Data usage auditing process method and system based on trusted data space

Abstract

The application discloses a data use auditing processing method and system based on a trusted data space, and relates to the technical field of data processing. The method comprises the following steps: acquiring an audit computing power pool for data use auditing, setting an auditing constraint, and distributing a plurality of first audit computing powers for auditing the data use of a plurality of users in the trusted data space; acquiring a plurality of user characteristics of the plurality of users, combining the plurality of first audit computing powers, performing privacy loss analysis and auditing security analysis, obtaining a first privacy loss parameter and a first auditing security parameter, and calculating the first auditing fitness of the plurality of first audit computing powers; performing iterative optimization of the audit computing power of the plurality of users to obtain a plurality of optimal audit computing powers, wherein the forward optimization parameter and the negative optimization parameter are configured according to the first auditing security parameter and the first privacy loss parameter, and optimization is performed; and the plurality of optimal audit computing powers are used to perform data use auditing processing on the plurality of users respectively. The application effectively improves the privacy of user data.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, in particular to a data usage audit processing method and system based on a trusted data space. BACKGROUND

[0002] With the development of digital economy, as a key carrier to guarantee data security sharing and compliance use, the trusted data space has been widely used in various industries. In the prior art, data usage audit is carried out through fixed computing power allocation or simple pre-configuration, relying on basic audit rules to check user data usage behavior, providing preliminary guarantee for data compliance.

[0003] However, the existing data usage audit technology has the problem of insufficient privacy. Although too frequent or intensive audit can improve security, it itself will infringe on user privacy due to frequent data retrieval and analysis, leading to loss of user trust in the data space, thereby affecting the audit effect. SUMMARY

[0004] The present application provides a data usage audit processing method and system based on a trusted data space, aiming to solve the technical problem of insufficient user data privacy in the prior art.

[0005] In view of the above problems, the present application provides a data usage audit processing method and system based on a trusted data space.

[0006] In a first aspect, the present application provides a data usage audit processing method based on a trusted data space, comprising:

[0007] Obtaining an audit computing power pool for data usage audit, and setting audit constraints, and distributing a plurality of first audit computing powers for auditing a plurality of user data usage in the trusted data space;

[0008] Obtaining a plurality of user features of a plurality of users, combining the plurality of first audit computing powers, and performing privacy loss analysis and audit security analysis to obtain a first privacy loss parameter and a first audit security parameter, and calculating a first audit fitness of the plurality of first audit computing powers;

[0009] Performing iterative optimization of the audit computing power of the plurality of users to obtain a plurality of optimal audit computing powers, wherein the forward optimization parameter and the negative optimization parameter are configured according to the first audit security parameter and the first privacy loss parameter, and optimization is performed;

[0010] Using the plurality of optimal audit computing powers to perform data usage audit processing on the plurality of users respectively.

[0011] In a second aspect, the present application provides a data usage audit processing system based on a trusted data space, comprising:

[0012] The computing power allocation module is configured to obtain an audit computing power pool of data usage auditing, and set an audit constraint to allocate a plurality of first audit computing powers for auditing a plurality of user data usages in a trusted data space.

[0013] The audit evaluation module is configured to obtain a plurality of user features of the plurality of users, combine the plurality of first audit computing powers, perform privacy loss analysis and audit security analysis, obtain a first privacy loss parameter and a first audit security parameter, and calculate a first audit fitness of the plurality of first audit computing powers.

[0014] The iterative optimization module is configured to perform iterative optimization of the plurality of users to obtain a plurality of optimal audit computing powers, wherein the forward optimization parameter and the negative optimization parameter are configured according to the first audit security parameter and the first privacy loss parameter, and optimization is performed.

[0015] The audit execution module is configured to perform data usage auditing processing on the plurality of users respectively by using the plurality of optimal audit computing powers.

[0016] One or more technical solutions provided in the present application have at least the following technical effects or advantages:

[0017] The present application provides a data usage auditing processing method and system based on a trusted data space, which dynamically allocates initial audit computing power, sets resource constraints to avoid excessive auditing, quantitatively analyzes privacy loss and audit security, establishes a comprehensive evaluation system, iteratively optimizes audit computing power allocation based on quantitative indicators, seeks an optimal solution for minimizing privacy loss, and performs optimal computing power configuration to achieve precise auditing. Finally, under the premise of ensuring necessary security supervision, the audit intensity and privacy impact are intelligently balanced, thereby effectively improving the privacy of user data. BRIEF DESCRIPTION OF DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings needed in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0019] Figure 1 A flowchart of a data usage auditing processing method based on a trusted data space provided by the embodiments of the present application is shown.

[0020] Figure 2 A structural schematic diagram of a data usage auditing processing system based on a trusted data space provided by the embodiments of the present application is shown.

[0021] In the drawings, the components represented by the numbers are described as follows:

[0022] The computing power allocation module 11, the audit evaluation module 12, the iterative optimization module 13, and the audit execution module 14. DETAILED DESCRIPTION

[0023] The application provides a data use audit processing method and system based on a trusted data space, which is used to solve the technical problem of insufficient user data privacy in the prior art.

[0024] The technical solutions in the embodiments of the application will be clearly and completely described below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, rather than all the embodiments of the application. Based on the embodiments in the application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the application.

[0025] It should be noted that the terms "comprising" and "having" are intended to cover non-exclusive inclusion, for example, a process, method, system, product or server comprising a series of steps or units need not be limited to only those steps or units clearly listed, but can include other steps or modules not clearly listed or inherent to the process, method, product or device.

[0026] In an embodiment, as shown in the accompanying drawings, the application provides a data use audit processing method based on a trusted data space, which comprises the following steps: Figure 1

[0027] S100: Obtain an audit computing power pool for data use audit, set audit constraints, and allocate a plurality of first audit computing powers for auditing a plurality of user data uses in a trusted data space.

[0028] In the embodiments of the application, an audit computing power pool for data use audit is obtained, audit constraints are set, and a plurality of first audit computing powers for auditing a plurality of user data uses in a trusted data space are allocated. Data use audit involves a large number of complex data analysis operations such as pattern recognition, anomaly detection and log tracing, which all need to consume significant computing resources. If these resource requirements are not uniformly managed and constrained, a plurality of concurrent audit tasks will compete for resources, leading to overall performance bottlenecks, and even affecting the normal operation of core business of the data space. Therefore, the abstract audit resource requirement must be quantified as a unified audit computing power pool, and initial allocation is made under this constraint to lay a foundation for subsequent fine optimization.

[0029] The step S100 in the method provided by the embodiments of the application comprises:

[0030] Obtaining an audit computing power pool for data use audit;

[0031] ​a total audit computing power for auditing data usage of the plurality of users is less than or equal to a total audit computing power of the audit computing power pool as an audit constraint;

[0032] a plurality of first audit computing powers of a plurality of users satisfying the audit constraint are randomly set in the audit computing power pool.

