Data backup method and system based on encryption algorithm

Abstract

The invention relates to the technical field of data processing, in particular to a data backup method and system based on an encryption algorithm, and the method comprises the steps: obtaining to-be-backed-up data of a plurality of users, and obtaining a C-dimensional vector of the content corresponding to each data item in the to-be-backed-up data; analyzing the value domain of each data item of each data unit by adopting a region growing algorithm to obtain a first leakage probability; classifying the data items according to the confusion degree of the content of each data item in all the to-be-backed-up data of the user; combining all data items in each data unit of each user in pairs to obtain a confidentiality demand of each data unit; the method comprises the following steps: clustering data units of all users to obtain a privacy budget of each data unit, adding random Laplace noise variables to word vectors and digits in the data units, restoring the word vectors and digits, encrypting restored data fragments, and performing backup. The invention aims to encrypt the to-be-backed-up data and enhance the data security.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a data backup method and system based on an encryption algorithm. Background Art

[0002] With the rapid development of informatization, various industries generate a large amount of production and life data every day. For example, in trading platforms, a large amount of sensitive personal data such as transaction records and user account information is involved. For security reasons, platforms usually need to back up databases regularly. However, through in-depth analysis, mining and machine learning modeling of these massive data, the platform can extract the potential value in the data, thereby gaining insight into user behavior and providing users with more personalized services.

[0003] Differential privacy is a technique that interferes with user data by introducing a privacy-preserving algorithm, usually performed in a trusted execution environment. In this method, by adding an appropriate amount of noise to the statistical analysis results, the sensitive information of an individual can be effectively obscured, making it impossible to trace it back to a specific personal identity. This method not only protects the user's privacy data, but also ensures that statistically significant general information can still be extracted during data analysis, meeting the platform's needs in data mining and user behavior analysis while avoiding potential data leakage risks.

[0004] However, existing differential privacy protection technologies have difficulty in selecting key parameters for differential privacy, that is, it is difficult to balance data availability and the degree of protection of personal information; when the added random noise falls within a very small interval, attackers can easily obtain accurate user sensitive data through differential attacks and probabilistic reasoning attack analysis. Summary of the invention

[0005] In view of the above, it is necessary to provide a data backup method and system based on encryption algorithm to solve the above problems.

[0006] The first aspect of the present application provides a data backup method based on an encryption algorithm, the method comprising: Acquire data to be backed up of several users, wherein the data to be backed up includes several data units; the data units include several data items, and obtain a C-dimensional vector for the content corresponding to each data item; wherein C is a preset value; A region growing algorithm is used on a set consisting of C-dimensional vectors corresponding to each data item in each data unit of all users, and a first leakage probability of each data item in each data unit of each user is obtained in combination with the number of elements in the set; According to the disorder degree of the C-dimensional vector corresponding to each data item of each data unit in the data to be backed up by all users, all data items of each user are classified in combination with the clustering algorithm; all data items in each data unit of each user are combined in pairs, and based on the first leakage probability of the data items in all combinations and the classification information, the confidentiality requirement of each data unit of each user is obtained; According to the similarity characteristics between all data items of any two data units belonging to different users, combined with the clustering algorithm and the confidentiality requirements, the privacy budget of each data unit of each user is obtained; based on the privacy budget, random Laplace noise variables are added to the word vectors and numbers in the data units and restored, and the restored data shards are encrypted and backed up.

[0007] The step of obtaining a C-dimensional vector for the content corresponding to each data item is as follows: If the content of the data item is a number, the number is placed in the first dimension of the C-dimensional vector, and the remaining dimensions are filled with 0; if the content of the data item is text, the text is segmented to obtain the word vector corresponding to the text, and each word vector is reduced to a C-dimensional vector. If it is less than C dimensions, 0 is added at the end, and the C-dimensional vectors of all word vectors are added together to obtain the C-dimensional vector of the corresponding data item.

