Data desensitization processing method and system

Abstract

The invention discloses a data desensitization processing method and system, and belongs to the technical field of data processing, and the processing method comprises the following specific steps: I, collecting original data from a plurality of groups of data sources, and preprocessing each group of data; iI, identifying sensitive information in each group of data to formulate a corresponding desensitization strategy; iII, converting the sensitive information according to the desensitization strategy and optimizing the conversion process in real time; iV, evaluating a desensitization result and adjusting a desensitization process according to the evaluation result; v, storing the desensitized data and continuously tracking the use condition of the data; according to the method, the desensitized data has higher privacy protection capability, the desensitization strategy is continuously optimized, the information loss is reduced, the data quality is ensured, the desensitization strategy can be dynamically adjusted according to data characteristics and application requirements, the overall desensitization effect can be optimized, the actual availability is relatively high, and the method is suitable for popularization and application. Differentiated desensitization strategy optimization can be implemented for different data types, so that the desensitization process is more refined, and the adaptability is higher.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a data desensitization processing method and system. Background Art

[0002] With the rapid development of information technology, the scale of data generation and application has grown exponentially. In the era of big data and cloud computing, a large amount of data involving personal privacy and sensitive information is widely used in various industries, such as finance, medical care, e-commerce and other fields. At the same time, the need to protect privacy data has become more urgent. Both personal privacy information and corporate confidential data are facing the risk of information leakage. To solve this problem, data desensitization technology has emerged. Its core idea is to transform sensitive data through a series of algorithms and technologies so that it can maintain the authenticity and availability of data while losing direct identifying information. Traditional data desensitization methods mainly rely on static rules and algorithms. Although they can protect data privacy to a certain extent, with the expansion of data scale, the diversification of application scenarios and the increase of data complexity, traditional methods are difficult to cope with the needs of efficient and secure data desensitization in actual scenarios. Therefore, more intelligent and dynamic desensitization strategies are urgently needed to improve the desensitization effect.

[0003] After searching, Chinese patent number CN118568772A discloses a data desensitization processing method and system. Although this invention improves the security of sensitive data to a certain extent, it cannot guarantee the privacy protection ability of the desensitized data, increases information loss, and reduces data quality; in addition, the existing data desensitization processing method and system cannot optimize the overall desensitization effect, and the actual usability is low. It can only implement differentiated desensitization strategy optimization for a single data type, and cannot achieve refined desensitization, and has low adaptability; for this reason, we propose a data desensitization processing method and system. Summary of the invention

[0004] The purpose of the present invention is to solve the defects in the prior art and to propose a data desensitization processing method and system.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] A data desensitization processing method, the specific steps of the processing method are as follows:

[0007] Ⅰ. Collect raw data from multiple data sources and preprocess each set of data;

[0008] II. Identify sensitive information in each group of data to develop corresponding desensitization strategies;

[0009] III. Convert sensitive information according to desensitization strategies and optimize the conversion process in real time;

[0010] IV. Evaluate the desensitization results and adjust the desensitization process based on the evaluation results;

[0011] V. Store desensitized data and continuously track data usage;

[0012] VI. Regularly audit processing procedures and update desensitization strategies.

[0013] As a further solution of the present invention, the specific steps of preprocessing each group of data in step I are as follows:

[0014] Q1.1: Collect each set of raw data collected, and identify and remove duplicate data information by comparing the unique identifiers of the data records. Then detect and locate the missing values ​​in each set of raw data, and fill the missing values ​​in the processed data by mean or median filling, and integrate the processed data into a set of data to be tested;

[0015] Q1.2: Calculate the first quartile and the third quartile of the data set to be tested, and calculate the interquartile range based on the two sets of quartiles. Then determine the upper and lower limits of the abnormal data based on the interquartile range, and identify the abnormal data in the data set to be tested that is less than the lower limit or greater than the upper limit, and directly remove the detected abnormal data from the data set to be tested;

[0016] Q1.3: After removing abnormal data, the data in the data set to be tested is formatted to unify the data formats, and then the Max-Min normalization method is used to standardize the processed data sets.