[0033] First, an audit computing power pool for data usage auditing is obtained. The audit computing power pool refers to a total computing resource quota that can be specially allocated by a trusted data space for performing all data usage auditing tasks within a specific time period. In this embodiment, the audit computing power pool is specifically embodied as a total bandwidth resource that can be used to perform all user data auditing tasks, representing the total capability of being able to process and analyze data flow within a unit time, and the audit computing power pool is normalized to a benchmark unit 1, representing the total auditing bandwidth that can be provided. By calling a resource management interface, a preset bandwidth quota configuration is read. For example, to guarantee core business, it is stipulated that the auditing service can occupy at most 10% of the total export bandwidth of the data center, and the 10% of bandwidth is abstractly quantified as 1 unit of the audit computing power pool, and the sum of bandwidth resources occupied by the auditing tasks of all users cannot exceed the normalized total capacity 1.

[0034] Second, a total audit computing power for auditing data usage of the plurality of users is less than or equal to a total audit computing power of the audit computing power pool as an audit constraint. The audit constraint is an optimized boundary condition, that is, the sum of the audit computing power distribution coefficients of all users cannot exceed 1. The sum of the plurality of first audit computing powers ≤ 1 is taken as the audit constraint, and is input into the audit resource control module as a verification basis for subsequent distribution results. For example, based on the audit computing power pool with a total distribution coefficient = 1, the audit constraint is set as: the sum of the bandwidth distribution coefficients of users A, B and C ≤ 1, to avoid the bandwidth resource distribution exceeding the total available amount.

[0035] Further, a plurality of first audit computing powers of a plurality of users satisfying the audit constraint are randomly set in the audit computing power pool.

[0036] The plurality of first audit computing powers of the plurality of users satisfying the audit constraint are randomly set in the audit computing power pool, including:

[0037] a plurality of first audit computing powers for auditing data usage of a plurality of users are randomly set in the audit computing power pool;

[0038] a first total audit computing power of the plurality of first audit computing powers is calculated, and it is judged whether the audit constraint is satisfied. If not, the audit computing power is re-set, and if yes, subsequent steps are performed.

[0039] Firstly, in the audit computing power pool, a plurality of first audit computing powers for auditing data use of a plurality of users are randomly set. The first audit computing power refers to an initial bandwidth allocation coefficient first allocated to each user. The larger the coefficient, the more audit bandwidth resources the user obtains. It is an initial benchmark for subsequent power optimization. Within the constraint that the sum of the allocation coefficients is less than or equal to 1, a random allocation algorithm is used to generate an initial bandwidth allocation coefficient for each user, i.e., the first audit computing power. For example, for three users A, B and C in the trusted data space, the first audit computing power is randomly allocated under the constraint that the sum of the allocation coefficients is less than or equal to 1, and the first allocation result is A: 0.4, B: 0.4 and C: 0.3.

[0040] Secondly, the first total audit computing power of the plurality of first audit computing powers is calculated, and it is determined whether the audit constraint is met. If not, the audit computing power is reset, and if yes, the subsequent step is performed. The first total audit computing power refers to the sum of the first audit computing powers of all users. By comparing the first total audit computing power with the audit constraint, if the sum is less than or equal to 1, the allocation result is determined, and if it exceeds, it is randomly allocated again until the constraint condition is met. For example, the first total audit computing power = 0.4 + 0.4 + 0.3 = 1.1 > 1, which exceeds the audit constraint; it is randomly allocated again, and the second allocation result is A: 0.3, B: 0.4 and C: 0.25, and the total audit computing power = 0.3 + 0.4 + 0.25 = 0.95 < 1, which meets the constraint condition, and the set of coefficients is determined as the final first audit computing power.

[0041] In the embodiments of the present application, by defining the audit computing power pool and the audit constraint, an explicit upper limit of resource consumption is set, which prevents resource overload caused by audit tasks and guarantees the overall stability of the trusted data space; by randomly generating and verifying the first audit computing power allocation scheme, an initial state that meets the global constraint condition and has randomness and exploratory is provided for the subsequent iterative optimization algorithm, which avoids the situation of being trapped in local optimum due to a fixed initial value.

[0042] S200: Obtain a plurality of user features of a plurality of users, combine the plurality of first audit computing powers, perform privacy loss analysis and audit security analysis, obtain a first privacy loss parameter and a first audit security parameter, and calculate a first audit fitness of the plurality of first audit computing powers.

[0043] In the embodiments of the present application, a plurality of user features of a plurality of users are obtained, privacy loss analysis and audit security analysis are performed in combination with a plurality of first audit computing forces, first privacy loss parameters and first audit security parameters are obtained, and first audit fitness of the plurality of first audit computing forces is calculated. The initial distribution of the first audit computing force has been completed in S100, but the data proportion difference of different users is not considered. The higher the user data proportion, the more sensitive data involved in the audit, and the privacy leakage risk and audit security demand are also different. In order to achieve the best balance between audit security and privacy protection, a set of quantitative evaluation system needs to be established, and the first audit fitness reflecting the comprehensive quality of the scheme is calculated through privacy loss analysis and audit security analysis, which is used as the explicit target of subsequent iterative optimization.

[0044] The method provided in the embodiments of the present application comprises the following step S200:

[0045] In the embodiments of the present application, a plurality of user features of a plurality of users are obtained, privacy loss analysis and audit security analysis are performed in combination with a plurality of first audit computing forces, first privacy loss parameters and first audit security parameters are obtained, and first audit fitness of the plurality of first audit computing forces is calculated. The initial distribution of the first audit computing force has been completed in S100, but the data proportion difference of different users is not considered. The higher the user data proportion, the more sensitive data involved in the audit, and the privacy leakage risk and audit security demand are also different. In order to achieve the best balance between audit security and privacy protection, a set of quantitative evaluation system needs to be established, and the first audit fitness reflecting the comprehensive quality of the scheme is calculated through privacy loss analysis and audit security analysis, which is used as the explicit target of subsequent iterative optimization.

[0046] A plurality of user features of a plurality of users are obtained, wherein each user feature comprises a user data proportion;

[0047] According to the plurality of first audit computing forces and the plurality of user features, a plurality of first user privacy loss parameters are calculated, a weight is assigned according to the size of the plurality of user features, and a first privacy loss parameter is obtained by weighted calculation, wherein the audit computing force and the user feature are positively correlated with the privacy loss parameter;

[0048] The plurality of user features and the plurality of first audit computing forces are input into an audit security analyzer, a plurality of user audit security parameters are output, and a first audit security parameter is obtained by weighted calculation.

[0049] First, a plurality of user features of a plurality of users are obtained, wherein each user feature comprises a user data proportion. The user feature is a key indicator for representing the inherent risk or data value of the user. The user feature mainly refers to the user data proportion, that is, the proportion of the data amount held by the user in the total amount of data in the trusted data space. The higher the data proportion of the user, the greater the overall impact that the data may cause if it is misused. Through the data statistical module of the trusted data space, the actual data storage amount, usage frequency and other data of each user are collected, the proportion of the total data amount is calculated, and a user feature set is formed. For example, continuing the three users A, B and C in S100, the user feature set is calculated by the data statistical module: the data proportion of user A is 0.3, the data proportion of user B is 0.4, and the data proportion of user C is 0.2, and the total proportion of the three is 0.9.

