A method and system for dynamically dividing a privacy boundary based on a trusted data space

Abstract

The application discloses a kind of privacy boundary dynamic division method and system based on trusted data space, it is related to privacy computing technical field, the method includes: obtaining the current task characteristics of multiple users in trusted data space and multiple current privacy boundaries, and obtaining privacy boundary range;Randomly adjust multiple current privacy boundaries in privacy boundary range, obtain multiple first privacy boundaries;Analysis multiple first data accuracy of multiple first privacy boundaries, combined with multiple current task characteristics, analysis obtains multiple first task success rate;Analysis multiple task matching degrees of multiple user permissions and multiple current task characteristics, combined with multiple first task success rate, carry out privacy division evaluation of multiple first privacy boundaries, obtain first privacy fitness, and carry out privacy boundary division optimization, obtain multiple optimal privacy boundaries, carry out privacy boundary dynamic division.The present application solves the technical problems of inaccurate privacy boundary division in the prior art.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of privacy computing, and particularly relates to a privacy boundary dynamic division method and system based on a trusted data space. BACKGROUND

[0002] In the data-driven era, the trusted data space is widely used in the fields of finance, medical treatment, scientific research and the like as a key infrastructure for guaranteeing data security sharing and cooperation. The core thereof is to realize the fuzzification processing of user data privacy through privacy boundary division, so as to ensure that sensitive information is not leaked in the sharing process. The existing technology usually adopts a fixed privacy boundary division strategy, that is, the privacy boundary of a user is defined according to preset rules or static parameters. This method is based on historical data or a general scene, and emphasizes the maximization of data security. However, this fixed division method has significant defects, such as being unable to adapt to dynamically changing user task characteristics and neglecting utility balance, which may cause waste of computing resources and decline of user experience. SUMMARY

[0003] The present application provides a privacy boundary dynamic division method and system based on a trusted data space, which is used to solve the technical problem of inaccurate privacy boundary division in the prior art.

[0004] In view of the above problems, the present application provides a privacy boundary dynamic division method and system based on a trusted data space.

[0005] In a first aspect, the present application provides a privacy boundary dynamic division method based on a trusted data space, which comprises the following steps:

[0006] obtaining current task characteristics of a plurality of users in a trusted data space and a plurality of current privacy boundaries, and obtaining a privacy boundary range when the plurality of current task characteristics change, wherein the task characteristics include a task importance level and a task type, and the plurality of users include a plurality of user permissions;

[0007] randomly adjusting the plurality of current privacy boundaries in the privacy boundary range to obtain a plurality of first privacy boundaries;

[0008] analyzing a plurality of first data accuracies of the plurality of first privacy boundaries, and analyzing a plurality of first task success rates in combination with the plurality of current task characteristics;

[0009] analyzing a plurality of task matching degrees of the plurality of user permissions and the plurality of current task characteristics, performing privacy division evaluation of the plurality of first privacy boundaries in combination with the plurality of first task success rates to obtain a first privacy adaptability, and performing privacy boundary division optimization to obtain a plurality of optimal privacy boundaries and perform privacy boundary dynamic division, wherein the optimization process is adjusted according to the plurality of task matching degrees and the task success rates.

[0010] In a second aspect, the application provides a system for dynamically dividing privacy boundaries based on a trusted data space, comprising:

[0011] A privacy boundary acquisition module is configured to acquire current task features of a plurality of users in the trusted data space and a plurality of current privacy boundaries, and acquire a privacy boundary range when the plurality of current task features change, wherein the task features include a task importance level and a task type, and the plurality of users include a plurality of user permissions.

[0012] A random adjustment module is configured to randomly adjust the plurality of current privacy boundaries within the privacy boundary range to obtain a plurality of first privacy boundaries.

[0013] A success rate analysis module is configured to analyze a plurality of first data accuracies of the plurality of first privacy boundaries, and analyze a plurality of first task success rates in combination with the plurality of current task features.

[0014] A boundary dynamic division module is configured to analyze a plurality of task matching degrees of the plurality of user permissions and the plurality of current task features, perform privacy division evaluation of the plurality of first privacy boundaries in combination with the plurality of first task success rates, obtain a first privacy adaptability, and perform privacy boundary division optimization to obtain a plurality of optimal privacy boundaries, and perform privacy boundary dynamic division, wherein the optimization process is adjusted according to the plurality of task matching degrees and the task success rates.

[0015] One or more technical solutions provided in the application have at least the following technical effects or advantages:

[0016] The application provides a method and system for dynamically dividing privacy boundaries based on a trusted data space. By real-time sensing changes in current task features of users and intelligently adjusting and optimizing iterations within a preset privacy boundary range, the dynamic balance between data privacy protection and data use utility is significantly improved. Compared with traditional methods, the technical solution provided by the application significantly enhances the flexibility and adaptability of privacy boundary division, achieving the technical effect of protecting data privacy and security while fully releasing the value of data use in complex environments. BRIEF DESCRIPTION OF DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the embodiments of the application, the following will briefly introduce the drawings needed to be used in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the application, and other drawings can be obtained by those skilled in the art without creative labor.