[0017] As a further solution of the present invention, the specific steps of identifying sensitive information in each group of data in step II are as follows:

[0018] Q2.1: Remove stop words and punctuation marks from text data, convert text data into numerical form using the TF-IDF method, and then design and build a recognition model based on the TensorFlow deep learning framework;

[0019] Q2.2: Collect each set of historical data, and then label the sensitive information in each set of historical data automatically and manually. Divide the labeled data into a test set and a validation set, and pass the test set data into the recognition model.

[0020] Q2.3: After the recognition model receives the test set data, it uses the forward propagation algorithm to pass each group of data through the model input layer, hidden layer, and output layer in turn, and outputs the prediction result. The cross entropy loss function is used to calculate the loss value between the prediction result and the actual value. Then, the back propagation algorithm is used to calculate the loss value and the gradient of each network layer of the model, and the Adam optimizer is used to adjust the model weights.

[0021] Q2.4: After each round of training, the recognition model is evaluated using the validation set data to calculate the accuracy, recall, and F1-score indicators. The recognition model hyperparameter information is adjusted based on the evaluation results. The recognition module is trained and verified repeatedly until the model loss value converges to the preset range, and then the training is stopped.

[0022] Q2.5: The latest data to be detected is input into the trained recognition model. The recognition model obtains the category probability of each data to be detected through the forward propagation algorithm. The output results of the model are then post-processed, and the data that exceeds the set threshold is extracted and marked as sensitive data. The identified sensitive data is then counted and analyzed.

[0023] As a further solution of the present invention, the specific steps of the real-time optimization conversion process described in step III are as follows:

[0024] Q3.1: Convert sensitive data according to the selected desensitization strategy, set optimization goals according to user needs and data conversion conditions, and construct the corresponding loss function, set the initial temperature, cooling rate and number of iterations. At the same time, take the current desensitization conversion process as the initial solution and calculate its loss value through the constructed loss function;

[0025] Q3.2: In each round of iteration, the current desensitization conversion process is randomly perturbed to generate a new desensitization conversion process, and the loss value of the strategy is calculated. The loss value of the original conversion process is compared with the loss value of the new conversion process. If the loss value of the new solution is lower than the current solution, the new solution is accepted. If the loss value of the new solution is higher than the current solution, the new solution is accepted according to the Metropolis criterion.

[0026] Q3.3: After each round of iteration, the current temperature is lowered according to the preset cooling rate, and random perturbations and process selections are repeated until the preset maximum number of iterations is reached. The iteration ends, and the optimal desensitization conversion process and its corresponding loss value are output. The conversion process and loss are saved to provide a basis for subsequent data storage and use.

[0027] A data desensitization processing system, comprising a collection and processing module, an information identification module, a strategy formulation module, a data conversion module, a result evaluation module, an optimization and adjustment module, a decision analysis module, a data storage module and an audit update module;

[0028] The collection and processing module is used to automatically collect raw data from different data sources, and clean and standardize the collected data;

[0029] The information identification module is used to analyze the processed data and identify sensitive information in each group of data;

[0030] The strategy formulation module is used to formulate a desensitization strategy based on the identified sensitive information;

[0031] The data conversion module is used to execute the desensitization strategy and convert sensitive information;

[0032] The result evaluation module is used to collect desensitization data information in real time and evaluate the desensitization results;

[0033] The optimization and adjustment module is used to monitor and adjust the desensitization process;

[0034] The decision analysis module is used to comprehensively evaluate the desensitization results by integrating multiple factors;

[0035] The data storage module is used to store and manage the desensitized data using encryption technology and access control;

[0036] The audit update module is used to continuously track the use of desensitized data, conduct regular audits, and regularly update desensitization strategies and algorithms based on new security requirements and technological developments.

[0037] As a further solution of the present invention, the specific steps of monitoring and adjusting the optimization and adjustment module are as follows:

[0038] Q4.1: Randomly generate a set of initial solutions, each of which represents a different combination of desensitization strategies. Construct a corresponding search space, calculate the fitness value of each initial solution based on the desensitization effect and data quality, compare the fitness values ​​of each initial solution, and take the initial solution with the highest fitness value as the current optimal solution;

[0039] Q4.2: Generate a random number in the range of 0 to 1 and calculate the position update amplitude of each initial solution. Then set a set of probability values. If the random number is less than the probability value, perform a local search. Otherwise, perform a global search.