[0008] The process of obtaining the first leakage probability of each data item of each data unit of each user is as follows: The set of all C-dimensional vectors corresponding to the same data item of all users in each data unit is recorded as the value range of the data item corresponding to each data unit; A region growing algorithm is used for the value range of each data item of each data unit to obtain several regions; The formula form of the first leakage probability of the jth data item of the i-th data unit of the e-th user is: ;in, represents the first leakage probability of the jth data item of the ith data unit of the eth user; The number of users in the region where the jth data item of the ith data unit of the eth user is located; Indicates the number of users; The number of C-dimensional vectors representing the value range of the j-th data item of the i-th data unit.

[0009] The step of using a region growing algorithm for the range of each data item of each data unit to obtain a plurality of regions is as follows: The C-dimensional vector corresponding to the j-th data item of the i-th data unit of the first acquired user is used as the growth seed; The growth condition of the region growing algorithm is that the cosine similarity between the C-dimensional vector corresponding to the j-th data item of the i-th data unit of the user and the growing seed is greater than a preset value.

[0010] The clustering algorithm is used to classify all data items of each user, specifically: Obtain the information entropy of the C-dimensional vector corresponding to the content of each data item in the data items of all users; A clustering algorithm is used to cluster all data items of each user to obtain a preset number of clusters; The information entropy mean corresponding to all data items in each cluster is obtained, and all data items in the cluster with the largest information entropy mean are recorded as commodity data items; and all data items in the cluster with the smallest information entropy mean are recorded as personal data items.

[0011] The confidentiality requirement of each data unit of each user is obtained as follows: Analyze the first leakage probability and category information of the data items in each combination to obtain the second leakage probability and interception probability of each combination of each data unit of each user; The product of the second leakage probability of each combination of each data unit of each user and the interception probability is used as the confidentiality requirement information of each combination of each data unit of each user; The maximum value of the confidentiality requirement information in each data unit is taken as the confidentiality requirement of each data unit.

[0012] The formula of the second leakage probability is specifically: ; The formula form of the intercept probability is specifically: ;in, represents the second leakage probability of the bth combination data item of the ith data unit of the eth user; represents the number of commodity data items in the bth combination data item of the ith data unit of the eth user, represents the number of individual data items in the bth combination of data items in the ith data unit of the eth user; u represents the number of all data items contained in a user; represents the first leakage probability of all commodity data items in the bth combination data item of the i-th data unit of the e-th user; The first leakage probability of all individual data items in the b-th combination data item of the i-th data unit of the e-th user; represents the normalization function; represents the interception probability of the bth combination of data items of the i-th data unit of the e-th user; The preset attack type required for each combination of data items representing each data unit of each user.

[0013] The process of obtaining the privacy budget of each data unit of each user is specifically as follows: According to the similarity characteristics between the C-dimensional vectors corresponding to all data items of any two data units belonging to different users, the difference between any two data units belonging to different users is obtained; The difference is used as the distance between data units in the clustering algorithm, the data units of all users are clustered, the mean confidentiality requirement of all data units in each cluster is calculated, and the product of the normalized result of the mean confidentiality requirement and the preset parameter value is used as the privacy budget of all data units in each cluster.

[0014] The obtaining of the difference between any two data units belonging to different users is specifically: The cumulative sum of cosine similarities between C-dimensional vectors corresponding to each data item of any two data units belonging to different users is calculated, and the negative correlation mapping results of all the cumulative sums of the any two data units are used as the difference between the any two data units.

[0015] In a second aspect, an embodiment of the present application also provides a data backup system based on an encryption algorithm, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor implements the steps of any one of the above methods when executing the computer program.

