l ), in this embodiment, the LTS algorithm (for the specific principle, refer to the paper: Rabbit: Efficient Comparison for Secure Multi-Party Computation) is used to implement. The input of the LTS algorithm is the random slices of the values x and y to be compared. One party executes LTS([x]0, [y]0) = r0, and the other party executes LTS([x]1, [y]1) = r1. Both parties disclose r0 and r1 and calculate r = r0 + r1. If r = 0, it means x < y. If r = 1, it means x ≥ y. For the screening rule r1 = (content1, sign1, value1), the specific screening steps for both parties are as follows:

[0114] In step S13, the model holder processes the slice data of the first classification result based on the rule model to obtain the first processing result.

[0115] Party B notifies Party A of the content content1 and the rule symbol sign1 to be screened, and randomly slices value1 as value1 = [value1]0 + [value1]1. Party B sends [value1]0 to Party A.

[0116] Party B and Party A determine which column of the data to screen according to content1. In this embodiment, the j-th column is taken as an example.

[0117] B processes all data using the LTS algorithm to obtain the first processing result, which is expressed as:

[0118] LTS)[f ij ]1,[value1]1)=r i1

[0119] Among them, r i1 Represents the first processing result; r i1 ∈r 11 , r 21 ,…,r n1 ; r 11 =LTS([f 1j ]1,[value1]1),r 21 =LTS([f 2j ]1,[value1]1),…,r n1 =LTS([f nj ]1,[value1]1); LTS() represents the LTS algorithm; value1 represents the rule value in the rule model.

[0120] A processes all data through the LTS algorithm to obtain the second processing result, which is expressed as:

[0121] LTS([f ij ]0,[value1]0)=r i0

[0122] Among them, r i0 Represents the second processing result; r i0 ∈r 10 , r 20 ,…,r n0 ; r 10 =LTS([f 1j ]0,[value1]0),r 20 =LTS([f 2j ]0,[value1]0),…,r n0 =LTS([f nj ]0,[value1]0).

[0123] Party A sends r i0 ∈r 10 , r 20 ,…,r n0 To Party B.

[0124] In step S14, the model holder receives the second processing result obtained based on the rule model sent by the data holder, and calculates the first processing result and the second processing result to obtain a third processing result.

[0125] r i =r i0 +r i1

[0126] Among them, r i Represents the third processing result; r i ∈r1,r2,…,r n , r1=r 10 +r 11 ,r2=r 20 +r 21 ,...,r n =r n0 +r n1 .

[0127] In step S15, the model holder compares the third processing result with the rule symbol in the rule model. If the third processing result is equal to the rule symbol, the ID corresponding to the third processing result is put into the set.

[0128] Party B replaces r1, r2, ..., r n Compare with the rule sign sign1, if r1, r2, …, r n If any value in is equal to sign1, the data ID corresponding to the current value is put into set S1. 14 If it is equal to sign1, the ID of the 14th data 14 Put it into set S1.

[0129] Using the method of step S13, using the rule model r2, r3, ..., r l Filter all data to obtain sets S2, S3,…, Sl.

[0130] Party B calculates the intersection S = S1 ∩ S2 ... ∩ S1. Set S contains the IDs corresponding to all the data filtered by Party B.

[0131] Data acquisition phase:

[0132] Party A and Party B run a hidden query protocol (the protocol principle refers to the paper Call Me By My Name: Simple, Practical Private Information Retrieval for Keyword Queries) to allow Party B to obtain the data corresponding to the ID in set S, and Party A cannot obtain what data Party B has obtained. Specifically:

[0133] In step S16, the model holder receives the encoded database sent by the data holder for encoding the database.

[0134] Party A runs the configuration algorithm of the hidden query protocol to preprocess the database D and generate the encoded database ppD, that is, pp D ←PIR.setup(1 λ ,D), where λ represents the security parameter; PIR() represents the hidden query protocol; setup() is encoded. Party A sends the encoded database ppD to Party B.

[0135] In step S17, the model holder generates a query value and a query status for the encoded database according to the data in the collection and based on the hidden query protocol, and sends the query value and the query status to the data holder.

[0136] Party B runs the query algorithm of the hidden query protocol to generate query q and query state st: (q,st)←PIR.query(pp D ,ID), where ID represents the ID of the element in set S; query represents the query algorithm. Party B sends the query q and the query status st to Party A.

[0137] In step S18, the model holder receives the response result generated by the data holder according to the query value and query status.

[0138] Party A runs the response algorithm of the hidden query protocol to generate a response r: r←PIR.respond(pp D ,D,q), where responde represents the response algorithm and D represents the database. Party A returns the response r to Party B.