[0018] Figure 1 A flowchart of a method for dynamically dividing privacy boundaries based on a trusted data space is provided for the embodiments of the application.

[0019] Figure 2 This is a schematic diagram of a privacy boundary dynamic partitioning system based on trusted data space, provided in an embodiment of this application.

[0020] The components represented by each number in the attached diagram are explained below:

[0021] Privacy boundary acquisition module 100, random adjustment module 200, success rate analysis module 300, and boundary dynamic division module 400. Detailed Implementation

[0022] This application provides a method and system for dynamically dividing privacy boundaries based on trusted data space, which addresses the technical problem of inaccurate privacy boundary division in existing technologies.

[0023] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0024] 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 that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or modules that are not explicitly listed or that are inherent to these processes, methods, products, or devices.

[0025] Example 1, as Figure 1 As shown, this application provides a method for dynamically partitioning privacy boundaries based on trusted data space, wherein the method includes:

[0026] S10: Obtain the current task characteristics and multiple current privacy boundaries of multiple users within the trusted data space. When the multiple current task characteristics change, obtain the range of the privacy boundaries. The task characteristics include the task importance level and task type, and the multiple users include multiple user permissions.

[0027] In multi-user collaborative computing scenarios within a trusted data space, users' task requirements often evolve dynamically with changes in business objectives or the external environment, such as a sudden shift from routine data retrieval to data analysis requiring higher data fidelity. However, existing privacy boundary management mechanisms often employ static configurations, which negatively impacts the effectiveness of data utilization.

[0028] Step S10 in the method provided in this application embodiment includes:

[0029] obtain current task features of a plurality of users in a trusted data space and a plurality of current privacy boundaries, wherein the plurality of users comprises a plurality of user authorities, and each user authority comprises a user level;

[0030] Specifically, it is determined whether the plurality of current task features and historical task features of the plurality of users are consistent, and a consistency ratio is obtained.

[0031] When the consistency ratio is less than or equal to a consistency ratio threshold, a privacy boundary range is obtained, wherein the privacy boundary range comprises a range of data privacy processing amplitude.

[0032] In the embodiments of the present application, current task features of a plurality of users in a trusted data space and a plurality of current privacy boundaries are obtained, and when the plurality of current task features change, a privacy boundary range is obtained, wherein the task features include a task importance level and a task type, and the plurality of users include a plurality of user authorities.

[0033] Specifically, current task features of a plurality of users in a trusted data space and a plurality of current privacy boundaries are obtained, wherein the plurality of users comprises a plurality of user authorities, and each user authority comprises a user level. For example, by calling the built-in API of the trusted data space, the current task features of each user currently being executed are retrieved in real time, for example, the user level of user A is level 2, the current task features are "importance: high, type: data retrieval", the user level of user B is level 1, and the current task features are "importance: medium, type: data analysis", and the current privacy boundary of each user is obtained, for example, the privacy boundary of user A is 10%, indicating that the data value is allowed to float ±10%, and the privacy boundary of user B is 15%, indicating that the data value is allowed to float ±15%. Wherein, the user level is used to represent the level of user authority, and the higher the user level, the greater the user authority.

[0034] Further, it is determined whether the plurality of current task features and the historical task features of the plurality of users are consistent, and a consistency ratio is obtained. For each user, the current task features are compared with the historical task features item by item, the number of consistent users is counted as a proportion of the total number of users, and the consistency ratio = consistent feature number ÷ total feature number. For example, there are 3 current task features, of which 2 current task features are consistent with the historical task features, and the consistency ratio = consistent feature number ÷ total feature number = 2 ÷ 3 = 0.67.

[0035] Further, when the consistency ratio is less than or equal to a consistency ratio threshold, a privacy boundary range is obtained, where the privacy boundary range includes a range of a privacy processing amplitude of the data. For example, the consistency ratio threshold is set to 0.6, that is, when the consistency ratio is less than or equal to 0.6, it is determined that the current task feature and the historical task feature are inconsistent, that is, the task changes and the privacy boundary range needs to be optimized. The privacy boundary range is obtained, where the privacy boundary range is a range of a privacy processing amplitude of the data, for example, one data in the data sequence is 10 and the privacy processing amplitude is 10%, and the display range is 10 ± 10%, that is, the data is randomly displayed as a value in the range of 9-11, and the data security is ensured through fuzzy privacy processing.

[0036] By obtaining the current task features of multiple users and the existing privacy boundaries in real time, and automatically triggering the acquisition of the privacy boundary range when detecting that the current task feature changes, a key transition from static configuration to dynamic perception is realized, and instead of blindly performing fixed adjustment, a response is made based on the actual change of the task, ensuring the accuracy basis of the adjustment behavior.

[0037] S20: In the privacy boundary range, the multiple current privacy boundaries are randomly adjusted to obtain multiple first privacy boundaries.