[0040] Q4.3: When searching locally, calculate the distance between the remaining solutions and the optimal solution in the search space, and make local adjustments based on the position update amplitude of each solution. Simultaneously, during the local search, simulate the spiral path using the spiral motion formula to obtain the position changes of the remaining solutions as they perform spiral motion around the optimal solution.

[0041] Q4.4: When searching globally, a new solution is randomly generated in the entire search space. After the new solution is generated, the fitness value of the new solution is calculated, and it is judged whether the new solution is better than the existing solution. If it is better, the existing solution is replaced by the new solution. Otherwise, a new random solution is generated for judgment;

[0042] Q4.5: In each iteration, check whether a better solution is found. If a better solution is found, update the current best solution. Otherwise, keep the current best solution unchanged and repeat the iteration. When the fitness value change is less than the predetermined threshold, stop the iteration and output the optimal desensitization strategy to replace the original desensitization strategy. Convert sensitive data according to the optimal desensitization strategy and monitor the conversion information in real time.

[0043] Compared with the prior art, the present invention has the following beneficial effects:

[0044] 1. The data desensitization processing method randomly generates a group of initial solutions, each of which represents a different combination of desensitization strategies. A corresponding search space is constructed. According to the desensitization effect and data quality of each initial solution, its fitness value is calculated, and the fitness value of each initial solution is compared. The initial solution with the highest fitness value is taken as the current optimal solution. A random number in the range of 0 to 1 is randomly generated, and the position update amplitude of each initial solution is calculated. Then a group of probability values ​​are set. If the random number is less than the probability value, a local search is performed, otherwise, a global search is performed. When searching locally, the distance between the remaining solutions and the optimal solution in the search space is calculated, and local adjustments are made through the position update amplitude of each solution. At the same time, during the local search process, the spiral path is simulated by the spiral motion formula to obtain the position changes of the remaining solutions when they are spirally moving around the optimal solution. During global search, new solutions are randomly generated in the entire search space. After the new solution is generated, the fitness value of the new solution is calculated, and it is judged whether the new solution is better than the existing solution. If it is, the existing solution is replaced by the new solution, otherwise, a random solution is regenerated for judgment. In each iteration, check whether a better solution is found. If a better solution is found, the current best solution is updated, otherwise, the current best solution is kept unchanged and the iteration is repeated. When the fitness value change is less than the predetermined threshold, the iteration is stopped, and the optimal desensitization strategy is output to replace the original desensitization strategy. Sensitive data is converted according to the optimal desensitization strategy, and the conversion information is monitored in real time to ensure that the desensitized data has higher privacy protection capabilities. By continuously optimizing the desensitization strategy, information loss is reduced, data quality is guaranteed, and the desensitization strategy can be dynamically adjusted according to data characteristics and application requirements.

[0045] 2. The present invention converts sensitive data according to the selected desensitization strategy, sets the optimization target according to user needs and data conversion conditions, constructs a corresponding loss function, sets the initial temperature, cooling rate and number of iterations, and takes the current desensitization conversion process as the initial solution, and calculates its loss value through the constructed loss function. In each round of iteration, the current desensitization conversion process is randomly perturbed to generate a new desensitization conversion process, and the loss value of the strategy is calculated, and the loss value of the original conversion process is compared with the loss value of the new conversion process. If the loss value of the new solution is lower than the current solution, the new solution is accepted. If the loss value of the new solution is higher than the current solution, the new solution is accepted according to the Metropolis criterion. After each round of iteration, the current temperature is reduced according to the preset cooling rate, and random perturbations and process selections are repeated until the preset maximum number of iterations is reached. The iteration ends, and the optimal desensitization conversion process and its corresponding loss value are output, and the conversion process and loss are saved to provide a basis for subsequent data storage and use, which can optimize the overall desensitization effect, have high practical availability, and can implement differentiated desensitization strategy optimization for different data types, so that the desensitization process is more refined and more adaptable. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention.

[0047] Figure 1 A flowchart of a data desensitization processing method proposed by the present invention;

[0048] Figure 2 This is a system block diagram of a data desensitization processing system proposed in the present invention. DETAILED DESCRIPTION

[0049] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0050] Example 1

[0051] Reference Figure 1 This embodiment discloses a data desensitization processing method, and the specific steps of the processing method are as follows:

[0052] Collect raw data from multiple data sources and preprocess each set of data.