[0139] In step S19, the model holder obtains the required data according to the response result.

[0140] Party B runs the processing algorithm of the hidden query protocol to obtain data d from the response r: d←PIR.process(pp D ,st,r), where process represents the processing algorithm.

[0141] Compared with the prior art, this embodiment has the following beneficial effects:

[0142] 1. In the process of filtering data, the data holder obtains random slices of the classification model, and random slices are random values ​​that have no meaning, thus protecting the privacy of the classification model of the model holder. In the filtering model, the data holder only obtains random slices of the filtering value, not the complete random value, so this solution also protects the random value.

[0143] 2. In the process of filtering data, the model holder only obtains the set S of IDs of the filtered data and random slices of all data, and random slices are a string of random numbers that have no meaning. In the data acquisition stage, the model holder only obtains the data corresponding to the ID in the set S, that is, the filtered data, and no other data.

[0144] 3. The sets S1, S2, …, S1 are always maintained by the model holder, and the data holder does not know any information about the set, thus ensuring that the data holder cannot know which data the model holder has filtered out. The hidden query protocol ensures that the data holder cannot know which data the model holder has finally obtained.

[0145] Embodiment 2

[0146] This embodiment provides a privacy-preserving data acquisition method which is different from the first embodiment in that:

[0147] This embodiment is described by taking the data holder as an example. Data acquisition includes:

[0148] S21. The data holder randomly slices the database and stores the second data obtained after random slicing;

[0149] S22. The data holder receives the first data obtained by random slicing of the LR model from the model holder, and the data holder calculates the first data and the second data to obtain slice data of the second classification result;

[0150] S23. The data holder processes the slice data of the second classification result based on the rule model to obtain a second processing result;

[0151] S24. The data holder receives the first processing result obtained based on the rule model sent by the model holder, and the data holder receives the third processing result obtained by calculating the first processing result and the second processing result by the model holder;

[0152] S25. The data holder encodes the database based on the hidden query protocol to obtain an encoded database, and sends the encoded database to the model holder;

[0153] S26. The data holder receives the query value and query status generated by the model holder according to the third processing result and the encoding database;

[0154] S27. The data holder generates a response result based on the query value and query status, and sends the corresponding result to the model holder. The model holder obtains the required data based on the response result.

[0155] It should be noted that the privacy-protected data acquisition method of this embodiment is similar to the implementation method of the first embodiment, and will not be elaborated in this embodiment.

[0156] Embodiment 3

[0157] This embodiment provides a data acquisition device based on privacy protection, which is used to execute a data acquisition method based on privacy protection in embodiment 1. The data acquisition device includes a processor, and the processor is provided with:

[0158] A slicing module, used by the model holder to randomly slice the LR model and store the first data obtained after random slicing;

[0159] A first calculation module is used for the model holder to receive second data sent by the data holder after random slicing of the database, and the model holder calculates the first data and the second data to obtain slice data of the first classification result;

[0160] A processing module, used by the model holder to process the slice data of the first classification result based on the rule model to obtain a first processing result;

[0161] A second calculation module is used for the model holder to receive the second processing result obtained based on the rule model sent by the data holder, and calculate the first processing result and the second processing result to obtain a third processing result;

[0162] A comparison module, used for the model holder to compare the third processing result with the rule symbol in the rule model, and if the third processing result is equal to the rule symbol, the ID corresponding to the third processing result is put into the set;

[0163] An encoding module, used for the model holder to receive an encoding database sent by the data holder for encoding the database;

[0164] The query module is used by the model holder to generate a query value and a query status for the encoded database according to the data in the collection and based on the hidden query protocol, and send the query value and the query status to the data holder;

[0165] The corresponding module is used for the model holder to receive the response result generated by the data holder according to the query value and query status;

[0166] The acquisition module is used by the model holder to obtain the required data based on the response results.

[0167] The processor described in this embodiment may include a central processing unit (CPU), or an application specific integrated circuit (ASIC), or may be configured to implement one or more integrated circuits of the embodiments of the present application.

