Differential privacy algorithm optimization method and device for enterprise credit data privacy protection

Abstract

The embodiment of the invention provides a differential privacy algorithm optimization method and device for enterprise credit data privacy protection, and the method comprises the steps: analyzing the query type of enterprise credit data, and adaptively adjusting a noise parameter; based on an improved noise mechanism, selectively increasing noise; combining the plurality of data queries into one data query; preprocessing the enterprise credit data subjected to noise processing so as to perform privacy protection on the enterprise credit data through a differential privacy algorithm introduced by noise; noise is flexibly adjusted according to specific query conditions, accuracy of results can be remarkably improved while privacy is protected, performance of data query and analysis is improved through selective noise addition, unnecessary noise interference is reduced, multiple queries are combined into one query, noise introduction frequency is reduced, and data query efficiency is improved. The comparability among different query results can be improved.

Description

Technical Field

[0001] The present application relates to the field of differential privacy optimization technology, and more specifically, to a differential privacy algorithm optimization method and device for protecting the privacy of enterprise credit data. Background Art

[0002] The differential privacy algorithm for protecting corporate credit data privacy is a cryptographic technology that protects corporate privacy data. It aims to ensure that individual information is not leaked when publishing statistical data or analyzing corporate credit results. Its core idea is to introduce random noise into the results of corporate credit data queries, so that external observers cannot accurately identify the participation of a specific individual, thereby protecting the privacy of corporate credit data.

[0003] The formal definition of differential privacy is based on the "ε-differential privacy" model. This model stipulates that for any two similar data sets (i.e., differing in only one record), the difference in the results for the same query will not exceed a small threshold (ε). Specifically, differential privacy requires that the distribution of query results should be almost the same regardless of whether a particular individual is included in the data set. This means that even if an attacker knows the query results, he cannot determine whether a certain individual is in the data set, thereby protecting the privacy of the user.

[0004] The key technology to achieve differential privacy is noise addition. Laplace noise or Gaussian noise is usually used, and the size of the noise is determined by calculating the sensitivity of the query result (that is, the degree of influence of a single record in the corporate credit data set on the query result). The higher the sensitivity, the more noise needs to be added to ensure the security of individual data.

[0005] Noise introduction is the core mechanism of the traditional differential privacy algorithm for protecting corporate credit data privacy, which aims to protect individual privacy by adding random noise to the query results. However, this process inevitably leads to a decrease in the practicality of corporate credit data, which is specifically reflected in the following aspects: First, the addition of noise will affect the accuracy of the results. When a large amount of noise is added to the query results, it may obscure the true trend of corporate credit data, thereby affecting the effectiveness of decision-making. Second, the introduction of noise will reduce the signal-to-noise ratio of the data, resulting in a decrease in the performance of the analysis model. In machine learning tasks, accurate data is needed when training models. If the noise in the training data is too large, the learning process of the model may be disturbed, which ultimately leads to a decrease in the generalization ability of the model in practical applications. In addition, the size of the noise often needs to be adjusted according to the sensitivity of the specific query, which may lead to a decrease in the comparability between different query results. When different queries are applied on the same data set, the inconsistency of the noise makes comparative analysis complicated and reduces the availability of the data. Summary of the invention

[0006] The purpose of the embodiments of the present application is to provide a differential privacy algorithm optimization method and device for protecting the privacy of enterprise credit data, so as to solve the problems of introducing noise when performing differential privacy algorithms in existing methods, causing unnecessary noise interference, and improving the comparability and accuracy of different query results.

[0007] In a first aspect, an embodiment of the present application provides a differential privacy algorithm optimization method for protecting the privacy of enterprise credit data, including: Analyze the query type of corporate credit data and adaptively adjust noise parameters; Selective noise addition based on improved noise mechanism; Merge multiple data queries into one data query; The enterprise credit data after noise processing is preprocessed to protect the privacy of the enterprise credit data through the differential privacy algorithm introduced by noise.

[0008] In the above implementation process, the embodiment of the present application analyzes the query type of enterprise credit data and adaptively adjusts the noise parameters; selectively increases noise based on the improved noise mechanism; merges multiple data queries into one data query; pre-processes the enterprise credit data after noise processing to protect the privacy of the enterprise credit data through the differential privacy algorithm introduced by noise; flexibly adjusts the noise size according to the specific circumstances of the query, which can significantly improve the accuracy of the results while protecting privacy, and reduces unnecessary noise interference through selective noise addition, thereby improving the performance of data query and analysis. At the same time, multiple queries are merged into one query to reduce the frequency of noise introduction, which can improve the comparability between different query results.

[0009] Furthermore, the query type of analyzing the enterprise credit data and adaptively adjusting the noise parameters include: Analyze the query types of corporate credit data and assess the sensitivity of corporate credit data; Determining influencing factors of the noise parameter, wherein the influencing factors include an allocated privacy budget, a sensitivity measure of the query, and a distribution characteristic of the data; Set noise parameters, dynamically adjust noise parameters, and adjust noise parameters according to application scenarios; Verify the settings of noise parameters and continuously optimize the noise parameters.

[0010] In the above implementation process, by flexibly adjusting the noise size according to the specific circumstances of the query, the accuracy of the results can be significantly improved while protecting privacy.

[0011] Furthermore, the selectively increasing noise based on the improved noise mechanism includes: Classify and standardize corporate credit data; Select the set noise type to add and determine the noise parameters; Determine the amount of noise based on data sensitivity and query type; Generate random noise and add the random noise to the data to update the data; Verify and evaluate the data after adding noise.

[0012] In the above implementation process, unnecessary noise interference is reduced and the performance of data query and analysis is improved by selectively adding noise.

[0013] Furthermore, merging the multiple data queries into one data query includes: Analyze query requirements, determine query relevance, and merge multiple data queries into one data query; Select a merge strategy and optimize query statements; Modify the query interface, process query results, and monitor query performance.

