Risk assessment method for medical institution health information sharing based on big data development

Abstract

The invention discloses a risk assessment method for medical institution health information sharing based on big data development, and particularly relates to the technical field of health information sharing. Data disturbance information, feature fuzzy information, precision loss information and data correlation information when a medical institution performs dynamic pseudonym operation on health information are acquired; an accumulative error risk assessment model is constructed, accumulative error risks introduced by dynamic pseudonym operation are quantified, potential influences of the pseudonym operation on overall analysis quality of health information data are determined, and an assessed accumulative error risk index is compared with a preset risk threshold value. And judging whether the health information subjected to the dynamic kana operation has a potential accumulative error risk or not, and after the accumulative error risk is detected, optimizing the configuration of a hash function by using a genetic algorithm so as to reduce errors introduced in the dynamic kana process to the greatest extent.

Description

Technical Field

[0001] The present invention relates to the technical field of health information sharing, and more specifically, to a risk assessment method for health information sharing of medical institutions developed based on big data. Background Art

[0002] Publication number CN115424733A discloses an active health management system and method based on intelligent algorithms for medical health big data, including: an application system for managing personal information, disease type data, indicators, and health risk screening; a management system for managing basic data and data sharing and exchange; an operation and maintenance service system for operation and maintenance management; and the application system, management system, and operation and maintenance service system are connected in sequence. The present invention makes full use of network communication technology, big data cloud computing platforms, and artificial intelligence algorithms as information management means to build a platform that is suitable for medical institutions at all levels and even individuals to operate simply. Through the "hospital-community-family / individual" H3C2H1 active health co-management service model of Doctors World, early detection, screening, and early warning of frequently occurring and chronic diseases, health risk screening, health management, health education, disease prevention, and other integrated services are promoted;

[0003] In the modern medical field, with the rapid development of information technology, medical institutions have accumulated a large amount of health data. The application of big data technology enables this data to be stored, processed and analyzed more efficiently, thus providing strong support for medical decision-making. However, in the process of sharing medical data, there are many risks and challenges, including data privacy leakage, data tampering and data abuse. These risks may not only damage the personal privacy of patients, but also affect the reputation and compliance of medical institutions.

[0004] When sharing personal health information, dynamic pseudonymization, as an advanced data protection technology, aims to reduce the probability of data re-identification and the risk of sensitive information exposure by using different pseudonyms for health information data in different usage scenarios or time periods. However, when medical institutions are subjected to large-scale and frequent unknown attacks and cannot stop sharing health information, they can only passively minimize the risk of sensitive information leakage by frequently updating pseudonyms. Pseudonymization of the same data point may cause potential cumulative error problems such as disturbance of health data, feature fuzziness, loss of precision, and destruction of correlation between health data. As the frequency of updating pseudonyms increases, these potential problems will gradually be exposed, thereby increasing the consequences of decreased analysis value and accuracy of shared health information, and even making it difficult to match the identity of the same patient in different data sets, affecting data integration and analysis across medical institutions. Summary of the invention

[0005] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides a risk assessment method for sharing health information of medical institutions developed based on big data to solve the problems raised in the above-mentioned background technology.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] The risk assessment method for sharing health information in medical institutions developed based on big data includes the following steps:

[0008] Step S1, obtaining data disturbance information, feature fuzziness information, precision loss information, and data correlation information when a medical institution performs a dynamic pseudonymization operation on health information;

[0009] Step S2, constructing a cumulative error risk assessment model to comprehensively assess whether the health information after the dynamic pseudonymization operation has a cumulative error risk to obtain a cumulative error risk assessment index;

[0010] Step S3, comparing the cumulative error risk assessment index with a preset cumulative error risk assessment index threshold to determine whether the health information after the dynamic pseudonymization operation has a cumulative error risk;

[0011] Step S4: When there is a risk of cumulative error in the health information after the dynamic pseudonymization operation, a genetic algorithm is introduced to optimize the hash function configuration to minimize the cumulative error risk of the dynamic pseudonymization process.

[0012] In a preferred embodiment, the data disturbance information includes a data disturbance coefficient, the feature fuzziness information includes a feature fuzziness coefficient, the precision loss information includes a precision loss coefficient, and the correlation information between data includes a correlation coefficient.

[0013] In a preferred embodiment, the logic for obtaining the data disturbance coefficient is as follows:

[0014] Get the user's original health information dataset D orig , calculate the information entropy H of the original health information data set orig , the expression is as follows Where x i represents the value of the i-th data point in the original health information dataset, p(x i ) represents x i The probability of occurrence in the original health information dataset;

[0015] Get the health information dataset D after dynamic pseudonymization operation pseudo , calculate the information entropy H of the health information dataset after the dynamic pseudonym operation pseudo , the expression is as follows In the formula represents the value of the i-th data point in the health information dataset after the dynamic pseudonymization operation, express The probability of occurrence in the original health information dataset;

[0016] Calculate the disturbance deviation value ΔH, the expression is as follows

[0017] Calculate the data disturbance coefficient sjr, the expression is as follows Where ΔH t,f Represents the disturbance deviation value of the fth scene in the tth period of time.