[0038] The traditional method may rely on fixed rules or linear adjustment strategies, for example, fixed scaling of the privacy boundaries of all users by a fixed ratio. However, this method has obvious exploration limitations and it is difficult to find a better boundary combination.

[0039] The step S20 in the method provided by the embodiment of the application includes:

[0040] The privacy boundary adjustment step is obtained.

[0041] In the privacy boundary range, the multiple current privacy boundaries are randomly adjusted by using the privacy boundary adjustment step to obtain multiple first privacy boundaries.

[0042] In the embodiment of the application, the multiple current privacy boundaries are randomly adjusted in the privacy boundary range to obtain multiple first privacy boundaries.

[0043] Specifically, first, the privacy boundary adjustment step is obtained, for example, the privacy boundary adjustment step is set to 2% in advance.

[0044] Further, within the privacy boundary range, the multiple current privacy boundaries are randomly adjusted by using a privacy boundary adjustment step to obtain multiple first privacy boundaries. Specifically, for each user, the current privacy boundary value is added or subtracted by an integer multiple of the adjustment step. New privacy boundary = current privacy boundary ± N × step, where N is a randomly generated positive integer. The random addition or subtraction operation and the multiple N are determined by calling the random number generation tool in the standard library of Python, such as the random library. For example, the current privacy boundary of user A is 10%, a random addition operation is performed, and N = 1, so the adjusted new privacy boundary is 10% + 1 × 2% = 12%. The current privacy boundary of user B is 15%, a random subtraction operation is performed, and N = 1, so the adjusted new boundary is 15% - 1 × 2% = 13%.

[0045] By introducing random adjustment as an exploration means, extensive and diverse sampling in the solution space is performed, the dependence on fixed adjustment rules is broken, and a series of first privacy boundary candidate schemes that are different in pattern and value are generated, providing sufficient selection materials for subsequent evaluation and optimization steps.

[0046] S30: Analyze multiple first data precisions of the multiple first privacy boundaries, and analyze to obtain multiple first task success rates in combination with the multiple current task characteristics.

[0047] If only the single dimension of privacy protection strength is evaluated, and the specific influence of these boundary settings on the upper-layer business task is ignored, the optimization process may lead to a scheme that is absolutely safe in privacy but unfeasible in practicality.

[0048] The step S30 in the method provided by the embodiments of the present application comprises:

[0049] According to the multiple first data privacy processing amplitudes in the multiple first privacy boundaries, multiple first data precisions are calculated and obtained;

[0050] The multiple first data precisions and the multiple current task characteristics are respectively combined and input into a task analyzer, and multiple first task success rates are output;

[0051] The training step of the task analyzer comprises:

[0052] According to the task execution records of the trusted data space in the historical time, a sample task characteristic set and a sample data precision set are collected, and the proportion of the number of times of successful task processing under different sample task characteristics and sample data precisions is collected to obtain a sample task success rate set;

[0053] Based on machine learning, a task analyzer is constructed, wherein the task analyzer comprises an input layer, a hidden layer, a full connection layer, and an output layer;

[0054] The task analyzer is supervised and trained using the sample task feature set, the sample data accuracy set, and the sample task success rate set, and the training is completed after the test accuracy rate is qualified.

[0055] In the embodiments of the present application, a plurality of first data accuracies of a plurality of first privacy boundaries are analyzed in combination with a plurality of current task features to obtain a plurality of first task success rates.

[0056] Specifically, first, a plurality of first data accuracies are calculated according to a plurality of first data privacy processing amplitudes in a plurality of first privacy boundaries. Exemplarily, there is an inverse relationship between data accuracy and privacy processing amplitude, that is, the greater the privacy processing amplitude, the lower the data accuracy. First data accuracy = 1 - first privacy processing amplitude. For example, if the first privacy processing amplitude of a user is 12%, then the first data accuracy = 1 - 12% = 88%. This calculation is performed for the first privacy boundary of each user to obtain the first data accuracy value corresponding to the first privacy boundary.

[0057] Further, the plurality of first data accuracies and the plurality of current task features are respectively combined and input into the task analyzer to output a plurality of first task success rates.

[0058] Specifically, the task analyzer is constructed. First, according to the task execution records of the trusted data space in the historical time, a sample task feature set and a sample data accuracy set are collected, and the number of times of successful task processing under different sample task features and sample data accuracies is collected to obtain a sample task success rate set. Among them, the sample task success rate = the number of times of successful task processing / the total number of task processing times.

[0059] Based on machine learning, a task analyzer is constructed, wherein the task analyzer comprises an input layer, a hidden layer, a fully connected layer, and an output layer. For example, the task analyzer is constructed based on Python Scikit-learn. The task analyzer is a multi-layer perception neural network, which includes an input layer for receiving an input vector composed of data precision and task feature encoding, a hidden layer using a ReLU activation function and including a plurality of neurons, a fully connected layer, and an output layer using a Sigmoid activation function to output a scalar representing success rate. The training uses a supervised learning method, with the prepared sample task feature set and sample data precision set as input features, the sample task success rate set as the true label, the Adam optimizer is used to minimize the mean square error between the model prediction value and the true label as the loss function, and the training set is iterated for multiple rounds. A part of the data is reserved as a test set during the training process to periodically evaluate the model performance; when the prediction accuracy of the model on the test set is stable and reaches the preset qualified standard, for example, the accuracy is more than 85%, the training process is terminated. Finally, the task analyzer model with fixed parameters is obtained.