[0053] Specifically, each group of collected original data is collected, and the data information of duplicate records is identified and removed by comparing the unique identifiers of the data records. Then, the position of the missing values ​​in each group of original data is detected and located, and the missing values ​​in the processed data are filled by the mean or the median, and the processed data are integrated into a set of data sets to be tested, the first quartile and the third quartile of the data sets to be tested are calculated, and the interquartile range is calculated based on the two sets of quartiles, and then the upper and lower limits of the abnormal data are determined according to the interquartile range, and the abnormal data in the data sets to be tested that is less than the lower limit or greater than the upper limit is identified, and the detected abnormal data is directly removed from the data sets to be tested, and the data in the data sets to be tested after the abnormal data are removed are formatted to unify the data formats, and then the processed data groups are standardized through the Max-Min normalization method.

[0054] Identify sensitive information in each group of data to develop corresponding desensitization strategies.

[0055] Specifically, the stop words and punctuation marks in the text data are removed, and then the text data is converted into numerical form through the TF-IDF method. Then, the recognition model is designed and constructed based on the TensorFlow deep learning framework, and each group of historical data is collected. Then, the sensitive information in each group of historical data is annotated automatically and manually, and the annotated data is divided into a test set and a verification set. The test set data is passed into the recognition model. After the recognition model receives the test set data, it passes each group of data through the model input layer, hidden layer, and output layer in turn through the forward propagation algorithm, and outputs the prediction result. The loss value between the prediction result and the actual value is calculated through the cross entropy loss function, and then the back propagation algorithm is used to calculate The loss value is compared with the gradient of each network layer of the model, and the weight of the model is adjusted through the Adam optimizer. After each round of training, the recognition model is evaluated through the verification set data, and the accuracy, recall rate and F1-score indicators are calculated. The recognition model hyperparameter information is adjusted according to the evaluation results. The training and verification of the recognition module are repeated until the model loss value converges to the preset range. The training is stopped and the latest data to be detected is input into the trained recognition model. The recognition model obtains the category probability of each data to be detected through the forward propagation algorithm. After that, the output results of the model are post-processed, and the data exceeding the set threshold are extracted and marked as sensitive data. Then, the identified sensitive data is counted and analyzed.

[0056] Convert sensitive information based on desensitization strategies and optimize the conversion process in real time.

[0057] Specifically, the sensitive data is converted according to the selected desensitization strategy, the optimization goal is set according to user needs and data conversion conditions, and the corresponding loss function is constructed, the initial temperature, cooling rate and number of iterations are set, and the current desensitization conversion process is used as the initial solution, and its loss value is calculated by the constructed loss function. In each round of iteration, the current desensitization conversion process is randomly perturbed to generate a new desensitization conversion process, and the loss value of the strategy is calculated, and the loss value of the original conversion process is compared with the loss value of the new conversion process. If the loss value of the new solution is lower than the current solution, the new solution is accepted. If the loss value of the new solution is higher than the current solution, the new solution is accepted according to the Metropolis criterion. After each round of iteration, the current temperature is lowered according to the preset cooling rate, and random perturbations and process selections are repeated until the preset maximum number of iterations is reached. The iteration ends, and the optimal desensitization conversion process and its corresponding loss value are output, and the conversion process and loss are saved to provide a basis for subsequent data storage and use.

[0058] Evaluate the desensitization results and adjust the desensitization process based on the evaluation results.

[0059] Store de-identified data and keep track of data usage.

[0060] Regularly audit processing procedures and update desensitization strategies.

[0061] Example 2

[0062] Reference Figure 2 This embodiment discloses a data desensitization processing system, including a collection and processing module, an information identification module, a strategy formulation module, a data conversion module, a result evaluation module, an optimization and adjustment module, a decision analysis module, a data storage module and an audit update module.

[0063] The collection and processing module is used to automatically collect raw data from different data sources, and to clean and standardize the collected data; the information identification module is used to analyze the processed data and identify sensitive information in each group of data; the strategy formulation module is used to formulate desensitization strategies based on the identified sensitive information.

[0064] The data conversion module is used to execute desensitization strategies and convert sensitive information; the result evaluation module is used to collect desensitized data information in real time and evaluate the desensitization results; the optimization and adjustment module is used to monitor and adjust the desensitization process.