[0014] In the above implementation process, multiple queries are merged into one query to reduce the frequency of noise introduction, which can improve the comparability between different query results and allow more accurate cross-query analysis, thereby helping enterprises make better decisions.

[0015] Furthermore, the preprocessing of the enterprise credit data after noise processing includes: Clarify the source of corporate credit data and conduct a preliminary review of corporate credit data; Detect outliers in corporate credit data; Remove noise from corporate credit data.

[0016] In the above implementation process, data cleaning and preprocessing are used to remove outliers and noise, thereby improving the quality of raw data, which directly affects the accuracy and effectiveness of the final analysis results, thereby improving support for business decisions.

[0017] Furthermore, the analysis of the query type of the enterprise credit data and the assessment of the sensitivity of the enterprise credit data include: Identify the types of queries that exist in the corporate credit data, where different query types have different data sensitivity and privacy requirements; Assess the sensitivity of corporate credit data, including the assessment of the company's financial status, default record, and credit rating; The factors affecting the determination of noise parameters include: Allocate a privacy budget to control the degree of added noise; Different shares of the privacy budget are allocated to different types of queries and data sensitivity; Establish a query sensitivity metric and determine the query sensitivity by setting a method, where the higher the sensitivity of the query, the larger the noise parameter added; Establish the distribution characteristics of corporate credit data and use statistical methods to analyze the distribution of data; The step of setting noise parameters, dynamically adjusting noise parameters, and adjusting noise parameters according to application scenarios includes: For different types of queries, formulate different noise parameter setting strategies; For range queries, the noise parameters are determined based on the query range and data distribution; for count queries, the noise parameters are set based on the total number of enterprises in the data set and the accuracy requirements of the query; for sum queries, the noise parameters are set based on the numerical range of the data and the importance of the query; Different noise parameters are set according to the sensitivity level of the data, wherein the sensitivity level includes a high sensitivity level, a medium sensitivity level and a low sensitivity level; Among them, the noise parameter set at the high sensitivity level is greater than the noise parameter set at the medium sensitivity level, and the noise parameter set at the medium sensitivity level is greater than the noise parameter set at the low sensitivity level; Regularly re-evaluate the data, analyze changes in query types and data sensitivity, and adjust noise parameters based on the evaluation results; Adjust noise parameters according to application scenarios; The verification of the setting of the noise parameters and the continuous optimization of the noise parameters include: Select a set amount of corporate credit data as samples, simulate different types of queries, and add noise according to the set noise parameters; Compare the difference between the query results after adding noise and the real results, and evaluate the impact of noise parameters on the accuracy of query results; If the difference exceeds the set threshold, adjust the noise parameters; Continuously optimize the settings of noise parameters; Establish a feedback mechanism to collect user feedback on the accuracy of query results and the degree of privacy protection; Based on feedback and actual application, the noise parameter setting strategy is continuously adjusted.

[0018] Furthermore, the classification and standardization of enterprise credit data includes: Classify corporate credit data and divide the data into different categories according to the sensitivity of the data and the type of query; Standardize the classified data; The step of selecting and adding a set noise type and determining noise parameters includes: Select the set noise type to add; Determine the noise parameters according to the sensitivity of the data and the query type; the noise parameters include the mean, variance and standard deviation of the noise; The step of generating random noise and adding the random noise to data to update the data includes: Generate random noise using a random number generator and add the generated random noise to the data; Update the data with random noise added to the database to replace the original data; The verification and evaluation of the data after adding noise includes: Use data validation tools to validate the data after adding random noise; Conduct privacy assessment on data after adding random noise; Performance evaluation is performed on data after adding random noise.

[0019] Furthermore, analyzing the query requirements, determining the query relevance, and combining multiple data queries into one data query includes: Obtain usage scenarios and business requirements for corporate credit data and determine query relevance; Analyze the correlation between different queries; If multiple queries involve the same data fields or the similarity of query conditions exceeds the set threshold, they are merged into one query; The step of selecting a merge strategy and optimizing a query statement includes: Select a merge strategy based on the query relevance and business needs. The merge strategies include: conditional merge, result merge, and function merge. When designing a merge query, optimize the query statement, wherein the methods of optimizing the query statement include avoiding repeated queries, using indexes, and limiting the query result set; The modifying of the query interface, processing of the query results, and monitoring of the query performance include: According to the designed merged query, modify the data query interface so that the query interface accepts the merged query request; After executing the merged query, process the query results; If the result merging strategy is used, multiple query results will be merged and sorted; If the function merging strategy is used, call the corresponding function for processing; After implementing the merged query, use a database performance monitoring tool or a log analysis tool to monitor the query performance, where the query performance includes the query execution time, CPU usage, and memory usage.

[0020] Furthermore, the sources of enterprise credit data shall be clarified and the preliminary review of enterprise credit data shall include: Determine the source channels of corporate credit data and understand the methods and time frame for data collection; Conduct a preliminary review of the collected data to understand the format, field meaning, and data type; The outlier detection of enterprise credit data includes: Calculate the basic statistics of the data, understand the distribution of the data through basic statistics, and find outliers; Use visualization tools to display the distribution of data; Using model methods to detect outliers, where the model methods include: cluster analysis, outlier detection algorithms; Wherein, removing noise from the enterprise credit data includes: Using filtering methods to remove noise from corporate credit data, where filtering methods include mean filtering, median filtering, and Gaussian filtering; If there is noise in the data that is correlated with other variables, use regression analysis to remove the noise; For time series data or data with a continuously changing trend, data smoothing methods are used to remove noise, where data smoothing methods include moving average method and exponential smoothing method.

[0021] In a second aspect, the embodiment of the present application provides a differential privacy algorithm optimization device for protecting the privacy of enterprise credit data, including: The query analysis module is used to analyze the query type of enterprise credit data and adaptively adjust the noise parameters; A noise adjustment module for selectively increasing noise based on an improved noise mechanism; A query merging module, used to merge multiple data queries into one data query; The data processing module is used to pre-process the enterprise credit data after noise processing so as to protect the privacy of the enterprise credit data through the differential privacy algorithm introduced by noise.

