A method, device and computer program product for protecting privacy of sensitive data

Abstract

The present invention discloses a privacy protection method, device and computer program product for sensitive data. The present invention identifies sensitive information by calculating the sensitivity of each data block in the target data, and after extracting the sensitive data blocks in the target data, groups the sensitive data blocks according to the correlation between the sensitive data blocks, thereby obtaining a plurality of data block groups; then, the sensitivity level of each data block group is determined; finally, according to the sensitivity level of each data block group, each data block group can be desensitized in a manner corresponding to the sensitivity level; therefore, compared with the traditional technology, the present invention not only reduces the workload and improves the recognition efficiency, but also realizes the hierarchical desensitization of data of different sensitivity levels during desensitization, based on which computing resources can be saved and the amount of data processed can be reduced, and therefore, the present invention is very suitable for large-scale application and promotion.

Description

Technical Field

[0001] The present invention belongs to the field of data processing technology, and specifically relates to a privacy protection method, device and computer program product for sensitive data. Background Art

[0002] With the development of Internet technology, more and more data are processed through Internet technology. In the process of data interaction or storage, some sensitive information will inevitably be involved, that is, some important information that is not convenient for public disclosure, such as personal ID number, bank account, etc.; Therefore, how to accurately locate the sensitive information in the data, mark the sensitive information that needs to be protected, and protect the sensitive information is a problem that needs to be solved in the data processing link.

[0003] Currently, data privacy protection mostly involves first identifying sensitive information, then using a specified desensitizing algorithm to desensitize the sensitive information to obtain desensitized sensitive information; finally, the desensitized sensitive information and non-sensitive information are stored or transmitted to the recipient through a public channel, thereby completing data privacy protection during the transmission or storage process.

[0004] However, the above-mentioned existing technologies have the following deficiencies: (1) The identification of sensitive information mostly adopts manual methods, which not only increases the workload but also reduces the processing efficiency; (2) The existing technologies usually adopt the same method for desensitization processing. However, due to the different confidentiality and value of different data, their sensitivity levels are also different. Using the same method for unified desensitization not only wastes resources but also increases the amount of data processing. Based on this, how to provide a sensitive data privacy protection method that has a fast sensitive information identification efficiency and can perform graded desensitization on data of different sensitivity levels has become a problem that needs to be solved urgently. Summary of the invention

[0005] The purpose of the present invention is to provide a privacy protection method, device and computer program product for sensitive data, so as to solve the problems of slow sensitive information identification efficiency in the prior art, waste of resources and large data processing volume caused by unified desensitization using the same method.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions:

[0007] In a first aspect, a method for protecting privacy of sensitive data is provided, comprising:

[0008] Acquire target data, and process the target data in blocks to obtain multiple data blocks;

[0009] Calculate the mutual information between each data block, and calculate the sensitivity of each data block based on the mutual information between each data block, so as to identify a number of sensitive data blocks from multiple data blocks according to the sensitivity of each data block;

[0010] For any sensitive data block among the several sensitive data blocks, the correlation between the any sensitive data block and the corresponding target data blocks is calculated, and after all the sensitive data blocks are polled, the correlation between each sensitive data block and the corresponding target data blocks is obtained, wherein the target data blocks corresponding to any sensitive data block are all sensitive data blocks among the several sensitive data blocks except the any sensitive data block;

[0011] According to the correlation between each sensitive data block and the corresponding target data blocks, a plurality of sensitive data blocks are grouped and processed to obtain a plurality of data block groups after the grouping processing;

[0012] Calculate the sensitivity of each data block group, and based on the sensitivity of each data block group, obtain the sensitivity level of each data block group;

[0013] According to the sensitivity level of each data block group, the desensitization method of each data block group is determined, and the desensitization method corresponding to each data block group is used to desensitize the sensitive data blocks in each data block group, so as to complete the privacy protection of the target data after the desensitization process.

[0014] Based on the above disclosed content, after acquiring the target data, the present invention first divides the target data into blocks to obtain multiple data blocks, and then calculates the mutual information between each data block, and based on this, determines the sensitivity of each data block; then, according to the sensitivity of each data block, several sensitive data blocks can be identified from the multiple data blocks; in this way, the machine is used to identify sensitive information based on the sensitivity of the data, which not only reduces the workload but also improves the recognition efficiency compared with manual identification; after obtaining the sensitive information in the target data, the present invention calculates the correlation between each sensitive data block and each of the other sensitive data blocks, and based on the correlation, performs grouping processing of the sensitive information, thereby obtaining several data block groups; then, the sensitivity of each data block group is calculated, and based on this, the sensitivity level of each data block group is obtained; then, according to the sensitivity level of each data block group, the desensitization method of each data block group can be determined; finally, based on the desensitization method corresponding to the sensitivity level, each data block group is desensitized to complete the privacy protection of the target data.

[0015] Through the above design, before desensitizing, the present invention first calculates the mutual information between data blocks, and based on this, obtains the sensitivity of each data block; then, the sensitive information is identified according to the sensitivity, thereby obtaining a number of sensitive data blocks; then, according to the correlation between the data blocks, the sensitive data blocks are grouped and processed to obtain a number of data block groups; then, the sensitivity of each data block group is calculated, and the sensitivity level of each data block is determined based on this; finally, the desensitization method of each data block group is determined according to the sensitivity level, and the desensitization method corresponding to each data block is used for desensitization, so as to complete the privacy protection of the data; therefore, compared with the traditional technology, the present invention not only reduces the workload and improves the recognition efficiency, but also groups the data of different sensitivities during desensitization, and according to the sensitivity level of each data block group, the desensitization method corresponding to the sensitivity level is used for desensitization, so as to realize the hierarchical desensitization of data of different sensitivity levels, based on which computing resources can be saved and the amount of data processed can be reduced, therefore, it is very suitable for large-scale application and promotion in the field of data processing.

[0016] In a possible design, the sensitivity of each data block is calculated based on the mutual information between each data block, including:

[0017] For any data block among the multiple data blocks, determine a median value of the mutual information of the any data block based on the mutual information between the any data block and each designated data block, wherein each designated data block is all data blocks among the multiple data blocks except the any data block;

[0018] Calculating the entropy value of any data block and the discrete entropy between any data block and each designated data block, and selecting the maximum discrete entropy from each discrete entropy;

[0019] According to the median of the mutual information, the maximum discrete entropy and the entropy value of any data block, and in accordance with the following formula (1), the sensitivity of any data block is calculated;

[0020]

[0021] In the above formula (1), M i Indicates the sensitivity of any data block, L max represents the maximum discrete entropy, C i represents the entropy value of any data block, D i Express the median of the mutual information.

[0022] In one possible design, the sensitivity of each data block group is calculated, including:

[0023] For any data block group, determine whether the kth sensitive data block in the any data block group is an associated data block, wherein the associated data block is used to indicate that the kth sensitive data block has an associated relationship with at least one designated sensitive data block, and each designated sensitive data block is a sensitive data block in the any data block group except the kth sensitive data block;

[0024] If so, filtering out the designated sensitive data block most relevant to the k-th sensitive data block from the designated sensitive data blocks associated with the k-th sensitive data block;

[0025] Calculate the actual sensitivity of the k-th sensitive data block based on the screened designated sensitive data block and the k-th sensitive data block;

[0026] Add 1 to k, and re-determine whether the kth sensitive data block in any data block group is an associated data block, until k is equal to n, and obtain the actual sensitivity of each sensitive data block in any data block group, wherein the initial value of k is 1, and n is the total number of sensitive data blocks in any data block group;

[0027] The sensitivity of any data block group is determined according to the actual sensitivity of each sensitive data block in the any data block group.

[0028] In a possible design, based on the screened designated sensitive data block and the k-th sensitive data block, calculating the actual sensitivity of the k-th sensitive data block includes:

[0029] Calculating the conditional entropy between the kth sensitive data block and the screened designated sensitive data block;

[0030] Obtaining an entropy value of the k-th sensitive data block, and determining an association relationship entropy of the k-th sensitive data block based on the entropy value of the k-th sensitive data block and the conditional entropy;

[0031] Obtaining the mutual information between the kth sensitive data block and each designated sensitive data block, and the discrete entropy of each sensitive data block in any one of the data block groups;

[0032] The actual sensitivity of the kth sensitive data block is calculated based on the association relationship entropy, the mutual information between the kth sensitive data block and each designated sensitive data block, and the discrete entropy of each sensitive data block in any data block group.

