A desensitization processing method and system for sensitive information of digital power services

Abstract

The present application relates to the field of data processing technology, and specifically to a method and system for desensitizing sensitive information of digital power services, the method comprising: obtaining information data of digital power services corresponding to each user, dividing it into first information and second information, segmenting the data of each information in the first information to obtain a subsequence of each information in the first information, and determining the difference coefficient of each subsequence; analyzing the similarity of the information data vector of each information in the first information and other information, and determining the information sensitivity coefficient of each information in the first information; determining the state comparison coefficient of each information in the second information according to the degree of difference between each information in the second information and other information with respect to the collected data, and then dividing the users; desensitizing the information data of users in each cluster for user query. The present application can improve the accuracy of dynamic desensitization processing of sensitive information of power service data.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a method and system for desensitizing sensitive information of digital power services. Background Art

[0002] In the field of power services, a large amount of user sensitive information is involved, including user power consumption data (real-time power, voltage, current, etc.), equipment operation data (transformer oil temperature, line load, etc.), marketing data (user payment records, business processing information, etc.), etc. A large amount of data provides strong support for the operation management, optimized scheduling and customer service of power companies, but it also increases the difficulty of sensitive data management. Different types of data have different sensitivities and business values. For example, the real-time power consumption in user power consumption data may involve the user's living habits and privacy, while the key parameters in the equipment operation data are related to the safe and stable operation of the power grid, which requires sophisticated sensitive data identification and protection strategies.

[0003] In the process of desensitizing sensitive information of power services through centralized or decentralized differential privacy methods in the prior art, the addition of random disturbances does not fully consider the differences in sensitive changes in information data of different types of users of actual power service data, resulting in a large processing deviation in the balance between sensitive data protection and data sharing availability for power data released at this stage, making the dynamic desensitization processing of sensitive information of power service data less accurate. Summary of the invention

[0004] In order to solve the above technical problems, the purpose of this application is to provide a method and system for desensitizing sensitive information of digital power services. The technical solutions adopted are as follows:

[0005] The present application embodiment provides a method for desensitizing sensitive information of digital power services, comprising the following steps:

[0006] Obtain data on various information of digital power services for each user, and encrypt sensitive attribute information;

[0007] Based on the data type of the information data, the other information of each user after removing the sensitive attribute information is divided into first information and second information, and the non-stationary time series mutation detection algorithm is used to segment the data of each information in the first information at all collection times to obtain the subsequence of each information in the first information; the difference coefficient of each subsequence is determined according to the difference between each subsequence and other subsequences;

[0008] Constructing information data vectors of each information in the first information according to the difference coefficients of each information subsequence in the first information, analyzing the similarity between each information in the first information and the information data vectors of other information, and determining the information sensitivity coefficient of each information in the first information; determining the state comparison coefficient of each information in the second information according to the degree of difference between each information in the second information and other information with respect to the collected data; and dividing the users by combining the information sensitivity coefficients of all information in the first information of all users and the state comparison coefficients of all information in the second information;

[0009] The information data of users in each cluster is desensitized for user query.

[0010] Preferably, the multiple information includes the user's name, identity information, real-time electricity consumption, electricity usage, electricity usage time, electricity usage address, payment amount, payment method, arrears record, power equipment operating parameters and status information, and the sensitive attribute information is the user's name, identity information and electricity usage address.

[0011] Preferably, the first information and the second information further include:

[0012] The information of numerical data in the remaining information after removing the sensitive attribute information is taken as the first information, and the first information includes real-time electricity consumption, electricity consumption time, payment amount, and operating parameters of power equipment. The information of non-numerical data in the remaining information after removing the sensitive attribute information is taken as the second information, and the second information includes payment method, arrears record and status information.

[0013] Preferably, the calculation formula for the difference coefficient of each subsequence is: ;in, represents the difference coefficient of the x-th subsequence of the w-th information in the first information; represents the DTW distance between the x-th and i-th subsequences of the w-th information in the first information; represents the absolute value of the difference between the x-th and i-th subsequences of the w-th information in the first information with respect to the number of data in the sequence, represents an exponential function with a natural constant as base; Represents the number of subsequences of the w-th information in the first information.

[0014] Preferably, determining the information sensitivity coefficient of each information in the first information further includes: taking the average value of the similarity between each information in the first information and all other information in the first information regarding the information data vector as the information sensitivity coefficient of each information in the first information.

[0015] Preferably, the determination of the state comparison coefficient of each information in the second information further includes:

[0016] A sequence composed of data of each information in the second information at all acquisition moments is taken as the information data sequence of each information, and the average value of the difference between each information in the second information and all other information in the second information regarding the information data sequence is taken as the state comparison coefficient of each information in the second information.