[0060] Further, the plurality of first data precisions and the plurality of current task features are combined and input into the task analyzer to output a plurality of first task success rates. For example, one input can be: data precision 88% + current task feature "importance: high, type: data retrieval". After receiving the input, the task analyzer model analyzes and outputs a probability value between 0 and 1 as the first task success rate, for example, the first task success rate is 0.92.

[0061] By analyzing the data precision corresponding to the first privacy boundary and combining the current task feature to predict the corresponding task success rate, a precise mapping from abstract privacy settings to specific business impact is achieved, which can predict the potential impact of different privacy boundary schemes on actual task execution results, so that the adjustment decision of the privacy boundary considers the impact of data value rather than only considering the security effect, and the accuracy of the adjustment decision is improved.

[0062] S40: Analyze the plurality of task matching degrees of the plurality of user permissions and the plurality of current task features, combine the plurality of first task success rates, perform privacy division evaluation of the plurality of first privacy boundaries, obtain a first privacy adaptability, and perform privacy boundary division optimization to obtain a plurality of optimal privacy boundaries, and perform dynamic privacy boundary division, wherein the optimization process is adjusted according to the plurality of task matching degrees and the task success rates.

[0063] After obtaining multiple candidate privacy boundary schemes and their estimated task success rates, the final decision needs to comprehensively consider the overall performance of the schemes in security and utility. If only the task success rate is used for selection, sensitive data may be exposed to unnecessary risks.

[0064] The step S40 in the method provided by the embodiments of the present application comprises:

[0065] obtaining a first historical task feature set of a user executing a task in a first user authority in the multiple user authorities;

[0066] statistically obtaining a first task matching degree of a first current task feature in the multiple current task features in the first historical task feature set, and obtaining multiple task matching degrees of the multiple users;

[0067] according to the multiple task matching degrees, assigning multiple task weights, performing weighted calculation on the multiple first task success rates, and obtaining a first task privacy fitness degree, wherein the task weight and the task matching degree are positively correlated;

[0068] according to the multiple task matching degrees and the multiple first privacy boundaries, calculating and obtaining a first secure privacy fitness degree;

[0069] wherein, according to the multiple task matching degrees and the multiple first privacy boundaries, calculating and obtaining the first secure privacy fitness degree comprises:

[0070] sorting the multiple task matching degrees from large to small to obtain a task matching degree sequence;

[0071] sorting the multiple first privacy boundaries from small to large to obtain a first privacy boundary sequence;

[0072] calculating the similarity of the sorting positions of the multiple users in the task matching degree sequence and the first privacy boundary sequence to obtain the first secure privacy fitness degree;

[0073] according to the first task privacy fitness degree and the first secure privacy fitness degree, calculating and obtaining a first privacy fitness degree;

[0074] continuously randomly adjusting the multiple first privacy boundaries to obtain multiple second privacy boundaries, processing and obtaining a second privacy fitness degree, and performing iterative optimization;

[0075] calculating the mean value of the multiple task matching degrees to obtain a task credibility coefficient;

[0076] according to the task credibility coefficient and a preset reserved number, calculating and obtaining an optimized reserved number, and reserving the privacy boundary combination of the first optimized reserved number of privacy fitness degrees in the optimization process until the optimization converges;

[0077] The multiple optimal privacy boundaries with the maximum privacy fitness in the reserved privacy boundary combination are obtained, and dynamic privacy boundary division is performed.

[0078] In the embodiments of the present application, the multiple task matching degrees of the multiple user permissions and the multiple current task characteristics are analyzed, the privacy division evaluation of the multiple first privacy boundaries is performed in combination with the multiple first task success rates, the first privacy fitness is obtained, and the privacy boundary division optimization is performed to obtain the multiple optimal privacy boundaries, and dynamic privacy boundary division is performed. In the optimization process, the optimization adjustment is performed according to the multiple task matching degrees and the task success rates.

[0079] Specifically, first, the first historical task characteristic set of a user executing a task in a first user permission in the multiple user permissions is obtained. For example, the first historical task characteristic set can be obtained by querying the historical task execution log of the trusted data space, and the historical task characteristic set records the types and importance levels of the tasks executed by the user in the past.