[0022] In a third aspect, an embodiment of the present application provides an electronic device, including: A processor, a memory and a bus, wherein the processor is connected to the memory via the bus, and the memory stores computer-readable instructions. When the computer-readable instructions are executed by the processor, they are used to implement the differential privacy algorithm optimization method for protecting the privacy of enterprise credit data as described above.

[0023] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a server, the differential privacy algorithm optimization method for protecting the privacy of enterprise credit data as described above is implemented. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments of the present application will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0025] Figure 1 A flowchart of a differential privacy algorithm optimization method for protecting the privacy of enterprise credit data provided in an embodiment of the present application; Figure 2 It is a structural schematic diagram of a differential privacy algorithm optimization device for protecting the privacy of enterprise credit data provided in an embodiment of the present application; Figure 3 It is a structural schematic diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0026] The technical solutions in the embodiments of the present application will be described below in conjunction with the drawings in the embodiments of the present application.

[0027] It should be noted that similar reference numerals and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings. At the same time, in the description of this application, the terms "first", "second", etc. are only used to distinguish the description and cannot be understood as indicating or implying relative importance.

[0028] Please see Figure 1 , Figure 1 A flowchart of a differential privacy algorithm optimization method for protecting the privacy of enterprise credit data provided in an embodiment of the present application. The differential privacy algorithm optimization method for protecting the privacy of enterprise credit data includes: 100. Analyze the query type of corporate credit data and adaptively adjust the noise parameters.

[0029] Specifically, the size of the noise is dynamically adjusted according to the specific situation of the data query. For example, for different types of queries or data sensitivity, the noise parameters can be flexibly set to balance privacy protection and data accuracy.

[0030] 110. Analyze the query types of corporate credit data and evaluate the sensitivity of corporate credit data.

[0031] Specifically, the query types existing in the enterprise credit data are determined, wherein different query types have different data sensitivity and privacy requirements.

[0032] Optionally, determine the query types that may exist in the enterprise credit data, such as range query (querying the number of enterprises within a specific credit score range), count query (the total number of enterprises with good credit in a specific industry), sum query (the total credit score of enterprises in a certain region), etc. Different query types may have different data sensitivity and privacy requirements. For example, range query may be more likely to expose the distribution characteristics of data, while count query may be relatively less sensitive to data.

[0033] Specifically, the sensitivity of corporate credit data is assessed, where the sensitivity assessment includes the assessment of the company's financial status, default record, and credit rating.

[0034] Optionally, assess the sensitivity of corporate credit data, taking into account specific information contained in the data, such as the company's financial status, default record, credit rating, etc. Generally speaking, data involving sensitive information such as core financial data and defaults has a higher sensitivity, while some relatively broad credit rating information may have a lower sensitivity.

[0035] 120. Determine influencing factors of the noise parameter, wherein the influencing factors include an allocated privacy budget, a sensitivity measure of the query, and a distribution characteristic of the data.

[0036] 121. Allocate a privacy budget to control the degree of added noise.

[0037] Understandably, differential privacy protection usually allocates a privacy budget to control the degree of noise addition. When faced with different types of queries and data sensitivity, the privacy budget needs to be reasonably allocated.

[0038] 122. Different shares of privacy budget should be allocated to different types of queries and data sensitivity. For example, for highly sensitive data and queries, a smaller privacy budget share can be allocated to increase the amount of noise and improve the degree of privacy protection; for low-sensitivity data and queries, a larger privacy budget share can be allocated to reduce the impact of noise on the accuracy of query results.

[0039] 123. Establish a measurement index for query sensitivity and determine the sensitivity of the query by setting a method, wherein the higher the sensitivity of the query, the larger the noise parameter added.

[0040] Optionally, a metric for query sensitivity is established, which is usually related to the degree of change in the query results. For example, for a count query, if the change of a record in the data set has a small impact on the query results, then the query sensitivity is low.

[0041] For example, the sensitivity of the query can be determined by analyzing the mathematical properties of the query or conducting experiments; the more sensitive the query is, the larger the noise parameter needs to be added.

[0042] 124. Establish the distribution characteristics of corporate credit data and use statistical methods to analyze the distribution of data.

[0043] Specifically, consider the distribution characteristics of corporate credit data. Alternatively, if the data distribution is relatively concentrated, then smaller noise may be sufficient to provide privacy protection; if the data distribution is relatively dispersed, larger noise may be required to hide the specific characteristics of the data.

[0044] Among them, statistical methods can be used to analyze the distribution of data, such as mean, variance, standard deviation, etc., to understand the degree of concentration and dispersion of the data.

[0045] 130. Set noise parameters, dynamically adjust noise parameters, and adjust noise parameters according to application scenarios.

[0046] Specifically, different noise parameter setting strategies are formulated for different types of queries.

[0047] Exemplarily, for range queries, the noise parameter is determined based on the query range size and data distribution. If the query range is small and the data distribution is relatively concentrated, the noise parameter can be appropriately reduced; if the query range is large or the data distribution is relatively dispersed, the noise parameter needs to be increased.

[0048] Exemplarily, for count queries, the noise parameters are set according to the total number of enterprises in the data set and the query precision requirements. If the total number of enterprises is large and the query precision requirements are not high, a smaller noise parameter can be used; if the total number of enterprises is small or the query precision requirements are high, the noise parameter needs to be increased.

[0049] For example, for sum queries, the noise parameter is set by considering the numerical range of the data and the importance of the query. If the numerical range of the data is large and the query result has a greater impact on the decision, the noise parameter can be increased to improve the degree of privacy protection; if the numerical range of the data is small or the query result has a smaller impact on the decision, the noise parameter can be appropriately reduced.

[0050] Specifically, different noise parameters are set according to the sensitivity level of the data, wherein the sensitivity level includes a high sensitivity level, a medium sensitivity level, and a low sensitivity level.