[0016] This application has at least the following beneficial effects: In order to solve the problems in existing differential privacy protection technologies, such as the difficulty in balancing data availability and the degree of protection of personal information, and the difficulty in determining the addition of random noise, the embodiments of the present application analyze the data to be backed up of platform users. First, a regional growing algorithm is used for the set consisting of the C-dimensional vectors corresponding to each data item in each data unit of all users, and the first leakage probability of each data item of each data unit of each user is obtained in combination with the number of elements in the set, which helps to evaluate the privacy leakage risk of each data item of the user. Furthermore, since the personal information of the user changes little in the data of various platforms, while the transaction information changes at any time, all data items of each user are classified according to the degree of confusion of the C-dimensional vectors corresponding to each data item in each data unit in the data to be backed up of all users in combination with the clustering algorithm, which helps to evaluate the privacy risk and leakage possibility of the data according to the changing characteristics of the data item information. Such evaluation can help to better understand the data The sensitivity of the items can be determined, and a basis can be provided for taking more targeted privacy protection measures. Because the combination of different data items can release more information, all data items in each data unit of each user are combined in pairs, and the confidentiality requirements of each data unit of each user are obtained based on the first leakage probability and classification information of the data items in all combinations, which helps to accurately evaluate the privacy protection requirements of each data unit, and facilitates the subsequent determination of relevant parameters in the differential privacy protection technology according to the leakage probability and the sensitivity of the data combination, and encryption is performed to further ensure the security of the data. The data is then backed up, so that even if the data is leaked, personal privacy data cannot be obtained, which is conducive to improving the security of backup data in the database. This method introduces a privacy protection mechanism directly on the user's device, ensuring that the sensitive information of the individual will not be exposed during the storage and processing of the data, effectively preventing possible privacy attacks, while retaining the analytical value of the data. This method significantly enhances data security and user privacy protection. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 A flowchart of a data backup method based on an encryption algorithm provided in one embodiment of the present application; Figure 2 A flowchart for obtaining a privacy budget provided for one embodiment of the present application. DETAILED DESCRIPTION

[0018] In the description of the embodiments of the present application, words such as "exemplary", "or", "for example" and the like are used to indicate examples, illustrations or descriptions. Any embodiment or design described as "exemplary" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary", "or", "for example" and the like is intended to present related concepts in a concrete manner.

[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art in the present application. The terms used in the specification of the present application are only for the purpose of describing specific embodiments and are not intended to limit the present application.

[0020] It should also be noted that the terms "first" and "second" in this application and the accompanying drawings are used to distinguish similar objects, rather than to describe a specific order or sequence. The method disclosed in the embodiments of the present application or the method shown in the flow chart includes one or more steps for implementing the method. Without departing from the scope of protection of the present application, the execution order of multiple steps can be interchanged with each other, and some steps can also be deleted.

[0021] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.

[0022] The following is a detailed description of a data backup method and system based on an encryption algorithm provided by the present application with reference to the accompanying drawings.

[0023] See also Figure 1 , which shows a flowchart of a data backup method based on an encryption algorithm provided by an embodiment of the present application, the method comprising the following steps: The first step is to obtain data to be backed up of several users, where the data to be backed up includes several data units; the data units include several data items, and a C-dimensional vector is obtained for the content corresponding to each data item.

[0024] This application aims to find a balance between ensuring the availability and effectiveness of data and protecting personal privacy in the process of using privacy algorithms to interfere with user data. Based on this, the data to be backed up of several users is obtained, and the data to be backed up includes several data units, and the data units include several data items.

[0025] In this embodiment, taking a trading platform as an example, the user's data to be backed up includes personal information and several transaction information. The personal information and transaction information are both a data unit. The personal information includes several data items such as name, gender, account ID, etc.; the transaction information includes several data items such as product name, order number, shipping address, and delivery address.

[0026] Since the content corresponding to the data item contains text data, this embodiment uses the maximum matching word segmentation algorithm to segment the text data, and converts any vocabulary into a word vector through the Word2Vec model. Among them, converting text data into a word vector is an existing well-known technology, and this application will not elaborate on it.