[0080] Further, the occurrence proportion of a first current task characteristic in the first historical task characteristic set in the multiple current task characteristics is counted as the first task matching degree, and the multiple task matching degrees of the multiple users are analyzed and obtained. Specifically, for the user in the first user permission in the multiple user permissions, the first historical task characteristic set of the user is obtained, for example, the historical characteristic set of user A contains 10 task records, of which 6 are “importance: high, type: data retrieval”, and then the proportion of the number of occurrences of the current task characteristic such as “importance: high, type: data retrieval” in the historical characteristic set to the total number of times is counted, and the task matching degree = the number of occurrences of the current task characteristic in the historical characteristic set / the total number of tasks in the historical characteristic set. For example, the current task characteristic of user A occurs 6 times in the historical characteristic set, and the total number of tasks in the historical characteristic set is 10, so the task matching degree is 0.6. The same calculation is performed for each user to obtain the multiple task matching degrees of the multiple users.

[0081] Further, according to the multiple task matching degrees, the multiple task weights are assigned, the weighted calculation of the multiple first task success rates is performed, and the first task privacy fitness is obtained, wherein the task weight and the task matching degree are positively correlated. For example, the weighted average value of the first task success rate is calculated using the task matching degree as the weight. For example, the task matching degree of user A is 0.6, and the task matching degree of user B is 0.8, so the task weight of user A = 0.6 / (0.6+0.8) = 0.43, and the task weight of user B = 0.8 / (0.6+0.8) = 0.57. The first task privacy fitness = ∑(task weight of user × first task success rate of user). For example, the task weight of user A is 0.43, the first task success rate is 0.92, the task weight of user B is 0.57, and the first task success rate is 0.88, so the first task privacy fitness = 0.43 × 0.92 + 0.57 × 0.88 = 0.9.

[0082] Further, the first security privacy fitness is calculated according to the plurality of task fitnesses and the plurality of first privacy boundaries. First, the plurality of task fitnesses are sorted from large to small to obtain a task fitness sequence. For example, the user C fitness is 0.4, the user B fitness is 0.8, and the user A fitness is 0.6. After sorting, the sequence is [user B, user A, user C].

[0083] Further, the plurality of first privacy boundaries are sorted from small to large to obtain a first privacy boundary sequence. For example, the user A privacy boundary is 12%, the user B privacy boundary is 13%, and the user C privacy boundary is 10%. After sorting, the sequence is [user C, user A, user B].

[0084] Further, the similarity of the sorting positions of the plurality of users in the task fitness sequence and the first privacy boundary sequence is calculated to obtain the first security privacy fitness. Illustratively, the sorting position similarity = 1 - (sorting position difference / maximum possible difference). For example, the user A is ranked second in the task fitness sequence and second in the first privacy boundary sequence, the difference is 0, and the maximum possible difference is 2. Therefore, the sorting position similarity of the user A is 1 - (0 / 2) = 1. Using the same algorithm, the average of the sorting position similarities of the plurality of users in the sequence is calculated. For example, the sorting position similarity of the user B is 0, and the sorting position similarity of the user C is 0. Therefore, the first security privacy fitness = (1 + 0 + 0) / 3 = 0.33.

[0085] Further, the first privacy fitness is calculated according to the first task privacy fitness and the first security privacy fitness. Illustratively, the average of the first task privacy fitness and the first security privacy fitness is calculated to obtain the first privacy fitness. For example, the first task privacy fitness is 0.9, and the first security privacy fitness is 0.33. Therefore, the first privacy fitness = (0.9 + 0.33) / 2 = 0.62.

[0086] Further, the plurality of second privacy boundaries are obtained by further randomly adjusting the plurality of first privacy boundaries, the second privacy fitness is obtained by processing, and iterative optimization is performed. Specifically, the plurality of second privacy boundaries are obtained by using the same method as step S20, and then the processes of steps S30 and S40 are repeated. The data precision is calculated according to the second privacy boundary, the second task success rate is predicted by the task analyzer in combination with the current task characteristics, the second task privacy fitness and the second security privacy fitness are calculated in combination with the task fitness, and finally the second privacy fitness is obtained.

[0087] Further, the average of the plurality of task fitnesses is calculated to obtain a task trust coefficient. For example, the task fitness of the user A is 0.6, and the task fitness of the user B is 0.8. Therefore, the task trust coefficient = (0.6 + 0.8) / 2 = 0.7.

[0088] Based on the task credibility coefficient and the preset retention quantity, the optimized retention quantity is calculated. During the optimization process, the privacy boundary combination of the privacy fitness of the original optimized retention quantity is retained until the optimization converges. For example, the preset retention quantity can be set to 200. The higher the task credibility coefficient, the more realistic the task may be, and a smaller optimized retention quantity can be retained to speed up the iteration. Optimized retention quantity = preset retention quantity × (1 - task credibility coefficient). For example, if the task credibility coefficient is 0.7, then the optimized retention quantity = 200 × (1 - 0.7) = 60.

[0089] Furthermore, the system obtains the multiple optimal privacy boundaries with the highest privacy fitness among the retained privacy boundary combinations and dynamically partitions the privacy boundaries. Specifically, if the privacy fitness of a privacy boundary combination is 0.85, which is higher than other privacy boundary combinations, then the privacy boundaries in the combination with a fitness of 0.85 are set as multiple optimal privacy boundaries. These optimal privacy boundaries are then applied to the trusted data space through the system interface to achieve dynamic updates of the privacy boundaries.