[0050] Secondly, according to the plurality of first audit computing power and the plurality of user characteristics, a plurality of first user privacy loss parameters are calculated, and the first privacy loss parameter is calculated by weighting according to the size of the plurality of user characteristics, wherein the audit computing power and the user characteristics are positively correlated with the privacy loss parameter. The privacy loss parameter is a quantitative value for measuring the overall risk level of user data privacy exposure caused by performing the audit task, and the value range is 0-1. The larger the privacy loss parameter, the higher the privacy risk. The audit computing power and the user characteristics are positively correlated with the privacy loss, and the user privacy loss parameter = the computing power allocation coefficient x the user data proportion. The data proportion of the user privacy loss is greater, so the user data proportion is weighted to sum the privacy loss parameters of all users to obtain the first privacy loss parameter. For example, it is known that the first audit computing power allocation coefficient: A = 0.3, B = 0.4, C = 0.25; the user data proportion: A = 0.3, B = 0.4, C = 0.2; the single user privacy loss parameter is calculated: A = 0.3 x 0.3 = 0.09, B = 0.4 x 0.4 = 0.16, C = 0.25 x 0.2 = 0.05; the first privacy loss parameter is calculated by weighting: 0.09 x 0.3 + 0.16 x 0.4 + 0.05 x 0.2 = 0.101.

[0051] Further, the plurality of user characteristics and the plurality of first audit computing power are input into the audit security analyzer, and the plurality of user audit security parameters are output, and the first audit security parameter is calculated by weighting.

[0052] The training process of the audit security analyzer comprises:

[0053] According to the audit log of different users in the trusted data space in the historical time, a sample user characteristic set and a sample audit computing power set are collected, and the proportion of user data loss of trust use under different sample user characteristics and sample audit computing power is collected as a sample audit risk parameter set, and a sample audit security parameter set is calculated;

[0054] Based on machine learning, an audit security analyzer is constructed;

[0055] The sample user characteristic set, the sample audit computing power set and the sample audit security parameter set are used as training data and test data to supervise the training and test of the audit security analyzer, and after the test loss converges, the trained audit security analyzer is obtained.

[0056] First, according to the audit logs of different users in the trusted data space within the historical time, a sample user feature set and a sample audit power set are collected, and the proportion of user data misuse under different sample user features and sample audit powers is collected as a sample audit risk parameter set, and a sample audit security parameter set is calculated. The sample user feature set refers to the user data proportion set reflecting the data properties of different users extracted from the historical audit logs. The sample audit power set refers to the audit bandwidth allocation coefficient set allocated to different users in the past extracted from the historical audit logs. The sample audit risk parameter refers to the proportion of user data misuse calculated for each sample user feature and sample audit power. The data misuse proportion refers to the proportion of misuse times to the total audit times, with a value range of 0-1. The sample audit security parameter set is a set of security quantitative values converted from the sample audit risk parameter set. The sample audit security parameter = 1 - sample audit risk parameter. The larger the value, the higher the audit security. Determine the data collection range: select the complete audit logs of the past 6-12 months in the trusted data space to ensure coverage of records of different user types and different data usage scenarios, and ensure sample diversity. Extract sample input data: filter the user data proportion from each log and aggregate to form the sample user feature set; simultaneously extract the audit bandwidth allocation coefficient allocated to the user in the corresponding log and aggregate to form the sample audit power set. Calculate the sample audit risk parameter: for each sample user feature and sample audit power, count the total audit times and data misuse times corresponding thereto; the data misuse proportion = misuse times / total audit times, and the sample audit risk parameter of each combination is calculated. Convert the sample audit security parameter: sample audit security parameter = 1 - sample audit risk parameter, convert all risk parameters to security parameters to form the sample audit security parameter set, which corresponds to the sample input data one by one.

[0057] For example, assuming that 1000 groups of valid samples are extracted from the historical logs, and the sample information of one group is as follows: log record: user data proportion 0.4, audit bandwidth allocation coefficient 0.3, total audit times 100, data misuse times 5; sample audit risk parameter = 5 / 100 = 0.05; sample audit security parameter = 1-0.05 = 0.95; finally, this group of samples forms a ternary group (sample user feature: 0.4, sample audit power: 0.3, sample audit security parameter: 0.95), and after aggregating the 1000 groups of sample data, the complete data set required for training the audit security analyzer is obtained.

[0058] Secondly, based on machine learning, an audit security analyzer is constructed. The machine learning model is a prediction model constructed based on a supervised learning algorithm, which is used to learn the mapping relationship between user features, audit capacity and audit security parameters. In combination with the demand for model interpretability and efficiency in the audit scene, a random forest model is used as the basic framework. The input layer sets 2 feature dimensions, which are bound to the user data proportion and the audit bandwidth allocation coefficient respectively, to ensure the matching of the input data format with the sample; the output layer sets 1 prediction dimension, which is bound to the audit security parameter, and the output value is limited to the range of 0-1 through the activation function. According to the model training experience, reasonable initial values of hyperparameters are set, such as 100 trees for the decision tree, 5 for the maximum tree depth, and 0.1 for the learning rate, to provide a basis for subsequent training. For example, the specific configuration of the constructed random forest model is as follows: model type: random forest regression model; input layer: 2 feature dimensions (x1: user data proportion, x2: audit bandwidth allocation coefficient); output layer: 1 prediction dimension (y: audit security parameter); hyperparameters: number of decision trees = 100, maximum tree depth = 5, minimum sample split number = 2, learning rate = 0.1; model mapping relationship: y = f(x1, x2), that is, through the integrated calculation of 100 decision trees, the input user features and audit capacity are mapped to the corresponding audit security parameter prediction value.

[0059] Further, the sample user feature set, sample audit capacity set and sample audit security parameter set are used as training data and test data to supervise the training and testing of the audit security analyzer. After the test loss converges, the trained audit security analyzer is obtained. The training data is a sample set used for model parameter learning, accounting for 70% of the total samples, through which the model learns the correlation between user features, audit capacity and security parameters. The test data is a sample set used to verify the model generalization ability, accounting for 30% of the total samples, which does not participate in model training and is only used to evaluate the prediction accuracy of the model on new data. The loss function is an index to measure the deviation between the model prediction value and the true value, and the mean square error (MSE) is selected, and the smaller the value, the higher the prediction accuracy. Convergence means that during the model training process, the loss value of the test data remains stable for several iterations, indicating that the model has learned the mapping rule and does not need to continue training.

[0060] For example, 1000 groups of samples are randomly sampled and divided in a ratio of 7:3, 70% as training data and 30% as test data, ensuring that the distribution characteristics of the two groups of data are consistent. The sample user data proportion and sample audit computing power in the training data are used as input, and the sample audit security parameter is used as the true label to input the constructed random forest model. The mean square error (MSE) is used as the loss function, and the decision tree parameters of the model are iteratively optimized through the back propagation algorithm. After each iteration, the model weight is updated. After completing 10 iterations, the test data is input into the model to calculate the loss value (MSE) of the test data, and the loss value trend is recorded. When the loss value of the test data remains stable for 3 consecutive iterations, such as loss value ≤ 0.01, and there is no obvious rebound, it is determined that the model training converges, the iteration is stopped, the current model parameters are saved, and the trained audit security analyzer is obtained. At this time, when the user features and audit computing power are input, the audit security analyzer can output the user audit security parameter.