[0027] Obtain a C-dimensional vector for the content corresponding to each data item: if the content of the data item is a number, place the number in the first dimension of the C-dimensional vector, and fill in the remaining dimensions with 0; if the content of the data item is text, multiple word vectors are obtained after word segmentation, then each word vector is first reduced to a C-dimensional vector using principal component analysis, and if it is less than C dimensions, 0 is added at the end, and then the C-dimensional vectors corresponding to each word vector are added together as the vector of the corresponding data item. Since there are some ultra-high-dimensional word vectors in the content of the data item, such as product names, the value of C should not be too small. In this embodiment, the value of C is 3000, and the implementer can adjust it according to actual conditions. It should be understood that in this embodiment, one data item corresponds to one C-dimensional vector.

[0028] The second step: using a region growing algorithm on a set consisting of C-dimensional vectors corresponding to each data item in each data unit of all users, combined with the number of elements in the set, to obtain a first leakage probability of each data item in each data unit of each user.

[0029] In the differential privacy protection of this application, the ideal goal of privacy protection is to use data units as units: even if a data unit is leaked, the user to which the data unit belongs cannot be accurately identified. For example, in a trading platform, when the Laplace mechanism is used for protection, since the privacy requirements of data items such as gender, account ID, product name, order number, etc. are different, the risk of leaking user privacy when different data items are queried will also be different. In addition, different types of goods also have different privacy risks. For example, there are differences in confidentiality requirements between daily necessities and niche goods. The latter have a higher risk of leaking user information due to their strong directionality, and the consequences of leakage are more serious. Therefore, it is necessary to set a different privacy budget for each data unit according to the privacy risks of different data units.

[0030] First, we analyze the value range of each data item in each data unit. For example, the value range of the data item gender contains 2 elements, the value range of the product name is relatively wide, and the value range of the account ID is related to the number of users. The value range represents the richness of the data item. The smaller the value range, the more monotonous the data item, and the smaller the user's optional range. Then, when the data item is obtained, the possibility of leaking user information is smaller. However, at the same time, although the value range of the account ID is large, it is directional, and each user's account ID is unique. Therefore, when the data item is obtained, the possibility of leaking user information is very high.

[0031] In the present application, a set consisting of all C-dimensional vectors corresponding to the same data item in each data unit of all users is obtained, which is recorded as the value range of the data item corresponding to each data unit; a regional growing algorithm is used for the value range of each data item in each data unit to obtain several regions, wherein the C-dimensional vector corresponding to the j-th data item of the i-th data unit of the user obtained first is used as a growth seed; the growth condition of the regional growing algorithm is: the cosine similarity between the C-dimensional vector corresponding to the j-th data item of the i-th data unit of the user and the growth seed is greater than a preset value. It should be noted that when the growth is terminated, the C-dimensional vector corresponding to the j-th data item of the i-th data unit of the user obtained first that has not grown is used as the growth seed to continue growing until all users have been grown; if there is a user who does not belong to any region, it will be incorporated into the cluster where the user with the largest cosine similarity is located.

[0032] The vector distribution in the region can reflect the data distribution characteristics of the data item. For clusters with more users, the first leakage probability is lower. On this basis, the weaker the directionality of the data item, the lower the first leakage probability. Based on this, according to the number of elements in the region where each C-dimensional vector corresponds to each data unit of each user and each data item, combined with the number of elements in the value range of each data item, the first leakage probability of each data unit of each user and each data item is obtained. The formula form of the first leakage probability of the jth data item of the ith data unit of the eth user is: ;in, represents the first leakage probability of the jth data item of the ith data unit of the eth user; The number of users in the region where the jth data item of the ith data unit of the eth user is located; Indicates the number of users; The number of C-dimensional vectors representing the value range of the j-th data item of the i-th data unit, The closer it is to 1, the stronger the directivity.

[0033] The third step: classify all data items of each user according to the degree of disorder of the C-dimensional vector corresponding to each data item of each data unit in the data to be backed up by all users, using a clustering algorithm; combine all data items in each data unit of each user in pairs, and obtain the confidentiality requirement of each data unit of each user based on the first leakage probability of the data items in all combinations and the classification information.