[0033] In a possible design, the desensitization method of each data block group is determined according to the sensitivity level of each data block group, including:

[0034] Obtain a desensitization algorithm database, wherein the desensitization algorithm database stores desensitization methods corresponding to different sensitivity levels, and the desensitization method corresponding to any sensitivity level includes a primary desensitization method and a secondary desensitization method. The higher the sensitivity level, the higher the complexity of the primary desensitization method and the secondary desensitization method corresponding to any sensitivity level, and the primary desensitization method is to use an improved genetic algorithm for primary desensitization;

[0035] According to the sensitivity level of each data block group, a desensitization method corresponding to the sensitivity level of each data block group is matched from the desensitization algorithm database, so as to obtain the desensitization method of each data block group after the matching is completed;

[0036] Accordingly, the desensitization method corresponding to each data block group is adopted to desensitize the sensitive data blocks in each data block group, which includes:

[0037] For the j-th sensitive data block in any data block group, encoding the j-th sensitive data block to obtain encoded data corresponding to the j-th sensitive data block;

[0038] Based on the coded data, determining an initial population of the j-th sensitive data block, wherein the initial population includes a number of initial individuals, each initial individual represents a coded data block, and all initial individuals are concatenated to obtain the coded data;

[0039] Using the improved genetic algorithm in the desensitization method corresponding to any one of the data block groups, perform initial desensitization processing on each initial individual in the initial population, so as to obtain a plurality of desensitized individuals after the initial desensitization processing;

[0040] Based on several desensitized individuals, generate the jth sensitive data block after initial desensitization;

[0041] Add 1 to j and re-encode the jth sensitive data block until j is equal to S, and obtain S sensitive data blocks after initial desensitization, where the initial value of j is 1 and S is the total number of sensitive data blocks in any data block group;

[0042] The secondary desensitization method in the desensitization method corresponding to any one of the data block groups is adopted to perform secondary desensitization processing on the S sensitive data blocks after the initial desensitization processing, so as to complete the desensitization processing of the any one of the data block groups after the secondary desensitization processing.

[0043] In a possible design, the improved genetic algorithm in the desensitization method corresponding to any data block group is used to perform initial desensitization processing on each initial individual in the initial population, so as to obtain a number of desensitized individuals after the initial desensitization processing, including:

[0044] Initialize the number of iterations t to 1, and perform the t-th individual selection operation on the initial population to use the selected initial individuals to form a parent population;

[0045] Selecting multiple groups of parent individuals from the parent population, and calculating the crossover operator at the t-th iteration, wherein each group of parent individuals includes two parent individuals;

[0046] According to the crossover operator at the tth iteration, an individual crossover operation is performed on each group of parent individuals, so as to obtain multiple crossover individuals after the individual crossover operation;

[0047] Using a plurality of crossover individuals and parent individuals that are not selected from the parent population to form a child population;

[0048] The mutation probability at the t-th iteration is calculated, and based on the mutation probability at the t-th iteration, an individual mutation operation is performed on each offspring individual in the offspring population, so as to obtain a mutant individual corresponding to each offspring individual after the individual mutation operation;

[0049] Use the mutant individuals corresponding to each offspring individual to form the desensitized population after the tth iteration;

[0050] Calculate the fitness of the desensitized population after the tth iteration, and determine whether the fitness is greater than or equal to the fitness threshold, or whether t reaches the maximum number of iterations. The fitness threshold and maximum number of iterations corresponding to different sensitivity levels are different. The fitness is used to characterize the similarity between the desensitized population after the tth iteration and the initial population, and the greater the fitness, the smaller the similarity;

[0051] If not, the initial population is updated to the desensitized population after the tth iteration;

[0052] Increment t by 1 and re-perform the t-th individual selection operation on the initial population until the iteration stopping condition is met, so as to obtain the several desensitized individuals based on the desensitized population that meets the iteration stopping condition, wherein the iteration stopping condition is that the fitness of the desensitized population after the t-th iteration is greater than or equal to the fitness threshold or t reaches the maximum number of iterations.

[0053] In one possible design, the crossover operator at the tth iteration is calculated, including:

[0054] Generate a random number for the tth iteration, where the value interval of the random number is [0,1];

[0055] Based on the random number, the crossover operator at the t-th iteration is calculated using the following formula (2);

[0056]

[0057] In the above formula (2), δ t represents the crossover operator at the t-th iteration, μ represents the random number, a and b represent the first crossover adjustment factor and the second crossover adjustment factor respectively;

[0058] Correspondingly, according to the crossover operator at the tth iteration, the individual crossover operation is performed on each group of parent individuals, which includes:

[0059] For any group of parent individuals, based on the crossover operator at the t-th iteration, and using the following formula (3), an individual crossover operation is performed on the any group of parent individuals, so as to obtain crossover individuals corresponding to two parent individuals in the any group of parent individuals after the individual crossover operation;

[0060]

[0061] In the above formula (3), represents a parent individual in any group of parent individuals, represents another parent individual in any group of parent individuals, express The corresponding crossover individuals, express The corresponding crossover individuals.

[0062] In one possible design, the secondary desensitization method corresponding to any sensitivity level is to perform secondary desensitization using a desensitization model, wherein the desensitization model includes a generator and a discriminator, and the generator is used to perform secondary desensitization on the input data of the desensitization model, and the higher the sensitivity level, the greater the number of generators in the desensitization model corresponding to the sensitivity level.

[0063] In a second aspect, a privacy protection device for sensitive data is provided, comprising:

[0064] A block division unit, used for acquiring target data and performing block processing on the target data to obtain multiple data blocks;

[0065] A sensitive information identification unit, used to calculate the mutual information between each data block, and based on the mutual information between each data block, calculate the sensitivity of each data block, so as to identify a number of sensitive data blocks from multiple data blocks according to the sensitivity of each data block;

[0066] A grouping unit, for calculating, for any sensitive data block among the plurality of sensitive data blocks, the correlation between the any sensitive data block and the corresponding target data blocks, and obtaining the correlation between each sensitive data block and the corresponding target data blocks after all sensitive data blocks have been polled, wherein the target data blocks corresponding to any sensitive data block are all sensitive data blocks among the plurality of sensitive data blocks except the any sensitive data block;

[0067] The grouping unit is further used to group the sensitive data blocks according to the correlation between each sensitive data block and the corresponding target data blocks, so as to obtain a plurality of data block groups after the grouping process;

[0068] A grading unit, used to calculate the sensitivity of each data block group, and based on the sensitivity of each data block group, derive the sensitivity level of each data block group;

[0069] The desensitizing unit is used to determine the desensitizing method of each data block group according to the sensitivity level of each data block group, and adopt the desensitizing method corresponding to each data block group to desensitize the sensitive data blocks in each data block group, so as to complete the privacy protection of the target data after the desensitizing process.

[0070] According to a third aspect, another privacy protection device for sensitive data is provided. Taking the device as an electronic device as an example, the device includes a memory, a processor and a transceiver which are communicatively connected in sequence, wherein the memory is used to store computer programs, the transceiver is used to send and receive messages, and the processor is used to read the computer program to execute the privacy protection method for sensitive data as in the first aspect or any one of the possible designs in the first aspect.

[0071] In a fourth aspect, a storage medium is provided, on which instructions are stored. When the instructions are executed on a computer, the privacy protection method for sensitive data as in the first aspect or any possible design of the first aspect is executed.

[0072] In a fifth aspect, a computer program product comprising instructions is provided, which, when executed on a computer, causes the computer to execute the privacy protection method for sensitive data as described in the first aspect or any possible design of the first aspect.