[0168] The processor described in the present embodiment is connected with a memory for storing data or instructions. The memory may include a large capacity memory for data or instructions. For example, but not limitation, the memory may include a hard disk drive (HDD), a floppy disk drive, a solid state drive (SSD), a flash memory, an optical disk, a magneto-optical disk, a magnetic tape, or a universal serial bus (USB) drive or a combination of two or more of these. In appropriate cases, the memory may include a removable or non-removable (or fixed) medium. In appropriate cases, the memory may be inside or outside the data processing device. In a specific embodiment, the memory is a non-volatile memory. In a specific embodiment, the memory includes a read-only memory (ROM) and a random access memory (RAM). Where appropriate, the ROM may be a mask-programmed ROM, a programmable ROM (Programmable Read-Only Memory, PROM for short), an erasable PROM (Erasable Programmable Read-Only Memory, EPROM for short), an electrically erasable PROM (Electrically Erasable Programmable Read-Only Memory, EEPROM for short), an electrically alterable ROM (Electrically Alterable Read-Only Memory, EAROM for short) or a flash memory (FLASH) or a combination of two or more of these. Under appropriate circumstances, the RAM can be a static random access memory (SRAM) or a dynamic random access memory (DRAM), wherein the DRAM can be a fast page mode dynamic random access memory (FPMDRAM), an extended data output dynamic random access memory (EDODRAM), a synchronous dynamic random access memory (SDRAM), etc.

[0169] The memory can be used to store or cache various data files that need to be processed and / or used for communication, as well as possible computer program instructions executed by the processor. The processor reads and executes the computer program instructions stored in the memory.

[0170] The processor described in this embodiment is connected to a transmission device for transmitting data with an external device. The transmission device can be used to receive or send data via a network. The above-mentioned specific examples of the network may include a wired or wireless network provided by a communication provider of the electronic device. In one example, the transmission device includes a network adapter (Network Interface Controller, referred to as NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device can be a radio frequency (Radio Frequency, referred to as RF) module, which is used to communicate with the Internet wirelessly.

[0171] The processor in this embodiment is connected to an input / output device for inputting or outputting information. The input / output device is used to input or output information. In this embodiment, the input information is a database, an LR model, and a rule model, and the output information is the required data obtained by the model holder.

[0172] Note that the above are only preferred embodiments of the present invention and the technical principles used. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and that various obvious changes, readjustments and substitutions can be made by those skilled in the art without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in more detail through the above embodiments, the present invention is not limited to the above embodiments, and may include more other equivalent embodiments without departing from the concept of the present invention, and the scope of the present invention is determined by the scope of the appended claims.

Claims

1. A data acquisition method based on privacy protection, characterized in that: include: S11. The model holder randomly slices the LR model and stores the first data obtained after random slicing; S12. The model holder receives the second data obtained by randomly slicing the database sent by the data holder, and the model holder calculates the first data and the second data to obtain slice data of the first classification result; S13. The model holder processes the slice data of the first classification result based on the rule model to obtain a first processing result; S14. The model holder receives the second processing result obtained based on the rule model sent by the data holder, and calculates the first processing result and the second processing result to obtain a third processing result; S15. The model holder compares the third processing result with the rule symbol in the rule model. If the third processing result is equal to the rule symbol, the ID corresponding to the third processing result is placed in the set; S16. The model holder receives the encoding database for encoding the database sent by the data holder; S17. The model holder generates a query value and a query status for the encoded database based on the data in the collection and the hidden query protocol, and sends the query value and the query status to the data holder; S18. The model holder receives the response result generated by the data holder according to the query value and query status; S19. The model holder obtains the required data based on the response result.

2. According to the data acquisition method based on privacy protection according to claim 1, it is characterized in that: The step S11 specifically includes: The model holder has a LR model model1 = (a1, a2, ..., a m ) to perform random slicing and obtain the result a after random slicing j =[a j ]0+[a j ]1; the model holder stores itself [a j ]1, send [a j ]0 to the data holder; where j∈1,2,…,m; model1 represents the LR model; a1,a2,…,a m Represents the m data of the model; a j Represents the j-th data of the model.

3. A data acquisition method based on privacy protection according to claim 2, characterized in that: The second data obtained by randomly slicing the database and sent by the data holder in step S12 is specifically: The data holder has a database D = d1, d2, ..., d n Each piece of data d i =(ID i ,f i1 ,f i2 ,…,f im ) i1 ,f i2 ,…,f im Perform random slicing to obtain the result f after random slicing ij =[f ij ]0+[f ij ]1Data holder’s own storage ij ]0, send [f ij ]1 to the model holder; where i∈1,2,…,n, j∈1,2,…,m; D represents the database; d1,d2,…,d n Represents n pieces of data in the database; d i Indicates the i-th data; ID i Indicates the ID of the i-th data; f i1 ,f i2 ,…,f im represents the m contents of the i-th data; f ij Represents the jth content of the i-th data.

4. A data acquisition method based on privacy protection according to claim 3, characterized in that: In step S12, the model holder calculates the first data and the second data to obtain slice data of the first classification result, specifically: The model holder will ij ]1 and the LR model [a j ]1 Calculate the slice data of the first classification result, expressed as: in, represents the multiplication operation between random slices; [y i ]1 represents the slice data of the first classification result; [a j ]1 represents the first data after random slicing of the LR model stored by the model holder itself; [f ij ]1 represents the second data after random slicing of database D sent by the data holder to the model holder.