[0061] Finally, multiple user features and multiple first audit computing powers are input into the audit security analyzer, and multiple user audit security parameters are output. The first audit security parameter is obtained by weighted calculation. According to the user data proportion, the weight is allocated, and the weighted average value is calculated as the first audit security parameter. For example, the user features and first audit computing powers of users A, B and C are input respectively: user A: user data proportion 0.3, first audit computing power 0.3; user B: user data proportion 0.4, first audit computing power 0.4; user C: user data proportion 0.2, first audit computing power 0.25. The audit security analyzer outputs the user audit security parameters of users A, B and C respectively: the audit security parameter of A = 0.85; the audit security parameter of B = 0.9; the audit security parameter of C = 0.8. The first audit security parameter = 0.85*0.3 + 0.9*0.4 + 0.8*0.2 = 0.775, which represents that the overall audit security level is 77.5% under the current computing power allocation.

[0062] The first audit fitness of the multiple first audit computing powers is calculated, including:

[0063] The first total audit computing power of the multiple first audit computing powers is calculated, and the first computing power fitness is calculated, wherein the first computing power fitness is negatively related to the first total audit computing power;

[0064] According to the first privacy loss parameter, the first privacy fitness is calculated;

[0065] The first audit security parameter is used as the first security fitness;

[0066] According to the first computing power fitness, the first privacy fitness and the first security fitness, the first audit fitness of the multiple first audit computing powers is calculated.

[0067] First, the first total audit computing power of the plurality of first audit computing powers is calculated, and the first computing power fitness is calculated, wherein the size of the first computing power fitness is negatively related to the first total audit computing power. The first computing power fitness is a quantitative value for evaluating the rationality of the first total audit computing power resource consumption, and the first computing power fitness = 1 - the first total audit computing power, with a value range of 0-1. For example, the first audit computing powers are A = 0.3, B = 0.4, and C = 0.25, the first total audit computing power = 0.3 + 0.4 + 0.25 = 0.95, and the first computing power fitness = 1 - 0.95 = 0.05. Since the first total audit computing power is close to 1, the resource consumption is more, and the first computing power fitness value is low. If the first total audit computing power is 0.5, the first computing power fitness is 0.5. The less the computing power consumption, the greater the computing power fitness.

[0068] Secondly, according to the first privacy loss parameter, the first privacy fitness is calculated. The first privacy fitness is a quantitative value for evaluating the overall privacy protection effect, and the first privacy fitness = 1 - the first privacy loss parameter, with a value range of 0-1. For example, the first privacy loss parameter is 0.101, and the first privacy fitness = 1 - 0.101 = 0.899. If the first privacy loss parameter is 0.5, the first privacy fitness is 0.5. The greater the privacy loss parameter, the worse the privacy protection effect, and the smaller the privacy fitness.

[0069] In addition, the first audit security parameter is taken as the first security fitness. The first security fitness reflects the overall audit security level. The greater the value, the safer the audit, and the lower the data loss risk. For example, according to S200, the first audit security parameter = 0.775, i.e. the first security fitness = 0.775.

[0070] Finally, according to the first computing power fitness, the first privacy fitness, and the first security fitness, the first audit fitness of the plurality of first audit computing powers is calculated. The first audit fitness is the final comprehensive evaluation score, which is the weighted sum of the fitness of the three dimensions of computing power, privacy, and security. Appropriate weights are allocated to the first computing power fitness, the first privacy fitness, and the first security fitness to reflect the emphasis degree of the method on different targets, and then the weighted sum is calculated to obtain the first audit fitness. For example, the first computing power fitness, the first privacy fitness, and the first security fitness are each allocated a weight of 1 / 3, and then the first audit fitness = (0.05 + 0.899 + 0.775) / 3 ≈ 0.5747.

[0071] In the embodiments of the present application, the abstract audit effect is converted into precise calculation indexes by constructing a quantitative evaluation system. First, the privacy loss and security level brought by the audit behavior are quantitatively analyzed, and then the three goals of computing power consumption, privacy protection and audit security are integrated to generate a unified audit fitness score, solving the internal conflict among multiple goals and providing a clear direction for subsequent optimization. At the same time, with the prediction ability of the audit security analyzer, a leap from static evaluation to dynamic intelligent decision-making is realized, laying a theoretical foundation for globally optimal resource allocation.

[0072] S300: Iterative optimization of audit computing power of multiple users is performed to obtain multiple optimal audit computing powers, wherein the positive and negative optimization parameters are configured according to the first audit security parameter and the first privacy loss parameter, and optimization is performed.

[0073] In the embodiments of the present application, iterative optimization of audit computing power of multiple users is performed to obtain multiple optimal audit computing powers, wherein the positive and negative optimization parameters are configured according to the first audit security parameter and the first privacy loss parameter, and optimization is performed. S200 has quantified the comprehensive rationality of the initial computing power allocation through the first audit fitness, but the initial allocation has not been dynamically adjusted in combination with the historical audit benchmark and the current parameter deviation, and there is a problem of imbalance in the adaptation of resource conservation, privacy protection and audit security. Therefore, optimization parameters need to be configured based on the deviation of the first audit security parameter, the first privacy loss parameter and the historical benchmark, and iterative adjustment is carried out to ultimately obtain optimal audit computing power that adapts to the needs of multiple users.

[0074] The method provided in the embodiments of the present application includes step S300:

[0075] Obtain the average audit security parameter and the average privacy loss parameter of the audit processing in the historical time as the benchmark audit security parameter and the benchmark privacy loss parameter;

[0076] Calculate the ratio of the first audit security parameter to the benchmark audit security parameter to obtain a positive optimization parameter, and calculate the ratio of the first privacy loss parameter to the benchmark privacy loss parameter to obtain a negative optimization parameter, wherein the positive and negative optimization parameters include a positive optimization quantity and a negative optimization quantity;

[0077] According to the positive and negative optimization parameters, the multiple first audit computing powers are respectively positively adjusted and negatively adjusted to obtain multiple second positive audit computing power sets and multiple second negative audit computing power sets;

[0078] Randomly combine the multiple second positive audit computing power sets and the multiple second negative audit computing power sets to process and obtain an audit fitness of an audit computing power combination that satisfies the audit constraint, and multiple audit computing powers corresponding to the maximum audit fitness are taken as the basis for subsequent optimization adjustment;

[0079] The iterative optimization is continuously performed until convergence, and a plurality of optimal audit computing powers are obtained.

[0080] Firstly, the average audit security parameter and the average privacy loss parameter in the historical time for performing the audit processing are obtained as the benchmark audit security parameter and the benchmark privacy loss parameter. The benchmark audit security parameter refers to the average value of the audit security parameter in the trusted data space in the historical time, and the value range is 0-1, reflecting the average level of the past audit security. The benchmark privacy loss parameter refers to the average value of the privacy loss parameter in the trusted data space in the historical time, and the value range is 0-1, reflecting the average level of the past privacy protection. Through the audit log statistics module of the trusted data space, the audit security parameter and the privacy loss parameter of all audit records in the historical period are extracted, and the average values are calculated as the benchmark reference for this optimization. For example, the historical audit data of the past 6 months is extracted, and the average audit security parameter = 0.8 is calculated, i.e. the benchmark audit security parameter = 0.8; the average privacy loss parameter = 0.15, i.e. the benchmark privacy loss parameter = 0.15.