[0073] Beneficial effects:

[0074] (1) Compared with the traditional technology, the present invention not only reduces the workload and improves the recognition efficiency, but also groups the data of different sensitivities during desensitization, and performs desensitization processing in a manner corresponding to the sensitivity level according to the sensitivity level of each data block group. In this way, hierarchical desensitization of data of different sensitivity levels is achieved. Based on this, computing resources can be saved and the amount of data to be processed can be reduced. Therefore, the present invention is very suitable for large-scale application and promotion in the field of data processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] Figure 1 A schematic diagram of the steps of the sensitive data privacy protection method provided by an embodiment of the present invention;

[0076] Figure 2 A schematic diagram of the structure of a privacy protection device for sensitive data provided by an embodiment of the present invention;

[0077] Figure 3 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0078] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the present invention will be briefly introduced below in combination with the drawings and the description of the embodiments or the prior art. Obviously, the following description of the structure of the drawings is only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. It should be noted that the description of these embodiments is used to help understand the present invention, but does not constitute a limitation of the present invention.

[0079] It should be understood that although the terms first, second, etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are only used to distinguish one unit from another unit. For example, a first unit can be referred to as a second unit, and similarly, a second unit can be referred to as a first unit without departing from the scope of the exemplary embodiments of the present invention.

[0080] It should be understood that the term "and / or" that may appear in this article is merely a description of the association relationship between associated objects, indicating that three relationships may exist. For example, A and / or B can represent three situations: A exists alone, B exists alone, and A and B exist at the same time. The term " / and" that may appear in this article describes another type of association object relationship, indicating that two relationships may exist. For example, A / and B can represent two situations: A exists alone, and A and B exist alone. In addition, the character " / " that may appear in this article generally indicates that the previous and next associated objects are in an "or" relationship.

[0081] Example:

[0082] See also Figure 1 As shown, the privacy protection method for sensitive data provided in this embodiment identifies sensitive information by calculating the sensitivity of each data block in the target data, and after extracting the sensitive data blocks in the target data, the sensitive data blocks are grouped according to the correlation between the sensitive data blocks, thereby obtaining a number of data block groups; then, the sensitivity level of each data block group is determined; finally, each data block group can be desensitized in a manner corresponding to the sensitivity level according to the sensitivity level of each data block group; thus, compared with the traditional technology, this method not only reduces the workload and improves the recognition efficiency, but also realizes the hierarchical desensitization of data of different sensitivity levels during desensitization, based on which computing resources can be saved and the amount of data processed can be reduced, therefore, it is very suitable for large-scale application and promotion in the field of data processing; wherein, for example, this method can be, but not limited to, run on the privacy protection end side, optionally, the privacy protection end can be, but not limited to, a personal computer (PC), a tablet computer or a smart phone, it can be understood that the aforementioned execution subject does not constitute a limitation on the embodiments of the present application, and accordingly, the running steps of this method can be, but not limited to, as shown in the following steps S1 to S6.

[0083] S1. Obtain target data, and process the target data in blocks to obtain multiple data blocks; in this embodiment, the target data may be, but is not limited to, data to be stored or data to be transmitted, and processing the target data in blocks means performing word segmentation processing on the target data to obtain each word block in the target data; of course, word segmentation processing of data can be implemented using NLP (natural language processing) method, which is a commonly used method for data processing, and its principle will not be repeated here.

[0084] After completing the block processing of the target data, it may be necessary to identify sensitive information; wherein, this embodiment identifies sensitive information by calculating the sensitivity of each data block; optionally, the sensitive information identification process may be but is not limited to as shown in the following step S2.

[0085] S2. Calculate the mutual information between each data block, and calculate the sensitivity of each data block based on the mutual information between each data block, so as to identify several sensitive data blocks from multiple data blocks according to the sensitivity of each data block; in the specific implementation, for any data block, the mutual information between the any data block and each of the remaining data blocks is calculated, so that the sensitivity of each data block can be determined based on the mutual information between each data block and the remaining data blocks.

[0086] In actual application, assuming that any data block is data block 1, the mutual information between it and data block 2 among multiple sensitive data blocks is calculated as follows: obtain the prior probability of data block 1 and the probability of data block 1 under the condition that data block 2 exists (i.e., the posterior probability); then, use the formula To calculate the mutual information between data block 1 and data block 2; in the above formula, Then they represent the posterior probability and the prior probability respectively; of course, the calculation method of mutual information is a common technique in information theory, and its principle will not be repeated here.

[0087] After the mutual information between the data blocks is obtained, the sensitivity can be calculated. Here, any data block is taken as an example for specific explanation. The sensitivity calculation process can be but is not limited to the following steps S21 to S23.

[0088] S21. For any data block among the multiple data blocks, based on the mutual information between the any data block and each designated data block, determine the median of the mutual information of the any data block, wherein each designated data block is all data blocks among the multiple data blocks except the any data block; in this embodiment, it is assumed that the multiple data blocks are data block 1, data block 2, data block 3 and data block 4, and if any of the aforementioned data blocks is data block 1, then the mutual information between data block 1 and data block 2, the mutual information between data block 1 and data block 3, and the mutual information between data block 1 and data block 4 are first sorted in descending order to obtain a sorted sequence; then, the median of the sorted sequence is taken as the median of the mutual information; in this way, after obtaining the median of the mutual information, the next step can be entered, as shown in the following step S22.

[0089] S22. Calculate the entropy value of any data block and the discrete entropy between any data block and each designated data block, and select the largest discrete entropy from each discrete entropy; in specific application, the entropy value of any data block specifically refers to information entropy, and the corresponding information entropy and discrete entropy calculation method is a common technology in information theory, and its principle is not repeated; wherein, after calculating the discrete entropy between any data block and each designated data block, the largest discrete entropy can be selected to calculate the sensitivity; wherein, the sensitivity calculation process can be but is not limited to as shown in the following step S23.

[0090] S23. Calculate the sensitivity of any data block according to the median of the mutual information, the maximum discrete entropy and the entropy value of any data block and in accordance with the following formula (1).

[0091]

[0092] In the above formula (1), Mi Indicates the sensitivity of any data block, L max represents the maximum discrete entropy, C i represents the entropy value of any data block, D i Express the median of the mutual information; in this embodiment, the smaller the sensitivity, the higher the sensitivity of the data block, and conversely, the greater the sensitivity, the less sensitive the data block.

[0093] In this way, based on the aforementioned steps S21 to S23, the sensitivity of each data block can be calculated; and after obtaining the sensitivity of each data block, sensitive information can be identified based on the sensitivity; wherein, the sensitive information identification process can be but is not limited to as shown in the following steps S24 to S28.

[0094] S24. Use data blocks with a sensitivity greater than or equal to the sensitivity threshold to form a suspected sensitive data set, and use data blocks with a sensitivity less than the sensitivity threshold to form an initial sensitive data set; in this embodiment, the sensitivity threshold can be set to 0.6, but is not limited to it; of course, it can be set specifically according to actual use, and is not limited to the aforementioned example.

[0095] S25. Perform association rule analysis on the initial sensitive data set and the suspected sensitive data set to obtain the association relationship of each suspected sensitive data block in the suspected sensitive data set; wherein, the association relationship of any suspected sensitive data block is empty, or has an association relationship with at least one initial sensitive data block; in specific applications, for example, but not limited to, April algorithm (i.e., association rule mining algorithm) can be used to perform association rule analysis on the initial sensitive data set and the suspected sensitive data set.

[0096] After obtaining the association relationship of each suspected sensitive data block, the suspected sensitive data set can be filtered based on the association relationship, and the process can be but not limited to the following step S26.

[0097] S26. For any suspected sensitive data block in the initial sensitive data set, based on the association relationship of any suspected sensitive data block, determine whether the number of initial sensitive data blocks that have an association relationship with any suspected sensitive data block is greater than or equal to a quantity threshold; in this embodiment, the quantity threshold can be but is not limited to one third of the total number of suspected sensitive data blocks.

[0098] S27. If yes, any suspected sensitive data block is taken as a sensitive data block, and after all suspected sensitive data blocks are polled, the filtering process of the suspected sensitive data set is completed to obtain the filtering result.

[0099] After completing the filtering process of the suspected sensitive data set and filtering out sensitive data blocks from the suspected sensitive data set, the filtered sensitive data blocks and the initial sensitive data set can be used to obtain several sensitive data blocks. The process can be but is not limited to as shown in the following step S28.