[0090] By comprehensively analyzing the matching degree between user permissions and current task characteristics, and combining the task success rate to evaluate and iteratively optimize the privacy boundary partitioning of candidate boundaries, dynamic and intelligent partitioning of privacy boundaries is ultimately achieved. The iterative optimization process automatically and continuously searches for the privacy boundary configuration with the highest overall fitness. The final output set of optimal privacy boundaries is not an extreme solution in either security or utility, but rather a dynamic balance point that maximizes overall performance under the current task scenario and user permission structure, thus achieving dual protection of data privacy and data usability.

[0091] Example 2, as Figure 2 As shown, based on the same inventive concept as the method for dynamically partitioning privacy boundaries based on trusted data space provided in Embodiment 1, this embodiment of the invention also provides a system for dynamically partitioning privacy boundaries based on trusted data space, comprising:

[0092] The privacy boundary acquisition module 100 is used to acquire the current task characteristics and multiple current privacy boundaries of multiple users in the trusted data space. When the multiple current task characteristics change, the privacy boundary range is acquired. The task characteristics include the task importance level and task type, and the multiple users include multiple user permissions.

[0093] The random adjustment module 200 is used to randomly adjust the plurality of current privacy boundaries within the privacy boundary range to obtain a plurality of first privacy boundaries;

[0094] The success rate analysis module 300 is used to analyze the precision of multiple first data of multiple first privacy boundaries, and combine the multiple current task features to analyze and obtain the success rate of multiple first tasks.

[0095] The boundary dynamic division module 400 is configured to analyze a plurality of task matching degrees of the plurality of user permissions and the plurality of current task features, combine the plurality of first task success rates, perform privacy division evaluation on the plurality of first privacy boundaries, obtain a first privacy adaptability, and perform privacy boundary division optimization to obtain a plurality of optimal privacy boundaries, and perform privacy boundary dynamic division. In the optimization process, the plurality of task matching degrees and the task success rates are used for optimization adjustment.

[0096] In one embodiment, the privacy boundary acquisition module 100 is further configured to:

[0097] acquire current task features of a plurality of users in a trusted data space and a plurality of current privacy boundaries, wherein the plurality of users includes a plurality of user permissions, and each user permission includes a user level;

[0098] determine whether the plurality of current task features and historical task features of the plurality of users are consistent, and obtain a consistency ratio;

[0099] when the consistency ratio is less than or equal to a consistency ratio threshold, acquire a privacy boundary range, wherein the privacy boundary range includes a range of data privacy processing amplitudes.

[0100] In one embodiment, the random adjustment module 200 is further configured to:

[0101] acquire a privacy boundary adjustment step;

[0102] randomly adjust the plurality of current privacy boundaries in the privacy boundary range by using the privacy boundary adjustment step to obtain a plurality of first privacy boundaries.

[0103] In one embodiment, the success rate analysis module 300 is further configured to:

[0104] obtain a plurality of first data precisions according to a plurality of first data privacy processing amplitudes in the plurality of first privacy boundaries;

[0105] combine the plurality of first data precisions and the plurality of current task features, respectively, and input them into a task analyzer to output a plurality of first task success rates;

[0106] The training step of the task analyzer includes:

[0107] According to the task execution records of the trusted data space in the historical time, a sample task feature set and a sample data precision set are collected, and a number of times of successful task processing under different sample task features and sample data precisions is collected to obtain a sample task success rate set;

[0108] A task analyzer is constructed based on machine learning, wherein the task analyzer includes an input layer, a hidden layer, a fully connected layer, and an output layer;

[0109] The task analyzer is trained under supervision using the sample task feature set, sample data accuracy set, and sample task success rate set, and training is completed after the accuracy test is passed.

[0110] In one embodiment, the boundary dynamic partitioning module 400 is further configured to:

[0111] Obtain the first historical task feature set of the user with the first user permission among multiple user permissions;

[0112] The proportion of the first current task feature appearing in the first historical task feature set among multiple current task features is used as the first task matching degree, and the matching degree of multiple tasks of multiple users is obtained by analysis.

[0113] Based on the matching degree of multiple tasks, multiple task weights are assigned, and the success rates of multiple first tasks are weighted and calculated to obtain the privacy fitness of the first task. Among them, the task weights and task matching degrees are positively correlated.

[0114] The first security and privacy fitness score is calculated based on multiple task matching scores and multiple first privacy boundaries.

[0115] The first security and privacy fitness degree is calculated based on multiple task matching degrees and multiple first privacy boundaries, including:

[0116] Sort the matching scores of multiple tasks from largest to smallest to obtain a task matching score sequence;

[0117] Sort the multiple first privacy boundaries in ascending order to obtain a sequence of first privacy boundaries;

[0118] Calculate the similarity between the ranking positions of multiple users within the task matching degree sequence and the first privacy boundary sequence to obtain the first security and privacy fitness degree;

[0119] The first privacy fitness score is calculated based on the first task privacy fitness score and the first security privacy fitness score.