[0017] Preferably, the dividing of users further comprises:

[0018] The mean of the information sensitivity coefficients of all information in each user's first information is taken as the first eigenvalue; the mean of the state contrast coefficients of all information in each user's second information is taken as the second eigenvalue, and the data pair consisting of the first eigenvalue and the second eigenvalue is taken as the coordinate value of each user. The coordinate values ​​of all users are clustered to obtain clusters, and the users corresponding to all coordinate values ​​in each cluster are regarded as the same category.

[0019] Preferably, the process of clustering the coordinate values ​​of all users to obtain clusters further includes:

[0020] The information sensitivity coefficient of all information in the first information of each user and the state comparison coefficient of all information in the second information are used to form the decision vector of each user. The product of the similarity of the decision vectors between different users and the Euclidean distance between the corresponding coordinate values ​​of different users is used as the metric distance in the clustering process.

[0021] An embodiment of the present application also provides a desensitizing processing system for sensitive information of digital power services, including a memory, a processor, and a computer program stored in the memory and running on the processor, and the processor implements the steps of any one of the above methods when executing the computer program.

[0022] From the above, it can be seen that the desensitization processing method and system for digital power service sensitive information provided by this application have at least the following beneficial effects:

[0023] The present application considers that in the desensitization processing of information data of digital power services, the traditional desensitization processing method based on differential privacy has a centralized effect, resulting in large differences in the availability of information data of power services after desensitization processing. Therefore, based on the above-mentioned considerations, the unpublished information data is obtained, and the information data of sensitive attributes therein is first format-preserving and encrypted, and the format in which the power information data is published is retained. Further comparative analysis is performed on the changing characteristics of the user's power information data, and the user's information data is characterized based on the comparative analysis results. The specific division process is judged by the comparative differences in different time periods and the differences in status information. The beneficial effect is that, combined with the differences in sensitive information reflected in the information data of the user during the power use process, the information data with similar sensitive characteristics in different time periods is accurately divided, and then the user's information data is accurately differentially privacy processed, reducing the centralized effect of the privacy processing process and the influence of overly dispersed differential privacy processing on the differences in data availability after desensitization processing, thereby improving the accuracy of dynamic desensitization processing of sensitive information of power service data. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present application or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0025] Figure 1 A flowchart of the steps of a method for desensitizing sensitive information of digital power services provided in this application;

[0026] Figure 2 Schematic diagram of the processing flow of power service information data provided for this application. DETAILED DESCRIPTION

[0027] In order to further explain the technical means and effects adopted by the present application to achieve the predetermined invention purpose, the following, in combination with the accompanying drawings and preferred embodiments, describes in detail the desensitization processing method and system for sensitive information of digital power services proposed in the present application, its specific implementation method, structure, characteristics and effects. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0028] Unless otherwise specified and limited, terms such as "comprises", "comprising" or any other variants thereof are intended to cover non-exclusive inclusion, so that a circuit structure, article or device including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such articles or devices. In the absence of further restrictions, an element defined by the sentence "comprising a ..." does not exclude the presence of other identical elements in the article or device including the element. In addition, the term "and\or" used herein includes any and all combinations of one or more related listed items. All technical and scientific terms used herein have the same meaning as those commonly understood by technicians in the technical field of this application.

[0029] The following is a detailed description of a method and system for desensitizing sensitive information of digital power services provided by the present application in conjunction with the accompanying drawings.

[0030] See also Figure 1 , which shows a flowchart of a method for desensitizing sensitive information of digital power services provided by an embodiment of the present application, including the following steps:

[0031] Step 1: Obtain data on various information of digital power services corresponding to each user, and encrypt the data of sensitive attribute information.

[0032] First, obtain the unpublished digital power service information data of the power grid power supply company. Specifically, in this embodiment, obtain Information data of users in power services, in this embodiment The value is 1000. The information includes the user's name, identity information, real-time power consumption, power consumption, power consumption time, power consumption address, payment amount, payment method, arrears record, power equipment operating parameters and status information, respectively represented by alphabetical order. To represent; for non-numeric information such as name, identity information, and device status information, UTF-8 encoding technology is used for processing. For numeric information, the number of each type of information collected is , The value of the embodiment is 500. The specific UTF-8 encoding technology is well known to those skilled in the art and will not be described in detail.