[0081] Secondly, the ratio of the first audit security parameter to the benchmark audit security parameter is calculated to obtain a positive optimization parameter, and the ratio of the first privacy loss parameter to the benchmark privacy loss parameter is calculated to obtain a negative optimization parameter, wherein the positive optimization parameter and the negative optimization parameter include a positive optimization quantity and a negative optimization quantity. The positive optimization parameter reflects the improvement space of the audit security level, and includes the positive optimization quantity, i.e. the number of users requiring positive adjustment of computing power. The positive optimization parameter = the first audit security parameter / the benchmark audit security parameter. If the positive optimization parameter < 1, it means that the current audit security is lower than the historical average, and part of the user computing power needs to be adjusted positively. The negative optimization parameter reflects the improvement demand of the privacy loss, and includes the negative optimization quantity, i.e. the number of users requiring negative adjustment of computing power. The negative optimization parameter = the first privacy loss parameter / the benchmark privacy loss parameter. If the negative optimization parameter > 1, it means that the current privacy loss is higher than the historical average, and part of the user computing power needs to be adjusted negatively. For example, it is known that the first audit security parameter = 0.775, the benchmark audit security parameter = 0.8; the first privacy loss parameter = 0.101, the benchmark privacy loss parameter = 0.15, then the positive optimization parameter = 0.775 / 0.8≈0.969, and the negative optimization parameter = 0.101 / 0.15≈0.673.

[0082] Further, according to the positive optimization parameter and the negative optimization parameter, the plurality of first audit computing powers are respectively positively adjusted and negatively adjusted to obtain a plurality of second positive audit computing power sets and a plurality of second negative audit computing power sets.

[0083] The first plurality of computing powers are respectively adjusted in the positive direction and in the negative direction according to the positive optimization parameter and the negative optimization parameter, to obtain a plurality of second positive audit computing power sets and a plurality of second negative audit computing power sets, including:

[0084] A preset adjustment quantity is obtained.

[0085] The positive adjustment quantity and the negative adjustment quantity are calculated and processed according to the positive optimization parameter, the negative optimization parameter and the preset adjustment quantity.

[0086] The first plurality of computing powers are respectively adjusted in the positive direction and in the negative direction according to the positive adjustment quantity and the negative adjustment quantity, to obtain a plurality of second positive audit computing power sets and a plurality of second negative audit computing power sets.

[0087] First, a preset adjustment quantity is obtained. The preset adjustment quantity is a basic magnitude set for the adjustment range of this iteration, used to control the search range and precision of the algorithm in the solution space. A fixed value is read from the algorithm configuration parameter, which is usually set according to experience to balance the breadth and efficiency of the search. For example, the preset adjustment quantity is 10, and the number of new solutions to be generated in this round will be calculated based on this quantity.

[0088] Second, the positive adjustment quantity and the negative adjustment quantity are calculated and processed according to the positive optimization parameter, the negative optimization parameter and the preset adjustment quantity. The positive adjustment quantity is the number of new candidate values generated for each user based on the positive fine-tuning of the current computing power, and the calculation formula is: positive adjustment quantity = round (preset adjustment quantity x positive optimization parameter). The negative adjustment quantity is the number of new candidate values generated for each user based on the negative fine-tuning of the current computing power, and the calculation formula is: negative adjustment quantity = round (preset adjustment quantity x negative optimization parameter). For example, the positive adjustment quantity = round (10 x 0.969) = 10; the negative adjustment quantity = round (10 x 0.673) = 7. The positive optimization parameter 0.969 is closer to 1, and the audit security needs to be optimized, so the number of new solutions in the positive direction is consistent with the preset; the negative optimization parameter 0.673 is smaller, and the privacy protection is better than the benchmark, so the number of new solutions in the negative direction is smaller, avoiding excessive adjustment.

[0089] Again, according to the positive adjustment quantity and the negative adjustment quantity, the plurality of first audit computing powers are respectively positively adjusted and negatively adjusted to obtain a plurality of second positive audit computing power sets and a plurality of second negative audit computing power sets. The second positive audit computing power set refers to a set of candidate values obtained by positively fine-tuning the current computing power of a single user. The second negative audit computing power set refers to a set of candidate values obtained by negatively fine-tuning the current computing power of a single user. By fine-tuning, the computing power neighborhood of each user is explored. Taking user B as an example: the first audit computing power = 0.4, the second positive audit computing power set is generated, containing 11 values: increasing the step 0.05 based on 0.4. To introduce diversity, a small random disturbance of ±0.01 can be added to the step. For example: [0.449, 0.451, 0.446, 0.454, 0.448, 0.452, 0.447, 0.453, 0.455, 0.445]; the second negative audit computing power set is generated, containing 7 values: decreasing the step 0.05 based on 0.4, and adding a small random disturbance. For example: [0.351, 0.349, 0.354, 0.348, 0.352, 0.347, 0.353]. Similarly, the same operation is performed on user A and user C: 10 positive candidate values around 0.35 are generated for user A, and 7 negative candidate values around 0.25 are generated for user A. 10 positive candidate values around 0.30 are generated for user C, and 7 negative candidate values around 0.20 are generated for user C.

[0090] Then, the plurality of second positive audit computing power sets and the plurality of second negative audit computing power sets are randomly combined, the audit fitness of the audit computing power combination satisfying the audit constraint is obtained, and the plurality of audit computing powers corresponding to the maximum audit fitness are taken as the basis for subsequent optimization adjustment. By random combination, the global solution space composed of individual computing power fine-tuning is explored on a large scale to find a new allocation scheme, and then the feasible candidate scheme is screened out and its comprehensive quality is evaluated to find the optimal one. From the second positive audit computing power set and the plurality of second negative audit computing power sets of each user, one computing power value is randomly selected from each set, and these selected values are combined together to form a complete new generation of audit computing power allocation scheme. Repeat this process to generate a candidate pool containing a plurality of schemes. Constraint verification: traverse each scheme in the candidate pool, calculate the total audit computing power, eliminate all schemes that do not satisfy the total audit computing power ≤1 audit constraint, and the remaining schemes that satisfy the audit constraint are the effective pool. For each combination in the effective combination pool, the audit fitness calculation logic of S200 is followed to calculate and record the comprehensive fitness of each combination one by one. Compare all the fitness values in the effective combination pool, and select the computing power combination corresponding to the maximum value as the optimal computing power of this round; if there are multiple combinations with the same maximum fitness value, select the combination with the total computing power closer to 1.

[0091] For example, users A, B, and C have 10 positive candidate values and 7 negative candidate values, respectively. A random combination is made: 0.28 can be drawn from the negative set of A, 0.451 from the positive set of B, and 0.295 from the positive set of C, forming a candidate scheme (A: 0.28, B: 0.451, C: 0.295). The total computing power of the candidate scheme (A: 0.28, B: 0.451, C: 0.295) is 1.026, which is greater than 1, and thus is eliminated. Another candidate scheme (A: 0.3, B: 0.45, C: 0.25) has a total computing power of 0.9, which meets the constraint, and is subjected to S200 analysis, assuming that its audit fitness is calculated to be 0.572. Finally, 82 effective combinations are screened out to form an effective combination pool. By comparing the fitness of the 82 effective combinations, the fitness of the candidate scheme (A: 0.3, B: 0.35, C: 0.25) ≈ 0.594 is the maximum value, and the total computing power 0.9 is close to 1, so it is determined as the optimal computing power of this round, which is used as the initial computing power basis for the next iteration.