[0065] Specifically, a group of initial solutions are randomly generated, each of which represents a different combination of desensitization strategies. A corresponding search space is constructed, and the fitness value of each initial solution is calculated according to the desensitization effect and data quality. The fitness value of each initial solution is compared, and the initial solution with the highest fitness value is taken as the current optimal solution. A random number in the range of 0 to 1 is randomly generated, and the position update amplitude of each initial solution is calculated. Then a group of probability values ​​are set. If the random number is less than the probability value, a local search is performed, otherwise, a global search is performed. When searching locally, the distance between the remaining solutions in the search space and the optimal solution is calculated, and local adjustments are made through the position update amplitude of each solution. At the same time, during the local search process, the spiral motion formula is used The spiral path is simulated to obtain the position changes of the remaining solutions when they make spiral motion around the optimal solution. When searching globally, a new solution is randomly generated in the entire search space. After the new solution is generated, the fitness value of the new solution is calculated, and it is judged whether the new solution is better than the existing solution. If it is better, the existing solution is replaced by the new solution, otherwise, a random solution is regenerated for judgment. In each iteration, it is checked whether a better solution is found. If a better solution is found, the current best solution is updated, otherwise, the current best solution is kept unchanged and the iteration is repeated. When the fitness value change is less than the predetermined threshold, the iteration is stopped, and the optimal desensitization strategy is output to replace the original desensitization strategy. The sensitive data is converted according to the optimal desensitization strategy, and the conversion information is monitored in real time.

[0066] The decision analysis module is used to comprehensively evaluate the desensitization results by integrating multiple factors; the data storage module is used to store and manage the desensitized data using encryption technology and access control; the audit update module is used to continuously track the use of desensitized data, conduct regular audits, and regularly update desensitization strategies and algorithms based on new security requirements and technological developments.

Claims

1. A data desensitization processing method, characterized in that: The specific steps of this processing method are as follows: Ⅰ. Collect raw data from multiple data sources and preprocess each set of data; II. Identify sensitive information in each group of data to develop corresponding desensitization strategies; III. Convert sensitive information according to desensitization strategies and optimize the conversion process in real time; IV. Evaluate the desensitization results and adjust the desensitization process based on the evaluation results; V. Store desensitized data and continuously track data usage; VI. Regularly audit processing procedures and update desensitization strategies.

2. A data desensitization processing method according to claim 1, characterized in that: The specific steps of preprocessing each group of data in step I are as follows: Q1.1: Collect each set of raw data collected, and identify and remove duplicate data information by comparing the unique identifiers of the data records. Then detect and locate the missing values ​​in each set of raw data, and fill the missing values ​​in the processed data by mean or median filling, and integrate the processed data into a set of data to be tested; Q1.2: Calculate the first quartile and the third quartile of the data set to be tested, and calculate the interquartile range based on the two sets of quartiles. Then determine the upper and lower limits of the abnormal data based on the interquartile range, and identify the abnormal data in the data set to be tested that is less than the lower limit or greater than the upper limit, and directly remove the detected abnormal data from the data set to be tested; Q1.3: After removing abnormal data, the data in the data set to be tested is formatted to unify the data formats, and then the Max-Min normalization method is used to standardize the processed data sets.

3. A data desensitization processing method according to claim 2, characterized in that: The specific steps for identifying sensitive information in each group of data in step II are as follows: Q2.1: Extract the text data from each group of pre-processed data to be tested. Q2.1: Remove stop words and punctuation marks from the text data, then convert the text data into numerical form using the TF-IDF method, and then design and build a recognition model based on the TensorFlow deep learning framework; Q2.2: Collect each set of historical data, and then label the sensitive information in each set of historical data automatically and manually. Divide the labeled data into a test set and a validation set, and pass the test set data into the recognition model. Q2.3: After the recognition model receives the test set data, it uses the forward propagation algorithm to pass each group of data through the model input layer, hidden layer, and output layer in turn, and outputs the prediction result. The cross entropy loss function is used to calculate the loss value between the prediction result and the actual value. Then, the back propagation algorithm is used to calculate the loss value and the gradient of each network layer of the model, and the Adam optimizer is used to adjust the model weights. Q2.4: After each round of training, the recognition model is evaluated using the validation set data to calculate the accuracy, recall, and F1-score indicators. The recognition model hyperparameter information is adjusted based on the evaluation results. The recognition module is trained and verified repeatedly until the model loss value converges to the preset range, and then the training is stopped. Q2.5: The latest data to be detected is input into the trained recognition model. The recognition model obtains the category probability of each data to be detected through the forward propagation algorithm. The output results of the model are then post-processed, and the data that exceeds the set threshold is extracted and marked as sensitive data. The identified sensitive data is then counted and analyzed.