[0100] S28. Based on the filtering result and the initial sensitive data set, several sensitive data blocks are determined; in this embodiment, if the filtering result is empty, then the initial sensitive data block in the initial sensitive data set is used as the final sensitive data block, and if the filtering result is not empty, that is, a sensitive data block is filtered out from the suspected sensitive data set, then the sensitive data block obtained by filtering and the initial sensitive data block in the initial sensitive data set are both used as the final sensitive data block; in this way, several sensitive data blocks can be identified from multiple data blocks.

[0101] Therefore, through the aforementioned steps S21 to S28, the identification of sensitive information in the target data can be completed on a machine basis, thereby obtaining a number of sensitive data blocks; then, it is necessary to determine the sensitivity level of each sensitive data block so as to perform graded desensitization processing based on the sensitivity level; wherein, the sensitivity level determination process can be but is not limited to as shown in the following steps S3 to S5.

[0102] S3. For any sensitive data block among the several sensitive data blocks, the correlation between the any sensitive data block and the corresponding target data blocks is calculated, and after all the sensitive data blocks are polled, the correlation between each sensitive data block and the corresponding target data blocks is obtained, wherein the target data blocks corresponding to any sensitive data block are all sensitive data blocks among the several sensitive data blocks except the any sensitive data block; in specific implementation, the correlation between the aforementioned any sensitive data block and the corresponding target data blocks is also the mutual information; at the same time, assuming that there are 5 sensitive data blocks, then the mutual information between the any sensitive data block and the remaining 4 sensitive data blocks (that is, the remaining 4 sensitive data blocks are target data blocks) is calculated, and the mutual information is used as the correlation between the any sensitive data block and the remaining 4 sensitive data blocks; similarly, the calculation process of the correlation between the remaining sensitive data blocks and their corresponding target data blocks is the same, which will not be repeated here.

[0103] After the correlation between each sensitive data block and each corresponding target data block is obtained, grouping processing can be performed based on this, and the process can be but is not limited to the following step S4.

[0104] S4. According to the correlation between each sensitive data block and the corresponding target data blocks, several sensitive data blocks are grouped and processed to obtain several data block groups after the grouping processing; in the present embodiment, the sensitive data blocks are directly divided according to the correlation; specifically, for any sensitive data block, the sensitive data blocks whose correlation with the any sensitive data block is greater than or equal to the correlation threshold are grouped into one group; then, the grouped sensitive data blocks are removed, and the aforementioned process is repeated until all sensitive data blocks are divided, at which time, several data block groups can be obtained.

[0105] The following is an example to illustrate:

[0106] Assume that the sensitive data blocks are: sensitive data blocks 1-5, among which, the correlation between sensitive data block 1 and sensitive data blocks 2-5 is; U1, U2, U3, U4, among which U1 and U2 are greater than the correlation threshold, then, sensitive data block 1, sensitive data block 2 and sensitive data block 3 are divided into a group; then, it is necessary to remove sensitive data blocks 1-3 from several sensitive data blocks, and then, take another sensitive data block to re-execute the above process until all sensitive data blocks are divided.

[0107] Furthermore, in order to improve the accuracy of grouping, this embodiment also provides a grouping method, namely:

[0108] The sensitive data blocks are grouped according to the correlation and the association relationship between them (similarly, the April algorithm can also be used to obtain the result), that is, the data blocks with high correlation and association relationship are grouped together; specifically, for any sensitive data block, based on the correlation between it and each target data block, determine whether any sensitive data block has a cross-association; if so, group the target data block and the target data block with which it has a cross-association into a group; otherwise, group the target data block with the highest correlation with any sensitive data block into a group, and also group the sensitive data block with which it has an association relationship with the target data block with the highest correlation into the group; remove the aforementioned sensitive data blocks that have been divided from a number of sensitive data blocks; then, re-execute the aforementioned steps until all sensitive data blocks are divided.

[0109] Specifically, the above step S4 is described with an example:

[0110] Assume that the sensitive data blocks are: sensitive data blocks 1-5, among which the correlation between sensitive data block 1 and sensitive data blocks 2-5 is; U1, U2, U3, U4, and U1>U2>U3>U4, among which U3 and U4 are 0 (indicating that they are not correlated with sensitive data blocks 4 and 5), and sensitive data block 2 is correlated with sensitive data block 3 (that is, the mutual information between the two is greater than 0), then, sensitive data blocks 1, 2 and 3 are cross-correlated. At this time, sensitive data blocks 1-3 can be divided into 1 group; then, sensitive data blocks 1-3 are removed from several sensitive data blocks, and another sensitive data block is taken to re-execute the above process until all sensitive data blocks are divided.

[0111] At the same time, if U1>U2>U3, and sensitive data block 2 is associated with sensitive data block 4, then sensitive data block 1 and sensitive data block 2 are divided into a group, and sensitive data block 4 is also divided into the same group; at this time, sensitive data blocks 1, 2 and 4 are removed from several sensitive data blocks; then, another sensitive data block is taken to re-execute the above process until all sensitive data blocks are divided.

[0112] In this way, the grouping of sensitive data blocks can be completed through the above-mentioned method, thereby obtaining a number of data block groups; and after obtaining the number of data block groups, it is necessary to determine the sensitivity level of each data block group, so as to perform graded desensitization based on the sensitivity level later; wherein, the process of determining the sensitivity level of each data block group can be but is not limited to as shown in the following step S5.

[0113] S5. Calculate the sensitivity of each data block group, and based on the sensitivity of each data block group, obtain the sensitivity level of each data block group; in the specific implementation, the following will be explained using any data block group as an example, and the sensitivity calculation process can be but is not limited to the following steps S51 to S55.

[0114] S51. For any data block group, determine whether the kth sensitive data block in the any data block group is an associated data block, wherein the associated data block is used to characterize that the kth sensitive data block has an associated relationship with at least one designated sensitive data block, and each designated sensitive data block is a sensitive data block other than the kth sensitive data block in the any data block group; in this embodiment, the April algorithm is still used to determine whether there is a sensitive data block associated with the kth sensitive data block in the any data block group; wherein, if so, the kth sensitive data block is an associated data block, otherwise it is not; optionally, when the kth sensitive data block is an associated data block, it is necessary to calculate the actual sensitivity of the kth sensitive data block in combination with the designated sensitive data block associated with it, and the process may be but is not limited to the following steps S52 and S53.

[0115] S52. If yes, then select the designated sensitive data block that is most relevant to the k-th sensitive data block from the designated sensitive data blocks that are associated with the k-th sensitive data block; in specific implementation, the strength of the association relationship (i.e., the association relationship between the k-th sensitive data block and each designated sensitive data block) is measured by support and confidence (which has been obtained when the April algorithm is used to construct the association relationship), then, a weighted method can be used to obtain the strength of the designated sensitive data block that is associated with the k-th sensitive data block; if the k-th sensitive data block is associated with two designated sensitive data blocks, then, based on the support and confidence of the two associations, the strength of the two associations is weightedly calculated (i.e., the support and confidence are multiplied by their respective weights and then summed), and then, the designated sensitive data block with the greatest strength is used as the designated sensitive data block that is most relevant to the k-th sensitive data block; of course, if the number of designated sensitive data blocks that are associated with the k-th sensitive data block is 1, it is directly used as the most relevant designated sensitive data block.

[0116] In this way, after obtaining the designated sensitive data block most relevant to the k-th sensitive data block, the actual sensitivity of the k-th sensitive data block can be calculated based on this, and the process can be but is not limited to the following step S53.

[0117] S53. Based on the screened designated sensitive data block and the k-th sensitive data block, the actual sensitivity of the k-th sensitive data block is calculated; in specific implementation, the conditional entropy between the k-th sensitive data block and the screened designated sensitive data block is first calculated; then, the entropy value of the k-th sensitive data block is obtained, and based on the entropy value of the k-th sensitive data block and the conditional entropy, the association relationship entropy of the k-th sensitive data block is determined; then, the mutual information between the k-th sensitive data block and each designated sensitive data block, and the discrete entropy of each sensitive data block in any one data block group are obtained; finally, the actual sensitivity of the k-th sensitive data block can be calculated based on the association relationship entropy, the mutual information between the k-th sensitive data block and each designated sensitive data block, and the discrete entropy of each sensitive data block in any one data block group.