[0092] Further, the iteration optimization is continued until convergence, and multiple optimal audit computing powers are obtained. The optimal audit computing power refers to the user computing power distribution scheme that meets the audit constraint and has a stable maximum fitness after iteration convergence. The optimal computing power screened out for the first time is used as the initial computing power of the first round, the initial optimal fitness is recorded, and the fitness sequence and continuous convergence counter are initialized. The optimal computing power of the above round is used as the initial computing power, and the optimization parameters are repeatedly calculated, the adjusted computing power set is generated, the optimal combination is screened out, and the optimal computing power and fitness of this round are obtained. The difference between the fitness of this round and the last round is calculated, and if it is ≤0.01, the counter is incremented by 1, and if it is ≥3 times, the iteration is terminated; otherwise, the counter is reset and the iteration is continued. After the iteration is terminated, the optimal audit computing power is output.

[0093] For example, users A, B, and C have initial computing powers A=0.3, B=0.4, and C=0.25, and the convergence condition is that the difference is ≤0.01 for 3 consecutive rounds. Initialization: the initial computing powers of the first round are A=0.3, B=0.45, and C=0.25, and the fitness is 0.595; the second round: the optimal computing powers are A=0.35, B=0.4, and C=0.25, the fitness is 0.602, and the difference is 0.007≤0.01, so the counter is 1; the third round: the optimal computing powers are unchanged, the fitness is 0.605, and the difference is 0.003≤0.01, so the counter is 2; the fourth round: the fitness is 0.606, and the difference is 0.001≤0.01, so the counter is 3, which meets the convergence condition. The output optimal audit computing power is A=0.35, B=0.4, and C=0.25, the total computing power is 1.0, and the fitness is 0.606.

[0094] In the embodiments of the present application, through the closed-loop optimization mechanism of historical benchmark calibration, bidirectional adjustment solution generation, combination screening and iterative convergence, dynamic and accurate optimization of audit computing power is realized. Based on the deviation quantitative configuration of the first audit security parameter, the first privacy loss parameter and the historical benchmark, the positive / negative optimization parameters are configured, which breaks the limitation of static computing power allocation, and makes the optimization intensity and the actual adaptation demand accurate match; through the positive / negative computing power set generation and random combination screening, the solution space is fully explored, and invalid schemes are filtered through audit constraint verification, ensuring that the optimization direction covers the three dimensions of resources, privacy and security; relying on the quantitative convergence standard to ensure the stability of optimization, the finally output optimal audit computing power not only meets the audit constraints, but also realizes the optimal balance of resource saving, privacy protection and audit security, and provides accurate computing power support for subsequent efficient audit processing.

[0095] S400: Adopting the plurality of optimal audit computing powers, respectively performing data use audit processing on the plurality of users.

[0096] In the embodiments of the present application, the plurality of optimal audit computing powers are adopted to respectively perform data use audit processing on the plurality of users. S300 has output the optimal audit computing power adapted to the multi-user scenario, and the computing power is essentially a bandwidth resource allocation scheme matching user data attributes, privacy protection demand and audit security target. High computing power allocation corresponding to high-frequency audit is easy to lead to resource waste, and low computing power allocation corresponding to high-frequency audit is easy to cause audit delay. Therefore, it is necessary to establish a mapping relationship between the optimal audit computing power and the audit frequency through S400, to convert the optimized computing power resources into an efficient audit execution scheme, and to ensure that the audit intensity is highly adapted to resource allocation and user demand.

[0097] The method provided in the embodiments of the present application comprises the following steps S400:

[0098] According to the plurality of optimal audit computing powers, the audit frequency is set, and the data use audit processing is performed on the plurality of users.

[0099] First, set the audit frequency according to the optimal audit computing power. The audit frequency refers to the number of audits of user data usage behavior per unit time. The higher the frequency, the more intensive the audit monitoring, which can timely discover dishonest usage behavior, but needs to be matched with corresponding computing power support. The higher the optimal audit computing power, the higher the audit frequency that can be supported, ensuring that the computing power resources match the audit task volume. A mapping rule library is established to preset the corresponding relationship between the computing power interval and the audit frequency through historical audit test data; the optimal audit computing power of each user output by S300 is substituted into the mapping rule library to match the audit frequency of the corresponding user; and the sum of the audit frequencies of all users is ensured to be less than the maximum processing capacity of the audit system to avoid overload. For example, the optimal audit computing power: A = 0.35, B = 0.4, C = 0.25. The mapping rule library: 0.3-0.35: 2 times / hour; 0.35-0.45: 3 times / hour; 0.2-0.3: 1 time / hour. Frequency matching: A = 0.35, the corresponding audit frequency is 3 times / hour, B = 0.4, the corresponding audit frequency is 3 times / hour, C = 0.25, the corresponding audit frequency is 1 time / hour; verification: total frequency = 3 + 3 + 1 = 7 times / hour ≤ audit system upper limit 10 times / hour, matching is valid.

[0100] Second, data usage audit processing is performed on multiple users respectively. The data usage audit processing refers to a process of periodically collecting user's original data usage records according to the audit frequency, comparing with the preset compliance usage rules, and determining whether there is dishonest usage behavior. According to the matched audit frequency, a timed audit task is generated for each user. For example, A user is audited once every 20 minutes, B user is audited once every 20 minutes, and C user is audited once every 60 minutes. User data usage records are collected on time, audit rule engine is called, and compliance is determined by comparison; each audit result is written into an audit result log, and if it is determined to be dishonest, a real-time warning is triggered; log synchronization: the audit result log is synchronized to the data storage module at regular intervals to provide historical data support for parameter iteration of S100-S300.

[0101] In the embodiment of the application, through the quantitative mapping of the optimal audit computing power and the audit frequency, the problems of resource waste of high computing power and low frequency and audit delay of low computing power and high frequency are avoided, so that each unit of computing power can be converted into matched audit monitoring strength, and accurate adaptation of computing power and audit execution is realized. Different audit frequencies are matched for different users according to their optimal computing power, high-frequency audit is implemented for users with high data proportion and strong security requirements, and low-frequency audit is implemented for users with low data proportion, taking into account efficiency and cost.

[0102] The embodiment of the application achieves the following technical effects through the specific implementation manner described above:

[0103] The application provides a data use audit processing method and system based on a trusted data space. By constructing an audit computing power dynamic optimization distribution mechanism, the problem of balancing the security and privacy protection of data use audit in the trusted data space is effectively solved. First, a quantitative evaluation system is established to comprehensively model the audit security, privacy loss and computing power consumption. Then, an intelligent optimization algorithm is used to automatically optimize the best distribution scheme under the global computing power constraint. Through differential audit strategy execution, precise audit on demand is realized. Finally, while ensuring effective supervision of data use behavior, the privacy leakage risk caused by excessive audit is effectively reduced, and the utilization efficiency of audit resources is improved, providing an intelligent audit solution for the trusted data space that balances security, privacy and efficiency.