4. A data desensitization processing method according to claim 3, characterized in that: The specific steps of the real-time optimization conversion process described in step III are as follows: Q3.1: Convert sensitive data according to the selected desensitization strategy, set optimization goals according to user needs and data conversion conditions, and construct the corresponding loss function, set the initial temperature, cooling rate and number of iterations. At the same time, take the current desensitization conversion process as the initial solution and calculate its loss value through the constructed loss function; Q3.2: In each round of iteration, the current desensitization conversion process is randomly perturbed to generate a new desensitization conversion process, and the loss value of the strategy is calculated. The loss value of the original conversion process is compared with the loss value of the new conversion process. If the loss value of the new solution is lower than the current solution, the new solution is accepted. If the loss value of the new solution is higher than the current solution, the new solution is accepted according to the Metropolis criterion. Q3.3: After each round of iteration, the current temperature is lowered according to the preset cooling rate, and random perturbations and process selections are repeated until the preset maximum number of iterations is reached. The iteration ends, and the optimal desensitization conversion process and its corresponding loss value are output. The conversion process and loss are saved to provide a basis for subsequent data storage and use.

5. A data desensitization processing system, used to implement the data desensitization processing method according to any one of claims 1 to 4, characterized in that: It includes collection and processing module, information identification module, strategy formulation module, data conversion module, result evaluation module, optimization and adjustment module, decision analysis module, data storage module and audit update module; The collection and processing module is used to automatically collect raw data from different data sources, and clean and standardize the collected data; The information identification module is used to analyze the processed data and identify sensitive information in each group of data; The strategy formulation module is used to formulate a desensitization strategy based on the identified sensitive information; The data conversion module is used to execute the desensitization strategy and convert sensitive information; The result evaluation module is used to collect desensitization data information in real time and evaluate the desensitization results; The optimization and adjustment module is used to monitor and adjust the desensitization process; The decision analysis module is used to comprehensively evaluate the desensitization results by integrating multiple factors; The data storage module is used to store and manage the desensitized data using encryption technology and access control; The audit update module is used to continuously track the use of desensitized data, conduct regular audits, and regularly update desensitization strategies and algorithms based on new security requirements and technological developments.

6. A data desensitization processing system according to claim 5, characterized in that: The specific steps of monitoring and adjusting the optimization and adjustment module are as follows: Q4.1: Randomly generate a set of initial solutions, each of which represents a different combination of desensitization strategies. Construct a corresponding search space, calculate the fitness value of each initial solution based on the desensitization effect and data quality, compare the fitness values ​​of each initial solution, and take the initial solution with the highest fitness value as the current optimal solution; Q4.2: Generate a random number in the range of 0 to 1 and calculate the position update amplitude of each initial solution. Then set a set of probability values. If the random number is less than the probability value, perform a local search. Otherwise, perform a global search. Q4.3: When searching locally, calculate the distance between the remaining solutions and the optimal solution in the search space, and make local adjustments based on the position update amplitude of each solution. Simultaneously, during the local search, simulate the spiral path using the spiral motion formula to obtain the position changes of the remaining solutions as they perform spiral motion around the optimal solution. Q4.4: When searching globally, a new solution is randomly generated in the entire search space. After the new solution is generated, the fitness value of the new solution is calculated, and it is judged whether the new solution is better than the existing solution. If it is better, the existing solution is replaced by the new solution. Otherwise, a new random solution is generated for judgment; Q4.5: In each iteration, check whether a better solution is found. If a better solution is found, update the current best solution. Otherwise, keep the current best solution unchanged and repeat the iteration. When the fitness value change is less than the predetermined threshold, stop the iteration and output the optimal desensitization strategy to replace the original desensitization strategy. Convert sensitive data according to the optimal desensitization strategy and monitor the conversion information in real time.

Images