[0118] Optionally, conditional entropy is also a commonly used measurement method in information theory, and its calculation principle is not repeated here. At the same time, the entropy value of the kth sensitive data block and the aforementioned conditional entropy can be summed to obtain the association entropy of the kth sensitive data block. Then, based on the mutual information between the kth sensitive data block and each designated sensitive data block, the median of the mutual information is obtained, and the maximum discrete entropy is determined from the discrete entropies of each sensitive data block. Finally, the aforementioned formula (1) is used to calculate the actual sensitivity of the kth sensitive data block. Of course, when calculating, it is only necessary to replace the entropy value in formula (1) with the aforementioned association entropy.

[0119] Meanwhile, if the kth sensitive data block is not an associated data block, then the entropy value, the corresponding median of the mutual information and the maximum discrete entropy are directly used to calculate the actual sensitivity (still using the aforementioned formula (1)).

[0120] Thus, through the aforementioned steps S51 to S53, the actual sensitivity of the kth sensitive data block can be obtained; then, using the same principle, the actual sensitivity of the remaining sensitive data blocks can be obtained; wherein, the cyclic calculation process can be but is not limited to as shown in the following step S54.

[0121] S54. Add 1 to k and re-judge whether the kth sensitive data block in any data block group is an associated data block, until k is equal to n, and the actual sensitivity of each sensitive data block in any data block group is obtained, wherein the initial value of k is 1, and n is the total number of sensitive data blocks in any data block group.

[0122] After the actual sensitivity of each sensitive data block in any data block group is calculated based on the aforementioned steps S51 to S54, the sensitivity of any data block group can be derived based on this, and the process can be but is not limited to as shown in the following step S55.

[0123] S55. Determine the sensitivity of any data block group based on the actual sensitivity of each sensitive data block in the any data block group; in this embodiment, the average of the actual sensitivities of all sensitive data blocks in the any data block group is used as the sensitivity of the any data block group.

[0124] Therefore, through the aforementioned steps S51 to S54, the sensitivity of each data block group can be calculated, and then sorted in order of sensitivity from low to high, so as to determine the sensitivity level of each data block group based on the sorting result, wherein the sensitivity level of the data block group ranked first is the highest, which may be sensitivity level one, and the sensitivity level of the data block group ranked second is lower than the sensitivity level of the data block group ranked first, which is sensitivity level two, and so on, to determine the sensitivity level of each data block group.

[0125] After determining the sensitivity level of each data block group, hierarchical desensitization processing may be performed based on the sensitivity level, and the processing process may be, but is not limited to, as shown in the following step S6.

[0126] S6. According to the sensitivity level of each data block group, determine the desensitization method of each data block group, and use the desensitization method corresponding to each data block group to desensitize the sensitive data blocks in each data block group, so as to complete the privacy protection of the target data after the desensitization. In a specific implementation, for example, a desensitization algorithm database is pre-stored in the privacy protection end, wherein the desensitization algorithm database stores desensitization methods corresponding to different sensitivity levels, and the desensitization method corresponding to any sensitivity level includes a primary desensitization method and a secondary desensitization method. The higher the sensitivity level, the higher the complexity of the primary desensitization method and the secondary desensitization method corresponding to any sensitivity level, and the primary desensitization method is to use an improved genetic algorithm for primary desensitization. Further, in this embodiment, the fitness threshold and the maximum number of iterations of the improved genetic algorithm corresponding to different sensitivity levels are different. In this way, the higher the sensitivity level, the larger the fitness threshold and the maximum number of iterations of the corresponding algorithm, and the higher the required desensitization degree.

[0127] Based on this, when matching the desensitization method, the desensitization algorithm database is first obtained, and then, according to the sensitivity level of each data block group, the desensitization method corresponding to the sensitivity level of each data block group is matched from the desensitization algorithm database, so that the desensitization method of each data block group is obtained after the matching is completed.

[0128] After obtaining the desensitization method corresponding to each data block group based on the sensitivity level of each data block group, desensitization processing can be performed on each data block group; wherein, any data block group is taken as an example for specific explanation below, and its desensitization process can be but is not limited to as shown in the following steps S61 to S66.

[0129] S61. For the j-th sensitive data block in any data block group, the j-th sensitive data block is encoded to obtain the encoded data corresponding to the j-th sensitive data block; in this embodiment, the j-th sensitive data block can be encoded by but not limited to a real number encoding method to obtain the corresponding encoded data; after the encoding process is completed, an initial population of the j-th sensitive data block can be generated, and the process can be but not limited to as shown in the following step S62.

[0130] S62. Based on the coded data, determine the initial population of the j-th sensitive data block, wherein the initial population includes a number of initial individuals, each initial individual represents a coded data block, and all the initial individuals are spliced ​​together to obtain the coded data; in specific applications, for example, but not limited to, dividing the coded data according to a preset length to obtain multiple coded data blocks, and then, each coded data block can be used as an initial individual; of course, if the last length is less than the preset length, 0 is used to supplement it to the preset length; in this way, after obtaining the initial population of the j-th sensitive data block, the improved genetic algorithm corresponding to the sensitivity level of any sensitive data block can be used to perform the initial desensitization processing of the initial population, and the process can be but not limited to as shown in the following step S63.

[0131] S63. Perform initial desensitization processing on each initial individual in the initial population by utilizing the improved genetic algorithm in the desensitization method corresponding to any data block group, so as to obtain a number of desensitized individuals after the initial desensitization processing; in this embodiment, the genetic desensitization process continuously performs individual selection, crossover and mutation operations until the iteration stop condition is reached; wherein, the genetic desensitization process can be but is not limited to the following steps S631 to S639.

[0132] S631. Initialize the number of iterations t to 1, and perform the t-th individual selection operation on the initial population to use the selected initial individuals to form a parent population; in this embodiment, for example, it can be but not limited to pre-setting the selection probability, and then selecting multiple initial individuals from the initial population based on the selection probability; at the same time, to ensure that the number of the parent population is the same as the number of individuals in the initial population, if the number of selected individuals is less than the total number of initial individuals, then an individual replication operation is performed to expand the selected initial individuals to the same scale as the initial population; based on this, the selected initial individuals, or the selected initial individuals and the obtained replicated individuals, can be used to form the parent population.

[0133] After obtaining the parent population, individual crossover operations can be performed. In order to ensure the diversity of the population, this embodiment improves the traditional algorithm, that is, the crossover operator is calculated for each iteration, and based on this, individual crossover operations are performed; the aforementioned process can be but is not limited to the following steps S632 and S633.

[0134] S632. Select multiple groups of parent individuals from the parent population, and calculate the crossover operator at the t-th iteration, wherein each group of parent individuals includes two parent individuals. In specific implementation, the crossover selection probability can be pre-set, and then multiple groups of parent individuals are selected based on this. At the same time, the calculation process of the crossover operator at the t-th iteration is taken as an example: (1) generate a random number at the t-th iteration, wherein the value interval of the random number is [0,1]; (2) based on the random number, calculate the crossover operator at the t-th iteration.

[0135] In a specific application, for example, but not limited to, the following formula (2) may be used to generate a crossover operator at the t-th iteration.

[0136]

[0137] In the above formula (2), δ t represents the crossover operator at the t-th iteration, μ represents the random number, a, b represent the first crossover adjustment factor and the second crossover adjustment factor respectively; in the present embodiment, a∈R, b>0, R represents a real number set; optionally, for example, but not limited to, obtaining the expectation and variance of the distribution of each parent individual in the parent population at the t-th iteration, and then taking the expectation as the first crossover adjustment factor, and dividing the variance by 2, and taking the square root of the result to obtain the second crossover adjustment factor; in this way, the crossover operator in each iteration process is related to the individual distribution in the parent population in the iteration process; based on this, the diversity of the population after the crossover can be adjusted according to the above two crossover adjustment factors.