[0120] Continue to randomly adjust multiple first privacy boundaries to obtain multiple second privacy boundaries, process them to obtain second privacy fitness, and perform iterative optimization;

[0121] Calculate the average of the matching degrees of multiple tasks to obtain the task credibility coefficient;

[0122] Based on the task credibility coefficient and the preset retention quantity, the optimized retention quantity is calculated. During the optimization process, the privacy boundary combination of the privacy fitness of the original optimized retention quantity is retained until the optimization converges.

[0123] Obtain a plurality of optimal privacy boundaries with maximum privacy fitness in the reserved privacy boundary combination, and perform dynamic division of the privacy boundaries.

[0124] To sum up, the embodiments of the present application have at least the following technical effects:

[0125] The present application provides a privacy boundary dynamic division method and system based on a trusted data space. By real-time sensing of the changes in the current task characteristics of the user, and intelligent random adjustment and optimization iteration within the preset privacy boundary range, the dynamic balance ability between data privacy protection and data use utility is significantly improved. Specifically, by analyzing the data accuracy under different adjustment schemes and predicting the corresponding task success rate, it is ensured that the privacy processing will not excessively damage the usability of the data. Further, by introducing the matching degree of user authority and current task characteristics as a key evaluation index, personalized privacy resource allocation matching the authority and demand is realized, so as to allocate appropriate data accuracy for high-authority or high-demand tasks under the premise of guaranteeing the overall data security framework, effectively avoiding the data value loss or low task execution efficiency caused by fixed privacy strategy. In addition, by constructing a privacy fitness combining task success rate and security, and based on this, iterative optimization and convergence selection are performed, the approximately optimal privacy boundary combination under a specific task scenario can be automatically and efficiently found, which not only improves the response speed, but also enhances the intelligent level of iterative optimization. Compared with the traditional method, the technical scheme provided by the present application significantly enhances the flexibility and adaptability of privacy boundary division, and achieves the technical effect of protecting data privacy and security while fully releasing the data use value in a complex environment.

[0126] It should be noted that the above 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.

[0127] 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 principles of the present application shall be included in the protection scope of the present application.

[0128] The specification and drawings are, of course, to be regarded in an illustrative rather than a restrictive sense. It is to be understood that any such modifications, variations, combinations or equivalents that fall within the scope of the application are intended to be embraced herein.

Claims

1. A method for dynamic privacy boundary partitioning based on trusted data space, characterized in that, The method comprises: obtaining the current task characteristics of a plurality of users in a trusted data space and a plurality of current privacy boundaries, and obtaining a privacy boundary range when the plurality of current task characteristics change, wherein the task characteristics include task importance levels and task types, and the plurality of users include a plurality of user permissions; randomly adjusting the plurality of current privacy boundaries within the privacy boundary range to obtain a plurality of first privacy boundaries; analyzing a plurality of first data accuracies of the plurality of first privacy boundaries, and analyzing a plurality of first task success rates in combination with the plurality of current task characteristics, comprising: calculating a plurality of first data accuracies according to a plurality of first data privacy processing amplitudes within the plurality of first privacy boundaries; inputting the plurality of first data accuracies and the plurality of current task characteristics into a task analyzer in combination, respectively, to output a plurality of first task success rates; analyzing a plurality of task matching degrees of the plurality of user permissions and the plurality of current task characteristics, performing privacy division evaluation of the plurality of first privacy boundaries in combination with the plurality of first task success rates, obtaining a first privacy fitness, and performing privacy boundary division optimization to obtain a plurality of optimal privacy boundaries, and performing dynamic privacy boundary division, wherein the optimization process is adjusted according to the plurality of task matching degrees and the task success rates, comprising: obtaining a first historical task characteristic set of a user executing a task in a first user permission in the plurality of user permissions; statistically obtaining a first task matching degree as a proportion of a first current task characteristic in the plurality of current task characteristics in the first historical task characteristic set, and analyzing a plurality of task matching degrees of the plurality of users; allocating a plurality of task weights according to the plurality of task matching degrees to perform weighted calculation on the plurality of first task success rates to obtain a first task privacy fitness, wherein the task weights and the task matching degrees are positively correlated; calculating a first secure privacy fitness according to the plurality of task matching degrees and the plurality of first privacy boundaries, comprising: sorting the plurality of task matching degrees from large to small to obtain a task matching degree sequence; sorting the plurality of first privacy boundaries from small to large to obtain a first privacy boundary sequence; calculating a similarity of the sorting positions of the plurality of users in the task matching degree sequence and the first privacy boundary sequence to obtain the first secure privacy fitness; calculating the first privacy fitness according to the first task privacy fitness and the first secure privacy fitness; continuing to randomly adjust the plurality of first privacy boundaries to obtain a plurality of second privacy boundaries, processing a second privacy fitness, and performing iterative optimization; calculating a mean value of the plurality of task matching degrees to obtain a task trust coefficient; calculating an optimized reserved number according to the task trust coefficient and a preset reserved number, reserving a privacy boundary combination of the first privacy fitness in the optimization process until the optimization converges; obtaining a plurality of optimal privacy boundaries with the maximum privacy fitness in the reserved privacy boundary combination, and performing dynamic privacy boundary division.