[0051] Among them, the noise parameter set at the high sensitivity level is greater than the noise parameter set at the medium sensitivity level, and the noise parameter set at the medium sensitivity level is greater than the noise parameter set at the low sensitivity level.

[0052] For example, for highly sensitive data, such as a company's core financial data and breach of contract records, a larger noise parameter is set to ensure that the privacy of the data is fully protected.

[0053] For example, for moderately sensitive data, such as a company's credit rating and some financial indicators, moderate noise parameters can be set to minimize the impact on the accuracy of query results while protecting privacy.

[0054] For example, for low-sensitivity data, such as basic information of an enterprise and some public credit data, a smaller noise parameter can be set to improve the accuracy of the query results.

[0055] Specifically, the data is re-evaluated regularly, changes in query types and data sensitivity are analyzed, and noise parameters are adjusted based on the evaluation results.

[0056] It is understandable that as time goes by and data changes, query types and data sensitivity may change, so a mechanism to dynamically adjust noise parameters is needed.

[0057] For example, the data may be re-evaluated regularly to analyze changes in query types and data sensitivity, and the noise parameters may be adjusted based on the evaluation results. For example, the data may be analyzed once a week or a month, and the noise parameters may be adjusted based on the analysis results.

[0058] Specifically, the noise parameters are adjusted according to the application scenario.

[0059] Optionally, consider the actual application scenarios of corporate credit data, such as credit assessment, risk analysis, regulatory reporting, etc. Different application scenarios have different requirements for data accuracy and privacy protection.

[0060] When setting noise parameters, the needs of the application scenario need to be considered comprehensively. For example, when conducting credit assessment, a higher query result accuracy may be required, so the noise parameter can be appropriately reduced; when submitting reports to regulators, more attention may be paid to data privacy protection, so the noise parameter can be increased.

[0061] 140. Verify the settings of noise parameters and continuously optimize the noise parameters.

[0062] Specifically, a set number of corporate credit data are selected as samples, different types of queries are simulated, and noise is added according to the set noise parameters; the difference between the query results after adding noise and the actual results is compared, and the impact of the noise parameters on the accuracy of the query results is evaluated; if the difference exceeds the set threshold, the noise parameters are adjusted until a reasonable balance is reached.

[0063] Specifically, continuously optimize the setting of noise parameters; establish a feedback mechanism to collect user feedback on the accuracy of query results and the degree of privacy protection; and continuously adjust the setting strategy of noise parameters based on feedback and actual application situations to improve the effectiveness of differential privacy algorithms in protecting the privacy of corporate credit data.

[0064] It is understandable that for different types of queries or data sensitivity, flexible setting of noise parameters requires comprehensive consideration of factors such as query type, data sensitivity, privacy budget allocation, query sensitivity measurement, data distribution characteristics, and adjustment and optimization based on actual application scenarios; by reasonably setting noise parameters, it is possible to protect the privacy of corporate credit data while minimizing the impact on the accuracy of query results.

[0065] As described above, the embodiments of the present application can significantly improve the accuracy of the results while protecting privacy by flexibly adjusting the noise level according to the specific circumstances of the query.

[0066] 200. Selectively increase noise based on an improved noise mechanism.

[0067] Specifically, reducing the addition of noise in certain insensitive or low-sensitivity data queries and increasing noise in high-sensitivity queries can improve data practicality while maintaining privacy.

[0068] 210. Classify and standardize corporate credit data.

[0069] Optionally, the enterprise credit data is classified into different categories according to the sensitivity of the data and the query type. For example, the data can be classified into high-sensitivity data, medium-sensitivity data, and low-sensitivity data, or divided into count query data, range query data, and sum query data.

[0070] Optionally, the classified data is standardized so that different types of data have the same scale and distribution. Exemplarily, this can be achieved by methods such as data normalization and standardization, wherein the standardization process can improve the effect of noise addition and the usability of data.

[0071] 220. Select the set noise type to add and determine the noise parameters.

[0072] Optionally, a set noise type is selected for addition. Exemplarily, common noise types include Gaussian noise, Laplace noise, etc. Gaussian noise is suitable for continuous data, while Laplace noise is suitable for discrete data; select a suitable noise type according to the type and characteristics of the data.

[0073] Optionally, the noise parameters are determined according to the sensitivity of the data and the query type; wherein the noise parameters include the mean, variance, and standard deviation of the noise. The selection of the noise parameters should be determined according to the sensitivity of the data and the query type. Exemplarily, for highly sensitive data and high-precision queries, a smaller noise parameter can be selected; for less sensitive data and less-precision queries, a larger noise parameter can be selected.

[0074] 230. Determine the size of the noise based on data sensitivity and query type.

[0075] Optionally, based on the data sensitivity strategy, for highly sensitive data, larger noise is added to improve the privacy protection of the data. The size of the noise can be determined according to the sensitivity level of the data. For example, for the most sensitive data, noise with a standard deviation of a certain proportion of the data range can be added.

[0076] For medium-sensitive data, moderate noise is added to minimize the impact on data availability while protecting privacy. The size of the noise can be determined based on the sensitivity of the data and the query type. For example, for medium-sensitive data, noise with a standard deviation that is a smaller proportion of the data range can be added.

[0077] For low-sensitivity data, less noise or no noise is added to improve data usability. Whether to add noise can be determined based on the sensitivity of the data and the query type. For example, for the lowest sensitive data, no noise or very little noise can be added.

[0078] Optionally, based on the query type strategy, for count queries, smaller noise can be added, because count queries are relatively insensitive to data changes. The size of the noise can be determined based on the accuracy requirements of the query and the total amount of data. For example, for count queries, noise with a standard deviation that is a certain proportion of the total amount of data can be added.

[0079] For range queries, moderate noise can be added because range queries are sensitive to data distribution. The size of the noise can be determined based on the query range and data distribution. For example, for range queries, noise with a standard deviation that is a certain percentage of the query range can be added.