[0138] After the crossover operator at the t-th iteration is calculated based on the above formula (2), an individual crossover operation can be performed, and the process can be but is not limited to the following step S633.

[0139] S633. According to the crossover operator at the t-th iteration, perform an individual crossover operation on each group of parent individuals, so as to obtain a plurality of crossover individuals after the individual crossover operation. In this embodiment, the individual crossover operation is described using any group of parent individuals, and the process may be but is not limited to: for any group of parent individuals, based on the crossover operator at the t-th iteration, and using the following formula (3), perform an individual crossover operation on any group of parent individuals, so as to obtain crossover individuals corresponding to two parent individuals in any group of parent individuals after the individual crossover operation.

[0140]

[0141] In the above formula (3), represents a parent individual in any group of parent individuals, represents another parent individual in any group of parent individuals, express The corresponding crossover individuals, express The corresponding crossover individuals.

[0142] Thus, through the aforementioned formula (3), the crossover operation of two parent individuals in each group of parent individuals can be completed; then, multiple crossover individuals and unselected parent individuals can be used to form a child population, and the process can be but is not limited to the following step S634.

[0143] S634. Use multiple crossover individuals and parent individuals that are not selected from the parent population to form a child population; after completing the individual crossover operation and obtaining the child population, an individual mutation operation can be performed, and the process can be but is not limited to the following step S635.

[0144] S635. Calculate the mutation probability at the t-th iteration, and based on the mutation probability at the t-th iteration, perform an individual mutation operation on each offspring individual in the offspring population, so as to obtain a mutant individual corresponding to each offspring individual after the individual mutation operation; in this embodiment, the following formula (4) can be used as an example but is not limited to calculate the mutation probability at the t-th iteration.

[0145]

[0146] In the above formula (4), τ t represents the mutation probability at the tth iteration, ξ represents the mutation adjustment factor, ξ∈(0,1), T represents the maximum number of iterations, τ 0 Represents the initial mutation probability, which is the preset value.

[0147] Thus, based on the above formula (4), the mutation probability at the tth iteration can be calculated; and from the above formula (4), it can be seen that the mutation probability in the initial stage is high, and the solution space can be explored quickly; and as the number of iterations increases, the mutation probability gradually decreases, based on which the optimal solution can be searched more accurately.

[0148] After obtaining the mutation probability at the tth iteration, the mutation operation of each offspring individual can be performed based on the mutation probability. For example, if the mutation probability is 0.3, then each offspring individual has a 30% chance of mutation. Furthermore, the specific process of mutation of any offspring individual disclosed below can be but is not limited to the following steps.

[0149] Step 1: Obtain the variation position of any offspring individual and obtain the code at the variation position.

[0150] Step 2: Move the code at the mutation position to the remaining positions on any of the offspring individuals in sequence to obtain multiple initial mutation individuals.

[0151] Step 3: Calculate the similarity between each initial variant individual and any of the offspring individuals, and use the initial variant individual with the smallest similarity as the variant individual corresponding to any of the offspring individuals.

[0152] Thus, through the aforementioned steps, the mutation operation of each offspring individual can be completed, and then, each mutated individual can be used to form a desensitized population after the tth iteration, and the process can be but not limited to as shown in the following step S636.

[0153] S636. Use the mutant individuals corresponding to each offspring individual to form the desensitized population after the tth iteration.

[0154] After obtaining the desensitized population after the tth iteration, its corresponding fitness can be calculated so as to determine whether to stop the iteration based on the fitness; wherein, the fitness calculation and judgment process can be but is not limited to as shown in the following step S637.

[0155] S637. Calculate the fitness of the desensitized population after the t-th iteration, and determine whether the fitness is greater than or equal to the fitness threshold, or whether t reaches the maximum number of iterations; wherein, the fitness threshold and the maximum number of iterations corresponding to different sensitivity levels are different, and the fitness is used to characterize the similarity between the desensitized population after the t-th iteration and the initial population, and the greater the fitness, the smaller the similarity; in this embodiment, the essence is to use the coded data blocks corresponding to each individual in the desensitized population after the t-th iteration to form the j-th sensitive data block after the t-th desensitization; then, calculate the similarity between the j-th sensitive data block after the t-th desensitization and the j-th sensitive data block (the similarity can be obtained through the vector); finally, use 1 minus the similarity to obtain the fitness; in this way, the greater the fitness, the smaller the similarity.

[0156] After obtaining the fitness, it can be determined whether the fitness is greater than or equal to the fitness threshold (which can be set in advance), or whether t reaches the maximum number of iterations; if any of the above conditions is not met, it is necessary to continue iterating, and the process can be but is not limited to as shown in the following steps S638 and S639.

[0157] S638. If not, update the initial population to the desensitized population after the tth iteration.

[0158] S639. Add 1 to t and re-perform the t-th individual selection operation on the initial population until the iteration stopping condition is met, so as to obtain the several desensitized individuals based on the desensitized population that meets the iteration stopping condition, wherein the iteration stopping condition is that the fitness of the desensitized population after the t-th iteration is greater than or equal to the fitness threshold or t reaches the maximum number of iterations.

[0159] In this way, through the aforementioned steps S631 to S639, the initial desensitization processing of the j-th sensitive data block can be completed, and then, based on a number of desensitized individuals, the j-th sensitive data block after initial desensitization can be generated; the process is shown in the following step S64.

[0160] S64. Based on several desensitized individuals, generate the j-th sensitive data block after initial desensitization; after completing the initial desensitization processing of the j-th sensitive data block, the initial desensitization processing of the remaining sensitive data blocks can be completed according to the same principle; wherein, the cyclic desensitization process can be but is not limited to as shown in the following step S65.

[0161] S65. Add 1 to j and re-encode the j-th sensitive data block until j is equal to S, thereby obtaining S initially desensitized sensitive data blocks, wherein the initial value of j is 1, and S is the total number of sensitive data blocks in any data block group; after completing the initial desensitization processing of any data block group based on the aforementioned steps S61 to S66, a secondary desensitization processing may be performed, and the process may be but is not limited to as shown in the following step S66.

[0162] S66. Use the secondary desensitization method in the desensitization method corresponding to any one of the data block groups to perform secondary desensitization on the S sensitive data blocks after the initial desensitization, so as to complete the desensitization on any one of the data block groups after the secondary desensitization. In this embodiment, the secondary desensitization method corresponding to any one of the sensitivity levels is to use a desensitization model for secondary desensitization, wherein the desensitization model includes a generator and a discriminator, and the generator is used to perform secondary desensitization on the input data of the desensitization model (i.e., generate false data according to the sensitive data and mix it with the real data), and the higher the sensitivity level is, the more generators there are in the desensitization model corresponding to the sensitivity level.

[0163] In this way, it is equivalent to inputting each sensitive data block after initial desensitization in any data block group into the desensitization model corresponding to the sensitivity level of any data block group for secondary desensitization processing, so as to obtain any data block group after desensitization; wherein, for example, the number of generators in the desensitization model corresponding to the lowest sensitivity level is 1, and the data volume of the generator increases by 1 for each additional sensitivity level. In this way, for the data block group with the lowest sensitivity level, each sensitive data block after initial desensitization therein only passes through one generator to generate the data after secondary desensitization, and for each additional sensitivity level, the number of generators passed by the corresponding sensitive data block after initial desensitization increases by 1 (that is, it passes through two, three or four generators to generate the data after secondary desensitization); based on this, the more sensitive the data, the more generators it passes through, the more times the secondary desensitization is performed, and the stronger the security protection capability is.

[0164] Of course, in this embodiment, the desensitization model is pre-trained, that is, after the training samples are processed by the above-mentioned generator, the desensitized data corresponding to each training sample is obtained, and then the discriminator is used for discrimination (that is, the error between the actual output and the expected output is calculated and then the discrimination probability is calculated); then, it is judged whether the discrimination probability meets the requirements; if the requirements are not met, the model parameters are adjusted and retrained until the requirements are met.

[0165] Based on the aforementioned steps S61 to S66, data block groups with different sensitivity levels can be desensitized using desensitization methods corresponding to the sensitivity levels; and this embodiment adopts a double desensitization method, which improves the desensitization effect compared to the single desensitization of traditional technology, thereby further ensuring the security of the data.