[0104] In an embodiment, as shown in Figure 2 The application provides a data use audit processing system based on a trusted data space, which comprises:

[0105] A computing power distribution module 11 is configured to obtain an audit computing power pool for data use audit, set an audit constraint, and distribute a plurality of first audit computing powers for auditing data use of a plurality of users in the trusted data space.

[0106] An audit evaluation module 12 is configured to obtain a plurality of user features of the plurality of users, analyze privacy loss and audit security in combination with the plurality of first audit computing powers, obtain a first privacy loss parameter and a first audit security parameter, and calculate a first audit fitness of the plurality of first audit computing powers.

[0107] An iterative optimization module 13 is configured to perform iterative optimization of the audit computing power of the plurality of users to obtain a plurality of optimal audit computing powers, wherein the forward optimization parameter and the negative optimization parameter are configured according to the first audit security parameter and the first privacy loss parameter, and optimization is performed.

[0108] An audit execution module 14 is configured to use the plurality of optimal audit computing powers to perform data use audit processing on the plurality of users respectively.

[0109] In an embodiment, the computing power distribution module 11 is further configured to:

[0110] obtain an audit computing power pool for data use audit;

[0111] set the total audit computing power for auditing data use of the plurality of users to be less than or equal to the total audit computing power of the audit computing power pool as the audit constraint;

[0112] randomly set a plurality of first audit computing powers of a plurality of users in the audit computing power pool to satisfy the audit constraint.

[0113] The first audit computing power of the plurality of users is randomly set in the audit computing power pool to meet the audit constraint, including:

[0114] The first audit computing power of the plurality of users is randomly set in the audit computing power pool to perform data usage auditing on the plurality of users;

[0115] The first total audit computing power of the plurality of first audit computing powers is calculated to determine whether the audit constraint is met, and if not, the audit computing power is re-set, and if yes, the subsequent step is performed.

[0116] In one embodiment, the audit evaluation module 12 is further configured to:

[0117] Obtain a plurality of user features of the plurality of users, wherein each user feature includes a user data proportion;

[0118] According to the plurality of first audit computing powers and the plurality of user features, a plurality of first user privacy loss parameters are calculated, a weight is allocated according to the size of the plurality of user features, and a first privacy loss parameter is calculated by weighting, wherein the audit computing power and the user feature are positively correlated with the privacy loss parameter;

[0119] The plurality of user features and the plurality of first audit computing powers are input into the audit security analyzer, and a plurality of user audit security parameters are output, and a first audit security parameter is calculated by weighting.

[0120] The training process of the audit security analyzer includes:

[0121] According to the audit log of different users in the trusted data space within the historical time, a sample user feature set and a sample audit computing power set are collected, and the proportion of users who appear data misuse under different sample user features and sample audit computing powers is collected as a sample audit risk parameter set, and a sample audit security parameter set is calculated;

[0122] Based on machine learning, an audit security analyzer is constructed;

[0123] The sample user feature set, the sample audit computing power set, and the sample audit security parameter set are used as training data and test data to supervise the training and test of the audit security analyzer, and after the test loss converges, a trained audit security analyzer is obtained.

[0124] The first audit fitness of the plurality of first audit computing powers is calculated, including:

[0125] The first total audit computing power of the plurality of first audit computing powers is calculated, and a first computing power fitness is calculated, wherein the size of the first computing power fitness is negatively correlated with the first total audit computing power;

[0126] According to the first privacy loss parameter, a first privacy fitness is calculated and obtained;

[0127] The first audit security parameter is taken as a first security fitness;

[0128] According to the first computing power fitness, the first privacy fitness and the first security fitness, a first audit fitness of a plurality of first audit computing powers is calculated and obtained.

[0129] In one embodiment, the iterative optimization module 13 is further configured to:

[0130] Obtain an average audit security parameter and an average privacy loss parameter in the historical time as a benchmark audit security parameter and a benchmark privacy loss parameter;

[0131] Calculate a ratio of the first audit security parameter and the benchmark audit security parameter to obtain a positive optimization parameter, and calculate a ratio of the first privacy loss parameter and the benchmark privacy loss parameter to obtain a negative optimization parameter, wherein the positive optimization parameter and the negative optimization parameter include a positive optimization number and a negative optimization number;

[0132] According to the positive optimization parameter and the negative optimization parameter, a plurality of second positive audit computing power sets and a plurality of second negative audit computing power sets are obtained by respectively performing positive adjustment and negative adjustment on the plurality of first audit computing powers;

[0133] The plurality of second positive audit computing power sets and the plurality of second negative audit computing power sets are randomly combined to obtain an audit fitness of an audit computing power combination satisfying the audit constraint, and a plurality of audit computing powers corresponding to a maximum audit fitness are taken as a basis for subsequent optimization adjustment;

[0134] The iterative optimization is continued until convergence, and a plurality of optimal audit computing powers are obtained.

[0135] According to the positive optimization parameter and the negative optimization parameter, a plurality of second positive audit computing power sets and a plurality of second negative audit computing power sets are obtained by respectively performing positive adjustment and negative adjustment on the plurality of first audit computing powers, including:

[0136] A preset adjustment number is obtained;

[0137] According to the positive optimization parameter, the negative optimization parameter and the preset adjustment number, a positive adjustment number and a negative adjustment number are calculated and obtained;

[0138] According to the positive adjustment number and the negative adjustment number, a plurality of second positive audit computing power sets and a plurality of second negative audit computing power sets are obtained by respectively performing positive adjustment and negative adjustment on the plurality of first audit computing powers.

[0139] In one embodiment, the audit execution module 14 is further configured to:

[0140] According to the optimal audit computing power, the audit frequency is set, and the data usage audit processing is performed on the plurality of users respectively.

[0141] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description, and does not represent the advantages and disadvantages of the embodiments. And the above describes the specific embodiments of the present application. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are also possible or can be advantageous.

[0142] The above only describes the preferred embodiments of the present application, and does not limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

[0143] The present application and the drawings are only exemplary descriptions of the present application, and are considered to cover any and all modifications, changes, combinations or equivalents within the scope of the present application. Obviously, those skilled in the art can make various modifications and changes to the present application without departing from the scope of the present application. Thus, if these modifications and changes of the present application belong to the scope of the present application and its equivalent technology, the present application intends to include these modifications and changes.