[0080] For sum queries, larger noise can be added because sum queries are sensitive to data changes. The size of the noise can be determined based on the accuracy requirements of the query and the total amount of data. For example, for sum queries, noise with a standard deviation that is a larger proportion of the total amount of data can be added.

[0081] 240. Generate random noise and add the random noise to the data, and update the data.

[0082] Optionally, a random number generator is used to generate random noise, and the generated random noise is added to the data. A pseudo-random number generator or a hardware random number generator may be used to ensure randomness and unpredictability of the noise.

[0083] Optionally, the data after adding random noise is updated to the database to replace the original data. When updating the data, the integrity and consistency of the data need to be ensured to avoid data loss or error.

[0084] 250. Verify and evaluate the data after adding noise.

[0085] Optionally, a data verification tool is used to verify the data after random noise is added. The data after noise is added is verified to ensure the validity and availability of the data. The data verification tool can be used or the data can be manually checked for accuracy and completeness.

[0086] Optionally, a privacy assessment is performed on the data after random noise is added to ensure that the privacy of the data is effectively protected. The privacy protection degree of the data can be assessed by using a privacy assessment tool or performing a simulated attack.

[0087] Optionally, a performance evaluation is performed on the data after random noise is added to evaluate the impact of noise addition on data query performance and data availability. Performance testing tools or actual application scenarios can be used to evaluate data performance and availability.

[0088] As described above, the embodiments of the present application reduce unnecessary noise interference and improve the performance of data query and analysis by selectively adding noise.

[0089] 300. Merge multiple data queries into one data query.

[0090] Specifically, through the joint analysis mechanism, multiple queries are merged into one query to reduce the frequency of noise introduction, so that the overall noise level can be reduced, thereby improving the accuracy of each individual query.

[0091] 310. Analyze query requirements, determine query relevance, and merge multiple data queries into one data query.

[0092] Optionally, obtain the usage scenarios and business requirements of corporate credit data and determine the query relevance. Specifically, deeply understand the usage scenarios and business requirements of corporate credit data, determine which queries are often used together, and the relationship between them. For example, in credit assessment, it may be necessary to query the company's financial status, credit rating, and default record at the same time.

[0093] Optionally, analyze the correlation between different queries; if multiple queries involve the same data field or the similarity of query conditions exceeds a set threshold, merge them into one query. For example, querying the sales of a company in a specific time period and querying the profit in the same time period may be merged into one query to query the company's financial status in that time period.

[0094] 320. Select a merge strategy and optimize query statements.

[0095] 321. Select a merge strategy based on the query relevance and business needs, where the merge strategies include: conditional merge, result merge, and function merge.

[0096] For example, condition merging: merging multiple query conditions into one complex query condition. For example, merging the query of enterprises with a credit rating of A and the query of enterprises with sales greater than 1 million into the query of enterprises with a credit rating of A and sales greater than 1 million.

[0097] For example, result merging: executing multiple queries separately and then merging the query results together. For example, querying the financial status of an enterprise and querying the credit rating of an enterprise, obtaining a financial statement and a credit rating report respectively, and then merging the two into a comprehensive enterprise information report.

[0098] For example, function merging: multiple query results are used as input, processed by a function, and the final query result is obtained. For example, query the sales of a company and query the cost of the company, and then calculate the profit function to obtain the profit of the company.

[0099] 322. When designing a merge query, optimize the query statement. Methods for optimizing the query statement include avoiding repeated queries, using indexes, and limiting the query result set.

[0100] For example, avoid repeated queries: If multiple queries involve the same data fields, try to avoid repeated queries on these fields. For example, you can use subqueries or join queries to reduce the number of times data is read.

[0101] For example, use indexes: Creating indexes for frequently queried fields can speed up queries, reduce query time, and reduce the introduction of noise.

[0102] For example, limit the query result set: If you only need to query part of the data, you can use the LIMIT statement to limit the size of the query result set to reduce the time of reading and processing data.

[0103] 330. Modify the query interface, process query results, and monitor query performance.

[0104] 331. According to the designed merged query, modify the data query interface so that the query interface accepts the merged query request.

[0105] For example, if a database query language is used, the SQL statement or stored procedure may need to be modified; if an API is used for query, the parameters and return values ​​of the API may need to be modified.

[0106] 332. After executing the merge query, process the query results; if the result merge strategy is used, merge and organize multiple query results; if the function merge strategy is used, call the corresponding function for processing.

[0107] 333. After implementing the merge query, use database performance monitoring tools or log analysis tools to monitor query performance, where query performance includes query execution time, CPU usage, and memory usage.

[0108] Specifically, after implementing merged queries, you need to monitor query performance to ensure that merged queries do not have a significant impact on system performance. You can use database performance monitoring tools or log analysis tools to monitor query execution time, CPU usage, memory usage, and other indicators. If you find that query performance has degraded, you can further optimize the query statement or adjust the merge strategy.

[0109] As described above, the embodiment of the present application combines multiple queries into one query, reduces the frequency of noise introduction, can improve the comparability between different query results, and allows more accurate cross-query analysis, thereby helping enterprises make better decisions.

[0110] 400. Preprocess the enterprise credit data after noise processing to protect the privacy of the enterprise credit data through the differential privacy algorithm introduced by noise.

[0111] 410. Clarify the source of enterprise credit data and conduct preliminary review of enterprise credit data.

[0112] 411. Determine the source channels of corporate credit data and understand the methods and time frame for data collection.

[0113] Specifically, determine the source channels of corporate credit data, such as internal databases, external data providers, questionnaires, etc. Understand the methods and time frame of data collection to better understand the characteristics and potential problems of the data.

[0114] 412. Conduct a preliminary review of the collected data to understand the data format, field meaning, and data type.

[0115] Specifically, conduct a preliminary review of the collected data to understand the data format, field meaning, data type, etc., check whether the data is complete, and whether there are problems such as missing values ​​and duplicate values.

[0116] 420. Detect outliers in corporate credit data.