[0166] Therefore, through the privacy protection method of sensitive data described in detail in the aforementioned steps S1 to S6, the present invention identifies sensitive information by calculating the sensitivity of each data block in the target data, and after extracting the sensitive data blocks in the target data, the sensitive data blocks are grouped according to the correlation between the sensitive data blocks, thereby obtaining a number of data block groups; then, the sensitivity level of each data block group is determined; finally, each data block group can be desensitized according to the sensitivity level of each data block group in a manner corresponding to the sensitivity level; therefore, compared with traditional technologies, the present invention not only reduces the workload and improves the recognition efficiency, but also realizes hierarchical desensitization of data of different sensitivity levels during desensitization, based on which computing resources can be saved and the amount of data processed can be reduced, and therefore, it is very suitable for large-scale application and promotion.

[0167] like Figure 2 As shown, the second aspect of this embodiment provides a hardware device for implementing the privacy protection method for sensitive data described in the first aspect of the embodiment, including:

[0168] The block division unit is used to obtain target data and perform block processing on the target data to obtain multiple data blocks.

[0169] The sensitive information identification unit is used to calculate the mutual information between each data block, and based on the mutual information between each data block, calculate the sensitivity of each data block, so as to identify several sensitive data blocks from multiple data blocks according to the sensitivity of each data block.

[0170] A grouping unit is used to calculate the correlation between any sensitive data block among several sensitive data blocks and the corresponding target data blocks, and obtain the correlation between each sensitive data block and the corresponding target data blocks after all sensitive data blocks are polled, wherein the target data blocks corresponding to any sensitive data block are all sensitive data blocks among the several sensitive data blocks except the any sensitive data block.

[0171] The grouping unit is also used to group the sensitive data blocks according to the correlation between each sensitive data block and the corresponding target data blocks, so as to obtain a plurality of data block groups after the grouping process.

[0172] The grading unit is used to calculate the sensitivity of each data block group and obtain the sensitivity level of each data block group based on the sensitivity of each data block group.

[0173] The desensitizing unit is used to determine the desensitizing method of each data block group according to the sensitivity level of each data block group, and adopt the desensitizing method corresponding to each data block group to desensitize the sensitive data blocks in each data block group, so as to complete the privacy protection of the target data after the desensitizing process.

[0174] The working process, working details and technical effects of the device provided in this embodiment can be found in the first aspect of the embodiment and will not be described in detail here.

[0175] like Figure 3 As shown, the third aspect of this embodiment provides another privacy protection device for sensitive data. Taking the device as an electronic device as an example, it includes: a memory, a processor and a transceiver that are communicatively connected in sequence, wherein the memory is used to store computer programs, the transceiver is used to send and receive messages, and the processor is used to read the computer program to execute the privacy protection method for sensitive data as described in the first aspect of the embodiment.

[0176] For example, the memory may include, but is not limited to, random access memory (RAM), read-only memory (ROM), flash memory, first-in-first-out memory (FIFO) and / or first-in-last-out memory (FILO), etc. Specifically, the processor may include one or more processing cores, such as a 4-core processor, an 8-core processor, etc. The processor may be implemented in at least one hardware form of DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), and PLA (Programmable Logic Array). At the same time, the processor may also include a main processor and a coprocessor. The main processor is a processor for processing data in the awake state, also known as a CPU (Central Processing Unit); the coprocessor is a low-power processor for processing data in the standby state.

[0177] In some embodiments, the processor may be integrated with a GPU (Graphics Processing Unit), which is responsible for rendering and drawing the content to be displayed on the display screen. For example, the processor may not be limited to a microprocessor of the STM32F105 series, a reduced instruction set computer (RISC) microprocessor, an X86 or other architecture processor, or a processor with an integrated embedded neural network processor (NPU); the transceiver may be, but is not limited to, a wireless fidelity (WIFI) wireless transceiver, a Bluetooth wireless transceiver, a general packet radio service technology (Genera l Packet Rad ioService, GPRS) wireless transceiver, a ZigBee protocol (a low-power LAN protocol based on the IEEE 802.15.4 standard, ZigBee) wireless transceiver, a 3G transceiver, a 4G transceiver and / or a 5G transceiver, etc. In addition, the device may also include, but is not limited to, a power module, a display screen and other necessary components.

[0178] The working process, working details and technical effects of the electronic device provided in this embodiment can be found in the first aspect of the embodiment and will not be described in detail here.

[0179] The fourth aspect of this embodiment provides a storage medium that stores instructions for the privacy protection method for sensitive data described in the first aspect of the embodiment, that is, the storage medium stores instructions, and when the instructions are run on a computer, the privacy protection method for sensitive data described in the first aspect of the embodiment is executed.

[0180] The storage medium refers to a carrier for storing data, which may include but is not limited to a floppy disk, a CD, a hard disk, a flash memory, a USB flash drive and / or a memory stick, etc. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices.

[0181] The working process, working details and technical effects of the storage medium provided in this embodiment can be found in the first aspect of the embodiment and will not be described in detail here.

[0182] A fifth aspect of this embodiment provides a computer program product comprising instructions, which, when executed on a computer, causes the computer to execute the sensitive data privacy protection method as described in the first aspect of the embodiment, wherein the computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device.

[0183] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the protection scope of the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A privacy protection method for sensitive data, characterized in that: include: Acquire target data, and process the target data in blocks to obtain multiple data blocks; Calculate the mutual information between each data block, and calculate the sensitivity of each data block based on the mutual information between each data block, so as to identify a number of sensitive data blocks from multiple data blocks according to the sensitivity of each data block; For any sensitive data block among the several sensitive data blocks, the correlation between the any sensitive data block and the corresponding target data blocks is calculated, and after all the sensitive data blocks are polled, the correlation between each sensitive data block and the corresponding target data blocks is obtained, wherein the target data blocks corresponding to any sensitive data block are all sensitive data blocks among the several sensitive data blocks except the any sensitive data block; According to the correlation between each sensitive data block and the corresponding target data blocks, a plurality of sensitive data blocks are grouped and processed to obtain a plurality of data block groups after the grouping processing; Calculate the sensitivity of each data block group, and based on the sensitivity of each data block group, obtain the sensitivity level of each data block group; Determine the desensitization method of each data block group according to the sensitivity level of each data block group, and use the desensitization method corresponding to each data block group to perform desensitization processing on the sensitive data blocks in each data block group, so as to complete the privacy protection of the target data after the desensitization processing; Based on the mutual information between each data block, the sensitivity of each data block is calculated, including: For any data block among the multiple data blocks, determine a median value of the mutual information of the any data block based on the mutual information between the any data block and each designated data block, wherein each designated data block is all data blocks among the multiple data blocks except the any data block; Calculating the entropy value of any data block and the discrete entropy between any data block and each designated data block, and selecting the maximum discrete entropy from each discrete entropy; According to the median of the mutual information, the maximum discrete entropy and the entropy value of any data block, and in accordance with the following formula (1), the sensitivity of any data block is calculated; In the above formula (1), M i Indicates the sensitivity of any data block, L max represents the maximum discrete entropy, C i represents the entropy value of any data block, D i Express the median of the mutual information.

2. The method according to claim 1, characterized in that Calculate the sensitivity of each data block group, including: For any data block group, determine whether the kth sensitive data block in the any data block group is an associated data block, wherein the associated data block is used to indicate that the kth sensitive data block has an associated relationship with at least one designated sensitive data block, and each designated sensitive data block is a sensitive data block in the any data block group except the kth sensitive data block; If so, filtering out the designated sensitive data block most relevant to the k-th sensitive data block from the designated sensitive data blocks associated with the k-th sensitive data block; Calculate the actual sensitivity of the k-th sensitive data block based on the screened designated sensitive data block and the k-th sensitive data block; Add 1 to k, and re-determine whether the kth sensitive data block in any data block group is an associated data block, until k is equal to n, and obtain the actual sensitivity of each sensitive data block in any data block group, wherein the initial value of k is 1, and n is the total number of sensitive data blocks in any data block group; The sensitivity of any data block group is determined according to the actual sensitivity of each sensitive data block in the any data block group.