2. The method of claim 1, wherein, Obtaining the current task characteristics of a plurality of users in a trusted data space and a plurality of current privacy boundaries, and obtaining a privacy boundary range when the plurality of current task characteristics change, comprising: obtaining current task features of a plurality of users in a trusted data space and a plurality of current privacy boundaries, wherein the plurality of users includes a plurality of user permissions, and each user permission includes a user level; determining whether the plurality of current task features and historical task features of the plurality of users are consistent, and obtaining a consistency ratio; when the consistency ratio is less than or equal to a consistency ratio threshold, obtaining a privacy boundary range, wherein the privacy boundary range includes a range of data privacy processing amplitude.

3. The method of claim 1, wherein, In the privacy boundary range, the plurality of current privacy boundaries are randomly adjusted to obtain a plurality of first privacy boundaries, including: obtaining a privacy boundary adjustment step; In the privacy boundary range, the plurality of current privacy boundaries are randomly adjusted using the privacy boundary adjustment step to obtain a plurality of first privacy boundaries.

4. The method of claim 1, wherein, The training step of the task analyzer includes: According to the task execution record of the trusted data space in the historical time, the sample task feature set and the sample data precision set are collected, and the number of times of successful task processing under different sample task features and sample data precision is collected to obtain a sample task success rate set; Based on machine learning, a task analyzer is constructed, wherein the task analyzer includes an input layer, a hidden layer, a full connection layer and an output layer; The sample task feature set, the sample data precision set and the sample task success rate set are used to supervise the training of the task analyzer, and the training is completed after the test accuracy is qualified. 5.A system for dynamic privacy boundary partitioning based on trusted data space, characterized in that, The system for implementing the method for dynamically dividing the privacy boundary based on the trusted data space according to any one of claims 1-4, the system comprising: a privacy boundary acquisition module for acquiring current task features of a plurality of users in a trusted data space and a plurality of current privacy boundaries, and acquiring a privacy boundary range when the plurality of current task features change, wherein the task features include task importance level and task type, and the plurality of users includes a plurality of user permissions; a random adjustment module for randomly adjusting the plurality of current privacy boundaries in the privacy boundary range to obtain a plurality of first privacy boundaries; a success rate analysis module for analyzing a plurality of first data precisions of the plurality of first privacy boundaries, and combining the plurality of current task features to analyze and obtain a plurality of first task success rates, including: calculating a plurality of first data precisions according to a plurality of first data privacy processing amplitudes in the plurality of first privacy boundaries; inputting the plurality of first data precisions and the plurality of current task features into a task analyzer respectively to output a plurality of first task success rates; a boundary dynamic division module for analyzing a plurality of task matching degrees of the plurality of user permissions and the plurality of current task features, combining the plurality of first task success rates to perform privacy division evaluation of the plurality of first privacy boundaries, obtaining a first privacy adaptability, and performing privacy boundary division optimization to obtain a plurality of optimal privacy boundaries for dynamic privacy boundary division, wherein the optimization process is adjusted according to the plurality of task matching degrees and the task success rates, including: obtaining a first historical task feature set of a user executing a task in a first user permission in the plurality of user permissions; The proportion of the first current task feature in the first historical task feature set is counted as a first task matching degree, and a plurality of task matching degrees of a plurality of users are obtained by analysis; According to the plurality of task matching degrees, a plurality of task weights are assigned, a plurality of first task success rates are weighted calculated, and a first task privacy adaptability is obtained, wherein the task weight and the task matching degree are positively correlated; According to the plurality of task matching degrees and the plurality of first privacy boundaries, a first security privacy adaptability is calculated and obtained, including: The plurality of task matching degrees are sorted from large to small to obtain a task matching degree sequence; The plurality of first privacy boundaries are sorted from small to large to obtain a first privacy boundary sequence; The similarity of the sorting positions of the plurality of users in the task matching degree sequence and the first privacy boundary sequence is calculated to obtain the first security privacy adaptability; According to the first task privacy adaptability and the first security privacy adaptability, a first privacy adaptability is calculated and obtained; Continue to randomly adjust the plurality of first privacy boundaries to obtain a plurality of second privacy boundaries, process a second privacy adaptability, and perform iterative optimization; The mean of the plurality of task matching degrees is calculated to obtain a task credibility coefficient; According to the task credibility coefficient and a preset retention number, an optimized retention number is calculated and obtained, and the privacy boundary combination of the privacy adaptability of the first optimization retention number is retained in the optimization process until the optimization converges; The plurality of optimal privacy boundaries with the maximum privacy adaptability in the retained privacy boundary combination are obtained, and the privacy boundary is dynamically divided.

Images