[0117] 421. Calculate the basic statistics of the data. Through the basic statistics, understand the distribution of the data and discover the outliers. Specifically, we calculate basic statistics of the data, such as mean, median, standard deviation, quartiles, etc. Through these statistics, we can get a preliminary understanding of the distribution of the data and find possible outliers. For example, if a data point deviates from the mean by more than a certain multiple of the standard deviation, or is located at the extreme position of the data distribution, it may be considered an outlier.

[0118] 422. Use visualization tools to display the distribution of data. Specifically, use visualization tools such as histograms, box plots, scatter plots, etc. to intuitively display the distribution of data. Through visual analysis, it is easier to find outliers and patterns in the data. For example, a box plot can display the median, quartiles, and outlier range of the data, so that possible outliers can be quickly identified.

[0119] 423. Use a model method to detect outliers, where the model method includes: cluster analysis and outlier detection algorithm.

[0120] Specifically, some model-based methods can be used to detect outliers, such as cluster analysis and outlier detection algorithms. These methods analyze the inherent structure and pattern of the data to find outliers that are different from most data points. For example, cluster analysis can divide data points into different groups, and outlier detection algorithms can identify outliers that are far away from other data points.

[0121] 430. Remove noise from corporate credit data.

[0122] 431. Use filtering methods to remove noise from corporate credit data, where filtering methods include mean filtering, median filtering, and Gaussian filtering; Specifically, filtering methods can be used to remove noise from data. Common filtering methods include mean filtering, median filtering, Gaussian filtering, etc. Among them, mean filtering replaces the value of the current data point with the average value of the surrounding data points, median filtering replaces the value of the current data point with the median value of the surrounding data points, and Gaussian filtering performs weighted averaging of the surrounding data points according to Gaussian distribution.

[0123] 432. If there is noise in the data that is correlated with other variables, use regression analysis to remove the noise; Specifically, if there is some noise in the data that is related to other variables, regression analysis can be used to remove the noise. By building a regression model between the data and other related variables, the true value of the data can be predicted, thereby removing the impact of the noise. For example, if an indicator in the corporate credit data is related to factors such as the size and industry of the enterprise, a regression model can be built to predict the true value of the indicator and remove noise that is unrelated to these factors.

[0124] 433. For time series data or data with a continuously changing trend, use data smoothing methods to remove noise, where data smoothing methods include moving average method and exponential smoothing method.

[0125] Specifically, for time series data or data with a continuous changing trend, data smoothing methods can be used to remove noise. Common data smoothing methods include moving average method, exponential smoothing method, etc. Among them, the moving average method replaces the value of the current data point with the average value of the data within a certain period of time, and the exponential smoothing method performs a weighted average of historical data, giving a higher weight to recent data.

[0126] As described above, the embodiments of the present application improve the quality of raw data by removing outliers and noise through data cleaning and preprocessing, which directly affects the accuracy and effectiveness of the final analysis results, thereby improving support for business decisions.

[0127] Thus, the embodiments of the present application analyze the query type of enterprise credit data and adaptively adjust the noise parameters; selectively increase noise based on the improved noise mechanism; merge multiple data queries into one data query; pre-process the enterprise credit data after noise processing to protect the privacy of the enterprise credit data through the differential privacy algorithm introduced by noise; flexibly adjust the noise size according to the specific circumstances of the query, which can significantly improve the accuracy of the results while protecting privacy, reduce unnecessary noise interference through selective noise addition, and improve the performance of data query and analysis. At the same time, multiple queries are merged into one query to reduce the frequency of noise introduction, which can improve the comparability between different query results.

[0128] The data accuracy of the embodiment of the present application is improved: through dynamic noise adjustment, the noise level can be flexibly adjusted according to the specific situation of the query, which can significantly improve the accuracy of the results while protecting privacy. For example, the noise is increased for high-sensitivity queries and the noise is reduced for low-sensitivity queries, so that users can obtain more reliable data results.

[0129] Performance optimization, through improved noise mechanism, selective noise addition, reducing unnecessary noise interference, improving data query and analysis performance. In this way, the speed and efficiency of data processing will be significantly improved, especially in scenarios that require real-time analysis.

[0130] Enhanced comparability: Through joint analysis, multiple queries are merged into one query, reducing the frequency of noise introduction, which can improve the comparability between different query results and allow more accurate cross-query analysis, thereby helping enterprises make better decisions.

[0131] Improving data quality by removing outliers and noise through data cleaning and preprocessing can improve the quality of raw data, directly affecting the accuracy and effectiveness of the final analysis results, thereby improving support for business decisions.

[0132] The above steps are not to be performed in a strict order as described in the numbers, but should be understood as an overall solution.

[0133] In the second aspect, based on the above embodiments, Figure 2 A schematic diagram of the structure of a differential privacy algorithm optimization device for protecting the privacy of enterprise credit data provided in an embodiment of the present application. Figure 2 The differential privacy algorithm optimization device for protecting the privacy of enterprise credit data provided in this embodiment specifically includes: a query analysis module 201, a noise adjustment module 202, a query merging module 203 and a data processing module 204.

[0134] Among them, the query analysis module 201 is used to analyze the query type of enterprise credit data and adaptively adjust the noise parameters; the noise adjustment module 202 is used to selectively increase noise based on the improved noise mechanism; the query merging module 203 is used to merge multiple data queries into one data query; the data processing module 204 is used to pre-process the enterprise credit data after noise processing, so as to protect the privacy of the enterprise credit data through the differential privacy algorithm introduced by noise.

[0135] As described above, the embodiments of the present application analyze the query type of enterprise credit data and adaptively adjust the noise parameters; selectively increase noise based on the improved noise mechanism; merge multiple data queries into one data query; pre-process the enterprise credit data after noise processing to protect the privacy of the enterprise credit data through the differential privacy algorithm introduced by noise; flexibly adjust the noise size according to the specific circumstances of the query, which can significantly improve the accuracy of the results while protecting privacy, reduce unnecessary noise interference through selective noise addition, and improve the performance of data query and analysis. At the same time, merge multiple queries into one query to reduce the frequency of noise introduction, which can improve the comparability between different query results.