[0033] Furthermore, the sensitive attribute information in the digital power service information data is initially processed, such as the user's name, identity information and electricity address, so the power service information data of each user is As input, the format-preserving encryption (FPE) algorithm is used to preliminarily process sensitive attribute information to obtain the information data of each user's electricity service. The encryption result is obtained and the original information format is retained.

[0034] In order to further analyze the changes in sensitive information in the information data of power services, the data of each type of user information at all collection moments are arranged in chronological order as a sequence, which is used as the information data sequence of each type of information.

[0035] At this point, the information data sequence of all users is obtained.

[0036] Step 2: Based on the data type of the information data, the other information of each user after removing the sensitive attribute information is divided into first information and second information, and the non-stationary time series mutation detection algorithm is used to segment the data of each information in the first information at all collection times to obtain the subsequence of each information in the first information; the difference coefficient of each subsequence is determined according to the difference between each subsequence and other subsequences.

[0037] Normally, when using differential privacy to desensitize digital information data of power services, there is a centralized feature, that is, random disturbances are added to all information data of power services, resulting in serious centralization of all data disturbance processing, and the availability of all power service information data after differential privacy may be poor; while the decentralized approach often uses differential privacy processing for the information data of each user in the power information service data. The addition of random disturbances is affected by the differences in changes in information data of different users in different time periods, resulting in a large deviation in the disturbance processing of the information data, which not only increases the complexity of data processing, but also does not significantly improve the data availability compared to centralized processing.

[0038] In response to the above problems, this embodiment obtains information data of the power service of the power supply company, and encrypts the information data with sensitive attributes in the information data with format preservation. Furthermore, considering that the sensitive changes reflected in the information data of different users in different time periods during actual electricity consumption are different, the information change characteristics of the users can be compared and analyzed, and the information data of the users can be characterized by the comparative analysis results.

[0039] First, feature extraction is performed on the power service information data corresponding to each user, and the sensitive features of the user information data are accurately analyzed based on the results of feature extraction. Specifically, the user information is classified, and the numerical data (real-time power consumption, power consumption, power consumption time, payment amount, power equipment operation parameters) in the remaining information after removing sensitive attribute information are all used as the first information, and the non-numerical data (payment method, arrears record and status information) in the remaining information are all used as the second information. The purpose of dividing the first information and the second information is to analyze the correlation differences based on the power information of users in different time periods, and highlight the sensitive characteristics of the information data of the user's power service.

[0040] Furthermore, for each user's first information and second information, for the information data sequence of each type of information in the first information, each information data sequence is taken as input, and the Bernaola Galvan segmentation algorithm (non-stationary time series mutation detection) is used to obtain the division result of each information data sequence, so as to analyze the difference in data changes in a short period of time for each user's first information, and for each subsequence in the division result, the DTW distance between each subsequence and each other subsequence is calculated. The larger the DTW distance, the greater the difference in changes in the power service information in different time periods.

[0041] Further, for each type of information in the first information, the difference in the number of data in each subsequence corresponding to each type of information and each other subsequence is calculated. The greater the difference in the number, the greater the difference in the duration of the impact range of the data change of the information in different time periods, that is, the greater the difference in the characteristics of the user information data in different time periods represented by the information data change, the less likely the current information data reflects the user's sensitive information characteristics; the difference coefficient of the sensitive feature comparison of the user information data is calculated by comparing the differences of each type of data of each user in different time periods. , the specific calculation formula is:

[0042] ;in, represents the difference coefficient of the x-th subsequence of the w-th information in the first information; represents the DTW distance between the x-th and i-th subsequences of the w-th information in the first information; represents the absolute value of the difference between the x-th and i-th subsequences of the w-th information in the first information with respect to the number of data in the sequence, represents an exponential function with a natural constant as base; Represents the number of subsequences of the wth type of information in the first information; the larger the calculated difference coefficient is, the greater the difference in local characteristics of the data of the information of the current user over time is during the data sensitive feature analysis of the information, reflecting that the data sensitive features of the user's information are relatively small.

[0043] Step three: construct the information data vector of each information in the first information according to the difference coefficient of each information subsequence in the first information, analyze the similarity between the information data vector of each information in the first information and other information, and determine the information sensitivity coefficient of each information in the first information; determine the state comparison coefficient of each information in the second information according to the degree of difference between each information in the second information and other information regarding the collected data; and divide the users according to the information sensitivity coefficients of all information in the first information of all users and the state comparison coefficients of all information in the second information.