[0034] When analyzing the data items in each data unit, the distribution characteristics of the data items need to be considered. For example, for data items such as product names, the first leakage probability is closely related to the popularity and demand level of the product. If a product is common and widely in demand, such as daily necessities, many people will buy it, then the risk of leaking user privacy is low because it has a strong non-directional nature. However, for some special products, such as experimental materials or large musical instruments, their purchases tend to be more directional and seasonal, so once they are acquired, the risk of leaking user information will be higher. In other words, the popularity and personalization of a product directly affect its first leakage probability. The risk of privacy leakage for common products is low, while the risk of leakage for personalized products is relatively high.

[0035] According to the degree of disorder of the content of each data item of each data unit in the data to be backed up by all users, all data items of each user are classified in combination with the clustering algorithm: the information entropy of the C-dimensional vector corresponding to the content of each data item in the data items of all users is obtained; all data items of each user are clustered by using the clustering algorithm to obtain a preset number of clusters; the mean value of the information entropy corresponding to all data items in each cluster is obtained, and all data items of the cluster with the largest mean value of the information entropy are recorded as commodity data items; all data items of the cluster with the smallest mean value of the information entropy are recorded as personal data items. In this embodiment, the preset number is 2, which is used to classify all data items of the user into two categories.

[0036] The first leakage probability of a data unit cannot be calculated simply by adding the first leakage probability of each data item. The reason is that the amount of information revealed by different combinations of data items is different. For example, when the data items such as goods, prices and delivery addresses are combined, the combination of goods and prices may reflect more relevant information about the goods themselves, while the combination of goods and delivery addresses may reveal more details about user behavior. For example, if goods are combined with delivery addresses, the user's purchasing habits and interests can be inferred, especially if it involves daily necessities, it may indicate that the user tends to buy goods on the online platform; if it is other niche goods, the combined information can more clearly reflect the user's personalized needs and behavior patterns. Therefore, the combination of different data items can release more information than just the content exposed by the individual data items.

[0037] In addition, the difficulty or probability of intercepting different combinations of data items is not a simple linear relationship. Taking differential attacks and repeated query attacks as examples, the difference between two data units of the same user lies in the product-related information, so only a small amount of differential attacks are needed to determine the product-related privacy. The repeated part of the two data units of the same user lies in personal related information, such as account nickname, mobile phone number, delivery address, etc. This information only needs to add a small amount of noise to achieve the purpose of privacy protection, but if the noise is increased within a small range, then the user's relevant privacy can be obtained through repeated queries.

[0038] For all data items in each data unit of each user, the second leakage probability and interception probability of each combination of each data unit of each user are obtained based on the first leakage probability and category information of the data items in each combination, where the formula form of the second leakage probability is: ,in, represents the second leakage probability of the bth combination data item of the ith data unit of the eth user; represents the number of commodity data items in the bth combination data item of the ith data unit of the eth user, represents the number of individual data items in the bth combination of data items in the ith data unit of the eth user; u represents the number of all data items contained in a user; represents the first leakage probability of all commodity data items in the bth combination data item of the i-th data unit of the e-th user; The first leakage probability of all individual data items in the b-th combination data item of the i-th data unit of the e-th user; Represents the normalization function.

[0039] The formula for intercept probability is: ;in, represents the interception probability of the bth combination of data items of the i-th data unit of the e-th user; represents the number of commodity data items in the bth combination data item of the ith data unit of the eth user, represents the number of individual data items in the bth combination of data items of the ith data unit of the eth user; It represents the preset attack type required for each combination of data items of each data unit of each user. The value of this embodiment is 2, and the implementer can adjust it according to the actual situation. exp() represents an exponential function with a natural constant as the base.