Claims

1. A data usage auditing process method based on a trusted data space, characterized in that, The method comprises: obtaining an audit computing power pool for data use auditing, setting an audit constraint, and allocating a plurality of first audit computing powers for auditing a plurality of user data uses in a trusted data space; obtaining a plurality of user features of a plurality of users, combining the plurality of first audit computing powers, performing privacy loss analysis and audit security analysis, obtaining a first privacy loss parameter and a first audit security parameter, and calculating a first audit fitness of the plurality of first audit computing powers, comprising: obtaining a plurality of user features of a plurality of users, wherein each user feature comprises a user data proportion; According to the plurality of first audit computing powers and the plurality of user features, a plurality of first user privacy loss parameters are calculated, the weights are allocated according to the sizes of the plurality of user features, and the first privacy loss parameter is obtained by weighted calculation, wherein the audit computing power and the user feature are positively correlated with the privacy loss parameter; inputting the plurality of user features and the plurality of first audit computing powers into an audit security analyzer, outputting a plurality of user audit security parameters, and obtaining a first audit security parameter by weighted calculation; calculating a first total audit computing power of the plurality of first audit computing powers, and obtaining a first computing power fitness, wherein the size of the first computing power fitness is negatively correlated with the first total audit computing power; According to the first privacy loss parameter, a first privacy fitness is calculated; the first audit security parameter is used as a first security fitness; According to the first computing power fitness, the first privacy fitness and the first security fitness, the first audit fitness of the plurality of first audit computing powers is calculated; The training process of the audit security analyzer comprises: According to the audit log of different users in the trusted data space in the historical time, a sample user feature set and a sample audit computing power set are collected, and the proportion of user data loss use under different sample user features and sample audit computing powers is collected as a sample audit risk parameter set, and a sample audit security parameter set is calculated; Based on machine learning, an audit security analyzer is constructed; The sample user feature set, the sample audit computing power set and the sample audit security parameter set are used as training data and test data to supervise the training and test of the audit security analyzer, and after the test loss converges, the trained audit security analyzer is obtained; iterative optimization of the plurality of user audit computing powers is performed to obtain a plurality of optimal audit computing powers, wherein the positive and negative optimization parameters are configured according to the first audit security parameter and the first privacy loss parameter, and optimization is performed; The plurality of optimal audit computing powers are used to perform data use auditing on the plurality of users respectively.

2. The data usage auditing process based on trusted data space according to claim 1, characterized in that, obtaining an audit computing power pool for data use auditing, setting an audit constraint, and allocating a plurality of first audit computing powers for auditing a plurality of user data uses in a trusted data space, comprising: obtaining an audit computing power pool for data use auditing; The total audit computing power for auditing the data use of the plurality of users is less than or equal to the total audit computing power of the audit computing power pool, which is used as an audit constraint; In the audit computing power pool, a plurality of first audit computing powers of a plurality of users satisfying the audit constraint are randomly set.

3. The data usage auditing process based on trusted data space according to claim 2, characterized in that, In the audit computing power pool, a plurality of first audit computing powers of a plurality of users satisfying the audit constraint are randomly set, comprising: In the audit computing power pool, a plurality of first audit computing powers for auditing data use of a plurality of users are randomly set; The first total audit computing power of the plurality of first audit computing powers is calculated to determine whether the audit constraint is met. If not, the audit computing power is re-set. If yes, the subsequent step is performed.

4. The data usage auditing process based on trusted data space according to claim 1, characterized in that, Iterative optimization of the audit computing power of the plurality of users is performed to obtain a plurality of optimal audit computing powers, including: Obtaining the average audit security parameter and the average privacy loss parameter in the historical time for audit processing as the benchmark audit security parameter and the benchmark privacy loss parameter; Calculating the ratio of the first audit security parameter to the benchmark audit security parameter to obtain a positive optimization parameter, and calculating the ratio of the first privacy loss parameter to the benchmark privacy loss parameter to obtain a negative optimization parameter, wherein the positive optimization parameter and the negative optimization parameter include a positive optimization quantity and a negative optimization quantity; According to the positive optimization parameter and the negative optimization parameter, the plurality of first audit computing powers are respectively adjusted positively and negatively to obtain a plurality of second positive audit computing power sets and a plurality of second negative audit computing power sets; The plurality of second positive audit computing power sets and the plurality of second negative audit computing power sets are randomly combined to process and obtain an audit fitness of an audit computing power combination that satisfies the audit constraint, and the plurality of audit computing powers corresponding to the maximum audit fitness are used as the basis for subsequent optimization adjustment; Iterative optimization is continued until convergence is achieved, and a plurality of optimal audit computing powers are obtained.

5. The trusted data space based data usage auditing process method according to claim 4, characterized in that, According to the positive optimization parameter and the negative optimization parameter, the plurality of first audit computing powers are respectively adjusted positively and negatively to obtain a plurality of second positive audit computing power sets and a plurality of second negative audit computing power sets, including: Obtaining a preset adjustment quantity; According to the positive optimization parameter, the negative optimization parameter, and the preset adjustment quantity, the positive adjustment quantity and the negative adjustment quantity are calculated and processed; According to the positive adjustment quantity and the negative adjustment quantity, the plurality of first audit computing powers are respectively adjusted positively and negatively to obtain a plurality of second positive audit computing power sets and a plurality of second negative audit computing power sets.

6. The trusted data space based data usage auditing process method according to claim 1, wherein, Using the plurality of optimal audit computing powers, data use audit processing is performed on the plurality of users, including: According to the plurality of optimal audit computing powers, an audit frequency is set, and data use audit processing is performed on the plurality of users.

7. A data usage auditing process system based on a trusted data space, characterized by, The system for implementing the data use audit processing method based on a trusted data space according to any one of claims 1-6, the system comprising: An algorithm power allocation module for obtaining an audit computing power pool for data use audit and setting an audit constraint, and allocating a plurality of first audit computing powers for auditing data use of a plurality of users in a trusted data space; An audit evaluation module for obtaining a plurality of user features of the plurality of users, combining the plurality of first audit computing powers, performing privacy loss analysis and audit security analysis, obtaining a first privacy loss parameter and a first audit security parameter, and calculating a first audit fitness of the plurality of first audit computing powers, including: Obtaining a plurality of user features of the plurality of users, wherein each user feature includes a user data proportion; According to the plurality of first audit computing powers and the plurality of user characteristics, a plurality of first user privacy loss parameters are calculated, and the first privacy loss parameters are obtained by weighted calculation according to the size of the plurality of user characteristics and the weight distribution; wherein the audit computing power and the user characteristic are positively correlated with the privacy loss parameter; The plurality of user characteristics and the plurality of first audit computing powers are input into the audit security analyzer, and a plurality of user audit security parameters are output, and the first audit security parameter is obtained by weighted calculation; The first total audit computing power of the plurality of first audit computing powers is calculated, and the first computing power fitness is obtained by calculation; According to the first privacy loss parameter, the first privacy fitness is obtained by calculation; The first audit security parameter is taken as the first security fitness; According to the first computing power fitness, the first privacy fitness and the first security fitness, the first audit fitness of the plurality of first audit computing powers is obtained by calculation; The training process of the audit security analyzer includes: According to the audit log of different users in the trusted data space in the historical time, the sample user characteristic set and the sample audit computing power set are collected, and the proportion of user data discreditable use under different sample user characteristics and sample audit computing power is collected as the sample audit risk parameter set, and the sample audit security parameter set is obtained by calculation; Based on machine learning, the audit security analyzer is constructed; The sample user characteristic set, the sample audit computing power set and the sample audit security parameter set are used as training data and test data to supervise the training and test of the audit security analyzer, and the trained audit security analyzer is obtained after the test loss converges; The iteration optimization module is used for iteration optimization of the plurality of user audit computing powers, and the plurality of optimal audit computing powers are obtained, wherein the positive and negative optimization parameters are configured according to the first audit security parameter and the first privacy loss parameter, and the optimization is carried out; The audit execution module is used for carrying out data use audit processing on the plurality of users by using the plurality of optimal audit computing powers.

Images