[0044] Based on the above process, the differences in data changes of each user's information data in different time periods can be analyzed. Furthermore, for each type of information in each user's first information, a vector composed of difference coefficients of all subsequences corresponding to each type of information is used as the information data vector of each type of information for sensitive feature analysis, that is, based on the differences in data changes of different types of information data of users in power services in different time periods, the sensitive features reflected by each type of information data are accurately analyzed.

[0045] Specifically, for each type of information in the first information of each user, the average value of the similarity between each type of information and all other information about the information data vector is used as the information sensitivity coefficient of each type of information. In this embodiment, the measure of the similarity of the information data vector adopts cosine similarity, and the implementer may adopt other calculation methods in the actual application process. The larger the information sensitivity coefficient is, the closer the difference characteristics of the information data between the current information and other information in different time periods are, that is, the more significant the corresponding information sensitivity characteristics are.

[0046] In order to further determine the information characteristics of each user, for each user's second information, for the information data sequence of each information in the second information, the average value of the difference between each information in the second information and all other information in the second information about the information data sequence is used as the state comparison coefficient of each information in the second information. Preferably, in this embodiment, the average value of the Levenshtein distance between each information in the second information and all other information in the second information about the information data sequence is used as the state comparison coefficient of each information in the second information. The larger the state comparison coefficient, the greater the difference in the data state change of the user's information, and the more significant the sensitive characteristics of the user's information data.

[0047] Furthermore, in order to accurately divide user information data, the average of the information sensitivity coefficients of all information in the first information of each user is used as the first eigenvalue of each user; the average of the state contrast coefficients of all information in the second information of each user is used as the second eigenvalue of each user, and the data pair consisting of the first eigenvalue and the second eigenvalue is used as the coordinate value mapped to each user. The purpose is to determine the information data differences between different users in combination with the sensitive change difference characteristics of the numerical data and non-numerical data of each user, which is used to divide user information and then accurately judge the information characteristics between users.

[0048] Through the above analysis, in this embodiment, the coordinate values ​​corresponding to all users are mapped to a two-dimensional rectangular coordinate system, and the density peak clustering algorithm is used to cluster all users. In order to accurately compare the sensitive features of different user information data, in the clustering process, the information sensitivity coefficient of all information in the first information of each user and the state comparison coefficient of all information in the second information are used to form the judgment vector of each user, and the similarity of the judgment vectors between different users and the product of the Euclidean distance between the corresponding coordinate values ​​of different users are used as the metric distance in the clustering process, and finally the division results of all users can be obtained, wherein the density peak clustering algorithm is a technology well known to those skilled in the art and will not be described in detail. In this embodiment, the similarity of the judgment vectors between different users is calculated using cosine similarity.

[0049] At this point, users are divided based on comparative analysis of their information data characteristics.

[0050] Step 4: Desensitize the information data of users in each cluster for user query.

[0051] Through the above process, users are divided, and user information data is further divided based on this. The information data sequences of all users in each cluster after clustering constitute the information data set of each cluster. The differential privacy algorithm is used to blur the data in the information data set of each cluster. In this embodiment, for the convenience of description and understanding, for the information data sequences of all users in each cluster, the desensitized data obtained after blurring all data is recorded as the perturbation data set.

[0052] At this point, according to the above process of this embodiment, the information data set within each cluster can be fuzzy processed separately through the differential privacy algorithm, which can improve the accuracy of information data processing for each category of users, and the desensitization processing effect of user information data can be improved through adaptive processing of user information data within each cluster.

[0053] Furthermore, based on the perturbation data set obtained through the above-mentioned desensitization processing, it can be provided for user query. The user can query through mean difference and send a mean query request. The server estimates the mean of the data based on the received mean query request to obtain the value of the mean query. It should be noted that in other embodiments, the implementer can also query through planned query and Top-k query, and the specific query process will not be repeated here.

[0054] Specifically, the processing flow of power service information data is as follows: Figure 2 As shown, Figure 2 In the figure, numbers 1, 2, 3, and N represent the information data of user 1, user 2, user 3, and user N in power services, respectively; numbers a1, a2, a3, and an represent the information data sets of cluster 1, cluster 2, cluster 3, and cluster n after clustering; numbers b1, b2, b3, and bn represent the data of the information data sets of cluster 1, cluster 2, cluster 3, and cluster n after privacy differential processing, respectively. The set of all data after privacy differential processing is taken as the perturbation data set, and can be queried through planned query, mean query, and Top-k query.

[0055] Based on the same inventive concept as the above method, an embodiment of the present application also provides a desensitizing processing system for sensitive information of digital power services, 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 methods for desensitizing processing sensitive information of digital power services are implemented.