3. The method according to claim 2, characterized in that Calculating the actual sensitivity of the k-th sensitive data block based on the screened designated sensitive data block and the k-th sensitive data block includes: Calculating the conditional entropy between the kth sensitive data block and the screened designated sensitive data block; Obtaining an entropy value of the k-th sensitive data block, and determining an association relationship entropy of the k-th sensitive data block based on the entropy value of the k-th sensitive data block and the conditional entropy; Obtaining the mutual information between the kth sensitive data block and each designated sensitive data block, and the discrete entropy of each sensitive data block in any one of the data block groups; The actual sensitivity of the kth sensitive data block is calculated based on the association relationship entropy, the mutual information between the kth sensitive data block and each designated sensitive data block, and the discrete entropy of each sensitive data block in any data block group.

4. The method according to claim 1, characterized in that According to the sensitivity level of each data block group, the desensitization method of each data block group is determined, including: Obtain a desensitization algorithm database, wherein the desensitization algorithm database stores desensitization methods corresponding to different sensitivity levels, and the desensitization method corresponding to any sensitivity level includes a primary desensitization method and a secondary desensitization method. The higher the sensitivity level, the higher the complexity of the primary desensitization method and the secondary desensitization method corresponding to any sensitivity level, and the primary desensitization method is to use an improved genetic algorithm for primary desensitization; According to the sensitivity level of each data block group, a desensitization method corresponding to the sensitivity level of each data block group is matched from the desensitization algorithm database, so as to obtain the desensitization method of each data block group after the matching is completed; Accordingly, the desensitization method corresponding to each data block group is adopted to desensitize the sensitive data blocks in each data block group, which includes: For the j-th sensitive data block in any data block group, encoding the j-th sensitive data block to obtain encoded data corresponding to the j-th sensitive data block; Based on the coded data, determining an initial population of the j-th sensitive data block, wherein the initial population includes a number of initial individuals, each initial individual represents a coded data block, and all initial individuals are concatenated to obtain the coded data; Using the improved genetic algorithm in the desensitization method corresponding to any one of the data block groups, perform initial desensitization processing on each initial individual in the initial population, so as to obtain a plurality of desensitized individuals after the initial desensitization processing; Based on several desensitized individuals, generate the jth sensitive data block after initial desensitization; Add 1 to j and re-encode the jth sensitive data block until j is equal to S, and obtain S sensitive data blocks after initial desensitization, where the initial value of j is 1 and S is the total number of sensitive data blocks in any data block group; The secondary desensitization method in the desensitization method corresponding to any one of the data block groups is adopted to perform secondary desensitization processing on the S sensitive data blocks after the initial desensitization processing, so as to complete the desensitization processing of the any one of the data block groups after the secondary desensitization processing.

5. The method according to claim 4, characterized in that Using the improved genetic algorithm in the desensitization method corresponding to any one of the data block groups, each initial individual in the initial population is subjected to initial desensitization processing, so as to obtain a number of desensitized individuals after the initial desensitization processing, including: Initialize the number of iterations t to 1, and perform the t-th individual selection operation on the initial population to use the selected initial individuals to form a parent population; Selecting multiple groups of parent individuals from the parent population, and calculating the crossover operator at the t-th iteration, wherein each group of parent individuals includes two parent individuals; According to the crossover operator at the tth iteration, an individual crossover operation is performed on each group of parent individuals, so as to obtain multiple crossover individuals after the individual crossover operation; Using a plurality of crossover individuals and parent individuals that are not selected from the parent population to form a child population; The mutation probability at the t-th iteration is calculated, and based on the mutation probability at the t-th iteration, an individual mutation operation is performed on each offspring individual in the offspring population, so as to obtain a mutant individual corresponding to each offspring individual after the individual mutation operation; Use the mutant individuals corresponding to each offspring individual to form the desensitized population after the tth iteration; Calculate the fitness of the desensitized population after the tth iteration, and determine whether the fitness is greater than or equal to the fitness threshold, or whether t reaches the maximum number of iterations. The fitness threshold and maximum number of iterations corresponding to different sensitivity levels are different. The fitness is used to characterize the similarity between the desensitized population after the tth iteration and the initial population, and the greater the fitness, the smaller the similarity; If not, the initial population is updated to the desensitized population after the tth iteration; Increment t by 1 and re-perform the t-th individual selection operation on the initial population until the iteration stopping condition is met, so as to obtain the several desensitized individuals based on the desensitized population that meets the iteration stopping condition, wherein the iteration stopping condition is that the fitness of the desensitized population after the t-th iteration is greater than or equal to the fitness threshold or t reaches the maximum number of iterations.

6. The method according to claim 5, characterized in that Calculate the crossover operator at the tth iteration, including: Generate a random number for the tth iteration, where the value interval of the random number is [0,1]; Based on the random number, the crossover operator at the t-th iteration is calculated using the following formula (2); In the above formula (2), δ t represents the crossover operator at the t-th iteration, μ represents the random number, a and b represent the first crossover adjustment factor and the second crossover adjustment factor respectively; Correspondingly, according to the crossover operator at the tth iteration, the individual crossover operation is performed on each group of parent individuals, which includes: For any group of parent individuals, based on the crossover operator at the t-th iteration, and using the following formula (3), an individual crossover operation is performed on the any group of parent individuals, so as to obtain crossover individuals corresponding to two parent individuals in the any group of parent individuals after the individual crossover operation; In the above formula (3), represents a parent individual in any group of parent individuals, represents another parent individual in any group of parent individuals, express The corresponding crossover individuals, express The corresponding crossover individuals.

7. The method according to claim 4, characterized in that The secondary desensitization method corresponding to any sensitivity level is to perform secondary desensitization using a desensitization model, wherein the desensitization model includes a generator and a discriminator, and the generator is used to perform secondary desensitization processing on the input data of the desensitization model. The higher the sensitivity level, the greater the number of generators in the desensitization model corresponding to the sensitivity level.

8. A privacy protection device for sensitive data, characterized in that: include: A block division unit, used for acquiring target data and performing block processing on the target data to obtain multiple data blocks; A sensitive information identification unit, used to calculate the mutual information between each data block, and based on the mutual information between each data block, calculate the sensitivity of each data block, so as to identify a number of sensitive data blocks from multiple data blocks according to the sensitivity of each data block; A grouping unit, for calculating, for any sensitive data block among the plurality of sensitive data blocks, the correlation between the any sensitive data block and the corresponding target data blocks, and obtaining the correlation between each sensitive data block and the corresponding target data blocks after all sensitive data blocks have been polled, wherein the target data blocks corresponding to any sensitive data block are all sensitive data blocks among the plurality of sensitive data blocks except the any sensitive data block; The grouping unit is further used to group the sensitive data blocks according to the correlation between each sensitive data block and the corresponding target data blocks, so as to obtain a plurality of data block groups after the grouping process; A grading unit, used to calculate the sensitivity of each data block group, and based on the sensitivity of each data block group, derive the sensitivity level of each data block group; A desensitization unit, for determining a desensitization method for each data block group according to the sensitivity level of each data block group, and performing desensitization processing on the sensitive data blocks in each data block group using the desensitization method corresponding to each data block group, so as to complete the privacy protection of the target data after the desensitization processing; Based on the mutual information between each data block, the sensitivity of each data block is calculated, including: For any data block among the multiple data blocks, determine a median value of the mutual information of the any data block based on the mutual information between the any data block and each designated data block, wherein each designated data block is all data blocks among the multiple data blocks except the any data block; Calculating the entropy value of any data block and the discrete entropy between any data block and each designated data block, and selecting the maximum discrete entropy from each discrete entropy; According to the median of the mutual information, the maximum discrete entropy and the entropy value of any data block, and in accordance with the following formula (1), the sensitivity of any data block is calculated; In the above formula (1), M i Indicates the sensitivity of any data block, L max represents the maximum discrete entropy, C i represents the entropy value of any data block, D i Express the median of the mutual information.

9. A computer program product comprising instructions, characterized in that When the instructions are executed on a computer, the computer is caused to execute the sensitive data privacy protection method as described in any one of claims 1 to 7.

Images