[0040] The product of the second leakage probability and the interception probability of each combination of each data unit of each user is used as the confidentiality requirement factor of each combination of each data unit of each user. The degree of data security lies in the weak point, so the maximum value of the confidentiality requirement information in each data unit is used as the confidentiality requirement of each data unit, which can ensure that all combined data items in the data unit can meet the confidentiality requirement.

[0041] The fourth step: based on the similarity characteristics between all data items of any two data units belonging to different users, combined with the clustering algorithm and the confidentiality requirements, obtain the privacy budget of each data unit of each user; based on the privacy budget, add random Laplace noise variables to the word vectors and numbers in the data units and restore them, encrypt the restored data shards and back them up.

[0042] The above steps only analyze a single user and do not take into account the differences between all users, which may lead to uneven distribution of privacy budgets among users. Therefore, it is necessary to analyze the number of data units of different users and build a unified privacy budget allocation model.

[0043] First, based on the similarity characteristics between the C-dimensional vectors corresponding to all data items of any two data units belonging to different users, the difference between any two data units belonging to different users is obtained: the cumulative sum of the cosine similarities between the C-dimensional vectors corresponding to each data item of any two data units belonging to different users is calculated, and the negative correlation mapping results of all the cumulative sums of the any two data units are used as the difference between the any two data units. In this embodiment, the cumulative sum of the cosine similarities between the C-dimensional vectors corresponding to each data item of any two data units belonging to different users is recorded as A, and the formula form of the difference between the any two data units is: .

[0044] Different users have similar purchased products, and similar products should have similar confidentiality requirements. The difference between data units is used as the distance, and the DBSCAN algorithm is used to cluster the data units of all users. In this embodiment, the neighborhood radius is , the minimum number of sample points is 2; get several clusters, record the data units contained in the same cluster as similar data units, and record the mean of the confidentiality requirements of similar data units as the confidentiality requirements of the cluster; calculate the privacy budget of the data units in each cluster , its formula form is: ,in: represents the confidentiality requirements of the data units of each cluster class, Indicates the preset parameter value. In this embodiment, the value is , which means that the privacy budget is controlled between 0 and 10, and implementers can adjust it according to specific circumstances; It should be understood that the privacy budgets of data units in the same cluster are equal. The flowchart for obtaining the privacy budget is as follows: Figure 2 shown.

[0045] Get a privacy budget After that, random Laplace noise variables are added to the word vectors and numbers in the data units. , and then restore the noisy word vector to the word with the largest cosine similarity to obtain the noisy data unit; where, Indicates that it obeys the scale parameter for The independent and identically distributed random Laplace noise variables are known in the art and will not be elaborated in this application.

[0046] The noise-added data unit is encrypted in pieces using the AES algorithm to complete the encryption, obtain the encrypted data unit, create a backup file storage database, and export the encrypted data unit to the backup file. This is an existing public technology and will not be elaborated on.

[0047] The present application adds noise to the backup data and encrypts it, so that even if the backup data is intercepted, the original data cannot be obtained.

[0048] Based on the same inventive concept as the above method, an embodiment of the present application also provides a data backup system based on an encryption algorithm, including a memory, a processor, and a computer program stored in the memory and running on the processor, and when the processor executes the computer program, the steps of any one of the above-mentioned data backup methods based on the encryption algorithm are implemented.