[0136] The differential privacy algorithm optimization device for protecting the privacy of enterprise credit data provided in the embodiment of the present application can be used to execute the differential privacy algorithm optimization method for protecting the privacy of enterprise credit data provided in the above embodiment, and has corresponding functions and beneficial effects.

[0137] In a third aspect, an embodiment of the present application further provides an electronic device that can integrate the differential privacy algorithm optimization device for protecting the privacy of enterprise credit data provided by an embodiment of the present application. Figure 3 Schematic diagram of the structure of an electronic device provided by an embodiment of the present application. Figure 3 The electronic device includes: an input device 33, an output device 34, a memory 32 and one or more processors 31; the memory 32 is used to store one or more programs; when the one or more programs are executed by the one or more processors 31, the one or more processors 31 implement the differential privacy algorithm optimization method for protecting the privacy of enterprise credit data as provided in the above embodiment. The input device 33, the output device 34, the memory 32 and the processor 31 can be connected via a bus or other means. Figure 3 The example of connecting through bus is taken in the following.

[0138] The processor 31 executes various functional applications and data processing of the device by running software programs, instructions and modules stored in the memory 32, that is, implements the differential privacy algorithm optimization method for protecting the privacy of enterprise credit data mentioned above.

[0139] The electronic device provided above can be used to execute the differential privacy algorithm optimization method for protecting the privacy of enterprise credit data provided in the above embodiment, and has corresponding functions and beneficial effects.

[0140] In a fourth aspect, an embodiment of the present application also provides a computer-readable storage medium, which includes a stored computer program; wherein, when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the differential privacy algorithm optimization method for protecting the privacy of enterprise credit data as described above, and can achieve the same beneficial effects.

[0141] Of course, the storage medium containing computer executable instructions provided in the embodiment of the present application, whose computer executable instructions are not limited to the differential privacy algorithm optimization method for protecting the privacy of enterprise credit data as described above, can also execute related operations in the differential privacy algorithm optimization method for protecting the privacy of enterprise credit data provided in any embodiment of the present application.

[0142] In a fifth aspect, the embodiments of the present application also provide a computer program product, and the methods described in the various embodiments of the present application can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer programs or instructions. When the computer program or instruction is loaded and executed on a computer, the processes or functions described in the various embodiments of the present application are executed in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, a network device, a user device, a core network device, an OAM (Open Application Model) or other programmable device.

[0143] The computer program or instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer program or instructions may be transmitted from one website, computer, server or data center to another website, computer, server or data center by wired or wireless means. The computer-readable storage medium may be any available medium that can be accessed by a computer or a data storage device such as a server or data center that integrates one or more available media. The available medium may be a magnetic medium, such as a floppy disk, a hard disk, or a magnetic tape; it may also be an optical medium, such as a digital video disk; it may also be a semiconductor medium, such as a solid-state drive. The computer-readable storage medium may be a volatile or non-volatile storage medium, or may include both volatile and non-volatile types of storage media.

[0144] In several embodiments provided in the present application, it should be understood that the disclosed devices and methods can also be implemented in other ways. The device embodiments described above are merely schematic. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architecture, functions and operations of the devices, methods and computer program products according to multiple embodiments 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 a code, and the module, a program segment or a part of a code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of boxes in the block diagram and / or flowchart can be implemented with a dedicated hardware-based system that performs a specified function or action, or can be implemented with a combination of dedicated hardware and computer instructions.

[0145] In addition, the functional modules in the various embodiments of the present application may be integrated together to form an independent part, or each module may exist separately, or two or more modules may be integrated to form an independent part.

[0146] If the function is implemented in the form of a software function module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for enabling an electronic device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, and other media that can store program codes.

[0147] The above description is only an embodiment of the present application and is not intended to limit the scope of protection of the present application. For those skilled in the art, the present application may have various changes and variations. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application should be included in the scope of protection of the present application. It should be noted that similar reference numerals and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in the subsequent drawings.

[0148] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0149] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

Claims

1. A differential privacy algorithm optimization method for protecting the privacy of enterprise credit data, characterized in that: The method comprises: Analyze the query type of corporate credit data and adaptively adjust noise parameters; Selective noise addition based on improved noise mechanism; Merge multiple data queries into one data query; The enterprise credit data after noise processing is preprocessed to protect the privacy of the enterprise credit data through the differential privacy algorithm introduced by noise.

2. The differential privacy algorithm optimization method for protecting the privacy of enterprise credit data according to claim 1 is characterized in that: The query type of analyzing the enterprise credit data and adaptively adjusting the noise parameters include: Analyze the query types of corporate credit data and assess the sensitivity of corporate credit data; Determining influencing factors of the noise parameter, wherein the influencing factors include an allocated privacy budget, a sensitivity measure of the query, and a distribution characteristic of the data; Set noise parameters, dynamically adjust noise parameters, and adjust noise parameters according to application scenarios; Verify the settings of noise parameters and continuously optimize the noise parameters.

3. The differential privacy algorithm optimization method for protecting the privacy of enterprise credit data according to claim 1 is characterized in that: The method of selectively increasing noise based on an improved noise mechanism includes: Classify and standardize corporate credit data; Select the set noise type to add and determine the noise parameters; Determine the amount of noise based on data sensitivity and query type; Generate random noise and add the random noise to the data to update the data; Verify and evaluate the data after adding noise.

4. The differential privacy algorithm optimization method for protecting the privacy of enterprise credit data according to claim 1 is characterized in that: The step of merging the multiple data queries into one data query includes: Analyze query requirements, determine query relevance, and merge multiple data queries into one data query; Select a merge strategy and optimize query statements; Modify the query interface, process query results, and monitor query performance.