5. A data acquisition method based on privacy protection according to claim 4, characterized in that: The step S13 is specifically as follows: The model holder obtains the screening content content1 based on the rule model, and processes the slice data of the first classification result based on the screening content content1 to obtain the first processing result, which is expressed as: LTS)[f ij ]1,[value1]1)=r i1 Among them, r i1 represents the first processing result; LTS() represents the LTS algorithm; value1 represents the rule value in the rule model.

6. A privacy protection-based data acquisition method according to claim 5, characterized in that: The step S14 is specifically as follows: The data holder shall ij ]0 and the LR model [a j ]0Calculate the slice data of the second classification result, expressed as: Among them, [y i ]0 represents the slice data of the second classification result; [a j ]0 represents the data after random slicing of the LR model sent by the model owner to the data owner; [f ij ]0 indicates data stored by the data holder itself; The model holder performs random slicing on the rule value value1, which is expressed as: value1=[value1]0+[value1]1 The model holder sends the filtered content content1, rule symbol sign1, and rule value [value1]0 to the data holder based on the rule model; The data holder processes the sliced ​​data of the second classification result according to the screening content content1 to obtain the second processing result, which is expressed as: LTS)[f ij ]0,[value1]0)=r i0 Among them, r i0 Indicates the second processing result; The data holder calculates the first processing result and the second processing result, which is expressed as: r i =r i0 +r i1 Among them, r i Indicates the third processing result.

7. A data acquisition method based on privacy protection according to claim 6, characterized in that: The coded database for encoding the database sent by the data holder in step S16 is represented as: pp D ←PIR.setup(1 λ ,D) Among them, pp D represents the encoded database; PIR() represents the hidden query protocol; setup() encoding; D represents the database; λ represents the security parameter.

8. A data acquisition method based on privacy protection according to claim 7, characterized in that: In step S19, the model holder obtains the required data according to the response result, which is expressed as: (q,st)←PIR.query(pp D ,ID) r←PIR.respond(pp D ,D,q) d←PIR.process(pp D ,st,r) Among them, d represents the required data; process represents the processing algorithm; st represents the query status; r represents the response result; responde represents the response algorithm; D represents the database; q represents the query value; query represents the query algorithm; ID represents the ID of the element in the collection.

9. A privacy-preserving data acquisition method, characterized in that: include: S21. The data holder randomly slices the database and stores the second data obtained after random slicing; S22. The data holder receives the first data obtained by random slicing of the LR model from the model holder, and the data holder calculates the first data and the second data to obtain slice data of the second classification result; S23. The data holder processes the slice data of the second classification result based on the rule model to obtain a second processing result; S24. The data holder receives the first processing result obtained based on the rule model sent by the model holder, and the data holder receives the third processing result obtained by calculating the first processing result and the second processing result by the model holder; S25. The data holder encodes the database based on the hidden query protocol to obtain an encoded database, and sends the encoded database to the model holder; S26. The data holder receives the query value and query status generated by the model holder according to the third processing result and the encoding database; S27. The data holder generates a response result based on the query value and query status, and sends the corresponding result to the model holder. The model holder obtains the required data based on the response result.

10. A data acquisition device based on privacy protection, used to execute a data acquisition method based on privacy protection according to any one of claims 1 to 8, the data acquisition device comprising a processor, characterized in that: The processor is provided with: A slicing module, used by the model holder to randomly slice the LR model and store the first data obtained after random slicing; A first calculation module is used for the model holder to receive second data sent by the data holder after random slicing of the database, and the model holder calculates the first data and the second data to obtain slice data of the first classification result; A processing module, used by the model holder to process the slice data of the first classification result based on the rule model to obtain a first processing result; A second calculation module is used for the model holder to receive the second processing result obtained based on the rule model sent by the data holder, and calculate the first processing result and the second processing result to obtain a third processing result; A comparison module, used for the model holder to compare the third processing result with the rule symbol in the rule model, and if the third processing result is equal to the rule symbol, the ID corresponding to the third processing result is put into the set; An encoding module, used for the model holder to receive an encoding database sent by the data holder for encoding the database; The query module is used by the model holder to generate a query value and a query status for the encoded database according to the data in the collection and based on the hidden query protocol, and send the query value and the query status to the data holder; The corresponding module is used for the model holder to receive the response result generated by the data holder according to the query value and query status; The acquisition module is used by the model holder to obtain the required data based on the response results.