[0049] In summary, the present application aims at the problem that it is difficult to balance data availability and the degree of protection of personal information, and it is difficult to determine the problem of adding random noise in the existing differential privacy protection technology. The data to be backed up of the platform users is analyzed. First, a regional growing algorithm is used for the set consisting of the C-dimensional vectors corresponding to each data item in each data unit of all users, and the first leakage probability of each data item of each data unit of each user is obtained in combination with the number of elements in the set, which helps to evaluate the privacy leakage risk of each data item of the user; further, since the user's personal information changes little in the data of various platforms, while the transaction information changes at any time, all data items of each user are classified according to the degree of confusion of the C-dimensional vectors corresponding to each data item of each data unit in the data to be backed up of all users, combined with the clustering algorithm, which helps to evaluate the privacy risk and leakage possibility of the data according to the changing characteristics of the data item information. This evaluation can help to better understand the data The sensitivity of the data items can be determined, and a basis can be provided for taking more targeted privacy protection measures. Because the combination of different data items can release more information, all data items in each data unit of each user are combined in pairs, and the confidentiality requirements of each data unit of each user are obtained based on the first leakage probability and classification information of the data items in all combinations, which helps to accurately evaluate the privacy protection requirements of each data unit, and facilitates the subsequent determination of relevant parameters in the differential privacy protection technology according to the leakage probability and the sensitivity of the data combination, and encrypts the data to further ensure the security of the data. Then the data is backed up, so that even if the data is leaked, personal privacy data cannot be obtained, which is conducive to improving the security of backup data in the database. This method introduces a privacy protection mechanism directly on the user's device, ensuring that the sensitive information of the individual will not be exposed during the storage and processing of the data, effectively preventing possible privacy attacks, and retaining the analytical value of the data. This method significantly enhances data security and user privacy protection.

[0050] The flowchart and block diagram in the accompanying drawings show the possible architecture, function and operation of the system, method and computer program product according to the embodiment of the present application. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of the code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. In some alternative implementations, the function marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two continuous boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, which can depend on the functions involved. In the description corresponding to the flowchart and the block diagram in the accompanying drawings, the operations or steps corresponding to different boxes can also occur in a different order from the order disclosed in the description, and sometimes there is no specific order between different operations or steps. For example, two continuous operations or steps can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, which can depend on the functions involved. Each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented by a dedicated hardware-based system that performs the specified functions or actions, or may be implemented by a combination of dedicated hardware and computer instructions.

[0051] For those skilled in the art, it is obvious that the present application is not limited to the details of the above exemplary embodiments, and the present application can be implemented in other specific forms without departing from the basic features of the present application. Therefore, no matter from which point of view, the above embodiments of the present application should be regarded as exemplary and non-restrictive; the technical solutions recorded in the above embodiments are modified, or some of the technical features are replaced by equivalents, which does not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.

Claims

1. A data backup method based on encryption algorithm, characterized in that: The method comprises the following steps: Acquire data to be backed up of several users, wherein the data to be backed up includes several data units; the data units include several data items, and obtain a C-dimensional vector for the content corresponding to each data item; wherein C is a preset value; A region growing algorithm is used on a set consisting of C-dimensional vectors corresponding to each data item in each data unit of all users, and a first leakage probability of each data item in each data unit of each user is obtained in combination with the number of elements in the set; According to the disorder degree of the C-dimensional vector corresponding to each data item of each data unit in the data to be backed up by all users, all data items of each user are classified in combination with the clustering algorithm; all data items in each data unit of each user are combined in pairs, and based on the first leakage probability of the data items in all combinations and the classification information, the confidentiality requirement of each data unit of each user is obtained; According to the similarity characteristics between all data items of any two data units belonging to different users, combined with the clustering algorithm and the confidentiality requirements, the privacy budget of each data unit of each user is obtained; based on the privacy budget, random Laplace noise variables are added to the word vectors and numbers in the data units and restored, and the restored data shards are encrypted and backed up.

2. A data backup method based on encryption algorithm as claimed in claim 1, characterized in that: The C-dimensional vector is obtained for the content corresponding to each data item, specifically: If the content of the data item is a number, the number is placed in the first dimension of the C-dimensional vector, and the remaining dimensions are filled with 0; if the content of the data item is text, the text is segmented to obtain the word vector corresponding to the text, and each word vector is reduced to a C-dimensional vector. If it is less than C dimensions, 0 is added at the end, and the C-dimensional vectors of all word vectors are added together to obtain the C-dimensional vector of the corresponding data item.