[0056] It is to be understood that the sequence of the embodiments of the present application described above is for description only and does not represent the advantages and disadvantages of the embodiments. The above describes specific embodiments of the present specification. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0057] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

[0058] The above content is only an implementation method of the present application and is not intended to limit the scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the present application specification and drawings, or directly or indirectly used in other related technical fields, are also included in the protection scope of the present application.

Claims

1. A method for desensitizing sensitive information of digital power services, characterized in that: The following steps are involved: Obtain data on various information of digital power services for each user, and encrypt sensitive attribute information; Based on the data type of the information data, the other information of each user after removing the sensitive attribute information is divided into first information and second information, and the non-stationary time series mutation detection algorithm is used to segment the data of each information in the first information at all collection times to obtain the subsequence of each information in the first information; the difference coefficient of each subsequence is determined according to the difference between each subsequence and other subsequences; Constructing an information data vector of each information in the first information according to the difference coefficient of each information subsequence in the first information, analyzing the similarity between each information in the first information and the information data vector of other information, and determining the information sensitivity coefficient of each information in the first information; Determine the status comparison coefficient of each information in the second information according to the difference between each information in the second information and other information about the collected data; divide the users by combining the information sensitivity coefficient of all information in the first information of all users and the status comparison coefficient of all information in the second information; Desensitize the information data of users in each cluster for user query; The dividing of users further comprises: The average value of the information sensitivity coefficient of all information in the first information of each user is taken as the first eigenvalue; the average value of the state contrast coefficient of all information in the second information of each user is taken as the second eigenvalue, and the data pair consisting of the first eigenvalue and the second eigenvalue is taken as the coordinate value of each user, and the coordinate values ​​of all users are clustered to obtain clusters, and the users corresponding to all coordinate values ​​in each cluster are taken as the same category; The process of clustering the coordinate values ​​of all users to obtain clusters further includes: The information sensitivity coefficient of all information in the first information of each user and the state comparison coefficient of all information in the second information are used to form the decision vector of each user, and the product of the similarity of the decision vectors between different users and the Euclidean distance between the corresponding coordinate values ​​of different users is used as the metric distance in the clustering process; The desensitizing process is performed on the information data of the users in each cluster for user query, and the specific steps include: A differential privacy algorithm is used to fuzzy the data in the information data set of each cluster to obtain desensitized data, and all desensitized data are recorded as perturbed data sets; The perturbed data set is provided for user query. The user queries through mean difference and sends a mean query request. The server estimates the mean of the data based on the received mean query request and obtains the value of the mean query.

2. A method for desensitizing sensitive information of digital power services as claimed in claim 1, characterized in that: The multiple information includes the user's name, identity information, real-time electricity consumption, electricity usage, electricity usage time, electricity usage address, payment amount, payment method, arrears record, power equipment operating parameters and status information, and the sensitive attribute information is the user's name, identity information and electricity usage address.

3. A method for desensitizing sensitive information of digital power services as claimed in claim 2, characterized in that: The first information and the second information further include: The information of numerical data in the remaining information after removing the sensitive attribute information is taken as the first information, and the first information includes real-time electricity consumption, electricity consumption time, payment amount, and operating parameters of power equipment. The information of non-numerical data in the remaining information after removing the sensitive attribute information is taken as the second information, and the second information includes payment method, arrears record and status information.

4. A method for desensitizing sensitive information of digital power services as claimed in claim 1, characterized in that: The corresponding calculation formula for the difference coefficient of each subsequence is: ;in, represents the difference coefficient of the x-th subsequence of the w-th information in the first information; represents the DTW distance between the x-th and i-th subsequences of the w-th information in the first information; represents the absolute value of the difference between the x-th and i-th subsequences of the w-th information in the first information with respect to the number of data in the sequence, represents an exponential function with a natural constant as base; Represents the number of subsequences of the w-th information in the first information.

5. A method for desensitizing sensitive information of digital power services as claimed in claim 1, characterized in that: Determining the information sensitivity coefficient of each information in the first information further includes: taking the average value of the similarity between each information in the first information and all other information in the first information regarding the information data vector as the information sensitivity coefficient of each information in the first information.

6. A method for desensitizing sensitive information of digital power services as claimed in claim 1, characterized in that: The determination of the state comparison coefficient of each information in the second information further includes: A sequence composed of data of each information in the second information at all acquisition moments is taken as the information data sequence of each information, and the average value of the difference between each information in the second information and all other information in the second information regarding the information data sequence is taken as the state comparison coefficient of each information in the second information.

7. A desensitization processing system for sensitive information of digital power services, 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 6 are implemented.

Images