5. The differential privacy algorithm optimization method for protecting the privacy of enterprise credit data according to claim 1 is characterized in that: The preprocessing of the enterprise credit data after noise processing includes: Clarify the source of corporate credit data and conduct a preliminary review of corporate credit data; Detect outliers in corporate credit data; Remove noise from corporate credit data.

6. The differential privacy algorithm optimization method for protecting the privacy of enterprise credit data according to claim 2 is characterized in that: The analysis of the query types of enterprise credit data and the assessment of the sensitivity of enterprise credit data include: Identify the types of queries that exist in the corporate credit data, where different query types have different data sensitivity and privacy requirements; Assess the sensitivity of corporate credit data, including the assessment of the company's financial status, default record, and credit rating; The factors affecting the determination of noise parameters include: Allocate a privacy budget to control the degree of added noise; Different shares of the privacy budget are allocated to different types of queries and data sensitivity; Establish a query sensitivity metric and determine the query sensitivity by setting a method, where the higher the sensitivity of the query, the larger the noise parameter added; Establish the distribution characteristics of corporate credit data and use statistical methods to analyze the distribution of data; The step of setting noise parameters, dynamically adjusting noise parameters, and adjusting noise parameters according to application scenarios includes: For different types of queries, formulate different noise parameter setting strategies; For range queries, the noise parameters are determined based on the query range and data distribution; for count queries, the noise parameters are set based on the total number of enterprises in the data set and the accuracy requirements of the query; for sum queries, the noise parameters are set based on the numerical range of the data and the importance of the query; Different noise parameters are set according to the sensitivity level of the data, wherein the sensitivity level includes a high sensitivity level, a medium sensitivity level and a low sensitivity level; Among them, the noise parameter set at the high sensitivity level is greater than the noise parameter set at the medium sensitivity level, and the noise parameter set at the medium sensitivity level is greater than the noise parameter set at the low sensitivity level; Regularly re-evaluate the data, analyze changes in query types and data sensitivity, and adjust noise parameters based on the evaluation results; Adjust noise parameters according to application scenarios; The verification of the setting of the noise parameters and the continuous optimization of the noise parameters include: Select a set amount of corporate credit data as samples, simulate different types of queries, and add noise according to the set noise parameters; Compare the difference between the query results after adding noise and the real results, and evaluate the impact of noise parameters on the accuracy of query results; If the difference exceeds the set threshold, adjust the noise parameters; Continuously optimize the settings of noise parameters; Establish a feedback mechanism to collect user feedback on the accuracy of query results and the degree of privacy protection; Based on feedback and actual application, the noise parameter setting strategy is continuously adjusted.

7. The differential privacy algorithm optimization method for protecting the privacy of enterprise credit data according to claim 3 is characterized in that: The classification and standardization of enterprise credit data includes: Classify corporate credit data and divide the data into different categories according to the sensitivity of the data and the type of query; Standardize the classified data; The step of selecting and adding a set noise type and determining noise parameters includes: Select the set noise type to add; Determine the noise parameters according to the sensitivity of the data and the query type; the noise parameters include the mean, variance and standard deviation of the noise; The step of generating random noise and adding the random noise to data to update the data includes: Generate random noise using a random number generator and add the generated random noise to the data; Update the data with random noise added to the database to replace the original data; The verification and evaluation of the data after adding noise includes: Use data validation tools to validate the data after adding random noise; Conduct privacy assessment on data after adding random noise; Performance evaluation is performed on data after adding random noise.

8. The differential privacy algorithm optimization method for protecting the privacy of enterprise credit data according to claim 4 is characterized in that: The analyzing query requirements, determining query relevance, and combining multiple data queries into one data query includes: Obtain usage scenarios and business requirements for corporate credit data and determine query relevance; Analyze the correlation between different queries; If multiple queries involve the same data fields or the similarity of query conditions exceeds the set threshold, they are merged into one query; The step of selecting a merge strategy and optimizing a query statement includes: Select a merge strategy based on the query relevance and business needs. The merge strategies include: conditional merge, result merge, and function merge. When designing a merge query, optimize the query statement, wherein the methods of optimizing the query statement include avoiding repeated queries, using indexes, and limiting the query result set; The modifying of the query interface, processing of the query results, and monitoring of the query performance include: According to the designed merged query, modify the data query interface so that the query interface accepts the merged query request; After executing the merged query, process the query results; If the result merging strategy is used, multiple query results will be merged and sorted; If the function merging strategy is used, call the corresponding function for processing; After implementing the merged query, use a database performance monitoring tool or a log analysis tool to monitor the query performance, where the query performance includes the query execution time, CPU usage, and memory usage.

9. The differential privacy algorithm optimization method for protecting the privacy of enterprise credit data according to claim 5 is characterized in that: The aforementioned clarification of the source of corporate credit data and preliminary review of corporate credit data include: Determine the source channels of corporate credit data and understand the methods and time frame for data collection; Conduct a preliminary review of the collected data to understand the format, field meaning, and data type; The outlier detection of enterprise credit data includes: Calculate the basic statistics of the data, understand the distribution of the data through basic statistics, and find outliers; Use visualization tools to display the distribution of data; Using model methods to detect outliers, where the model methods include: cluster analysis, outlier detection algorithms; Wherein, removing noise from the enterprise credit data includes: Using filtering methods to remove noise from corporate credit data, where filtering methods include mean filtering, median filtering, and Gaussian filtering; If there is noise in the data that is correlated with other variables, use regression analysis to remove the noise; For time series data or data with a continuously changing trend, data smoothing methods are used to remove noise, where data smoothing methods include moving average method and exponential smoothing method.

10. A differential privacy algorithm optimization device for protecting the privacy of enterprise credit data, characterized in that: include: The query analysis module is used to analyze the query type of enterprise credit data and adaptively adjust the noise parameters; A noise adjustment module for selectively increasing noise based on an improved noise mechanism; A query merging module, used to merge multiple data queries into one data query; The data processing module is used to pre-process the enterprise credit data after noise processing so as to protect the privacy of the enterprise credit data through the differential privacy algorithm introduced by noise.

Images