3. The data backup method based on encryption algorithm as claimed in claim 1, characterized in that: The process of obtaining the first leakage probability of each data item of each data unit of each user is as follows: The set of all C-dimensional vectors corresponding to the same data item of all users in each data unit is recorded as the value range of the data item corresponding to each data unit; A region growing algorithm is used for the value range of each data item of each data unit to obtain several regions; The formula form of the first leakage probability of the jth data item of the i-th data unit of the e-th user is: ;in, represents the first leakage probability of the jth data item of the ith data unit of the eth user; The number of users in the region where the jth data item of the ith data unit of the eth user is located; Indicates the number of users; The number of C-dimensional vectors representing the value range of the j-th data item of the i-th data unit.

4. A data backup method based on encryption algorithm as claimed in claim 3, characterized in that: The steps of using a region growing algorithm for the range of each data item of each data unit to obtain a plurality of regions are as follows: The C-dimensional vector corresponding to the j-th data item of the i-th data unit of the first acquired user is used as the growth seed; The growth condition of the region growing algorithm is that the cosine similarity between the C-dimensional vector corresponding to the j-th data item of the i-th data unit of the user and the growing seed is greater than a preset value.

5. The data backup method based on encryption algorithm as claimed in claim 1, characterized in that: The combined clustering algorithm is used to classify all data items of each user, specifically: Obtain the information entropy of the C-dimensional vector corresponding to the content of each data item in the data items of all users; A clustering algorithm is used to cluster all data items of each user to obtain a preset number of clusters; The information entropy mean corresponding to all data items in each cluster is obtained, and all data items in the cluster with the largest information entropy mean are recorded as commodity data items; and all data items in the cluster with the smallest information entropy mean are recorded as personal data items.

6. The data backup method based on encryption algorithm as claimed in claim 1, characterized in that: The confidentiality requirement of each data unit of each user is specifically: Analyze the first leakage probability and category information of the data items in each combination to obtain the second leakage probability and interception probability of each combination of each data unit of each user; The product of the second leakage probability of each combination of each data unit of each user and the interception probability is used as the confidentiality requirement information of each combination of each data unit of each user; The maximum value of the confidentiality requirement information in each data unit is taken as the confidentiality requirement of each data unit.

7. A data backup method based on encryption algorithm as claimed in claim 6, characterized in that: The formula form of the second leakage probability is specifically: ; The formula form of the intercept probability is specifically: ;in, represents the second leakage probability of the bth combination data item of the ith data unit of the eth user; represents the number of commodity data items in the bth combination data item of the ith data unit of the eth user, represents the number of individual data items in the bth combination of data items in the ith data unit of the eth user; u represents the number of all data items contained in a user; represents the first leakage probability of all commodity data items in the bth combination data item of the i-th data unit of the e-th user; The first leakage probability of all individual data items in the b-th combination data item of the i-th data unit of the e-th user; represents the normalization function; represents the interception probability of the bth combination of data items of the i-th data unit of the e-th user; The preset attack type required for each combination of data items representing each data unit of each user.

8. The data backup method based on encryption algorithm as claimed in claim 1, characterized in that: The process of obtaining the privacy budget of each data unit of each user is specifically as follows: According to the similarity characteristics between the C-dimensional vectors corresponding to all data items of any two data units belonging to different users, the difference between any two data units belonging to different users is obtained; The difference is used as the distance between data units in the clustering algorithm, the data units of all users are clustered, the mean confidentiality requirement of all data units in each cluster is calculated, and the product of the normalized result of the mean confidentiality requirement and the preset parameter value is used as the privacy budget of all data units in each cluster.

9. A data backup method based on encryption algorithm as claimed in claim 8, characterized in that: The difference between any two data units belonging to different users is obtained as follows: The cumulative sum of cosine similarities between C-dimensional vectors corresponding to each data item of any two data units belonging to different users is calculated, and the negative correlation mapping results of all the cumulative sums of the any two data units are used as the difference between the any two data units.

10. A data backup system based on an encryption algorithm, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 9 are implemented.

Images