[0018] In a preferred embodiment, the acquisition logic of the characteristic fuzzy coefficient is as follows:

[0019] The user's original health information dataset D orig Perform feature extraction to obtain the original feature vector Vorig, and perform feature extraction on the health information data set after each dynamic pseudonymization operation to obtain the pseudonymized feature vector V;

[0020] The similarity xs between the original feature vector Vorig and the pseudonymized feature vector V is calculated according to the cosine similarity calculation method. The expression is as follows Where Vorig·V represents the dot product operation of the original eigenvector Vorig and the eigenvector V after pseudonymization, ||Vorig|| represents the norm of the original eigenvector Vorig, and ||V|| represents the norm of the eigenvector V after pseudonymization;

[0021] Calculate the characteristic fuzzy value mh, the expression is as follows: mh = 1-xs;

[0022] Calculate the characteristic fuzzy coefficient tzm, the expression is as follows Wherein j represents the sequence number of the dynamic pseudonymization operation, j={1,2,...,m}, and m is a positive integer.

[0023] In a preferred embodiment, the logic for obtaining the precision loss coefficient is as follows:

[0024] Get the health information dataset D after dynamic pseudonymization operation pseudo , construct the accuracy loss function L(μ,σ 2 |D pseudo ), the expression is as follows Where μ represents the expected mean initial accuracy loss, σ 2 represents the expected initial accuracy loss variance;

[0025] For the accuracy loss function L(μ,σ 2 |Dpseudo ) Take the logarithm operation and get the logarithmic accuracy loss function, the expression is as follows

[0026] For μ and σ 2 Derivatives are taken separately and their derivatives are set equal to 0 to obtain the final mean accuracy loss And the final accuracy loss variance

[0027] Calculate the precision loss coefficient jdx, the expression is as follows

[0028] In a preferred embodiment, the logic for obtaining the correlation coefficient is as follows:

[0029] Get the user's original health information dataset D orig and the health information dataset D after dynamic pseudonymization pseudo , respectively for the original health information dataset D orig and the health information dataset D after dynamic pseudonymization pseudo The data points in are sorted from small to large to get the corresponding ranking;

[0030] Calculate ranking difference Where Rx i represents the ranking of the i-th data point in the original health information dataset, represents the ranking of the i-th data point in the health information dataset after the dynamic pseudonymization operation;

[0031] Calculate the correlation coefficient xgx, the expression is as follows

[0032] In a preferred embodiment, in step S2, a cumulative error risk assessment model is constructed to comprehensively assess whether the health information after the dynamic pseudonymization operation has a cumulative error risk to obtain a cumulative error risk assessment index, which is as follows:

[0033] The data disturbance coefficient, feature fuzzy coefficient, precision loss coefficient, and correlation coefficient are normalized, and a cumulative error risk assessment model is constructed based on the normalized data disturbance coefficient, feature fuzzy coefficient, precision loss coefficient, and correlation coefficient to generate a cumulative error risk assessment index ljw. The model is based on the following formula: Wherein a1, a2, a3, and a4 represent the preset proportional coefficients of the data disturbance coefficient, feature fuzzy coefficient, precision loss coefficient, and correlation coefficient, respectively, and a1, a2, a3, and a4 are all greater than 0.

[0034] In a preferred embodiment, in step S3, the cumulative error risk assessment index is compared with a preset cumulative error risk assessment index threshold to determine whether the health information after the dynamic pseudonymization operation has a cumulative error risk, as follows:

[0035] If the cumulative error risk assessment index is greater than the cumulative error risk assessment index threshold, a cumulative error risk signal is generated;

[0036] If the cumulative error risk assessment index is less than or equal to the cumulative error risk assessment index threshold, a cumulative error escape signal is generated without further optimizing the hash function configuration.

[0037] In a preferred embodiment, in step S4, when there is a cumulative error risk in the health information after the dynamic pseudonymization operation, a genetic algorithm is introduced to optimize the hash function configuration so as to minimize the cumulative error risk of the dynamic pseudonymization process, as follows:

[0038] Step S41, defining chromosomes of the genetic algorithm, encoding the configuration parameters of the hash function and using them as the encoding representation of the chromosome, where one chromosome represents a specific combination of configuration parameters of the hash function;

[0039] The configuration parameters of the hash function include but are not limited to salt value, hash algorithm selection, and output length;

[0040] Step S42, randomly generating a number of chromosomes based on a pseudo-random number generator, and forming an initial population with the number of randomly generated chromosomes;

[0041] Step S43, calculate the fitness value F, the expression is as follows In the formula, F represents the fitness value F, ljw represents the cumulative error risk assessment index, and ljw z represents the cumulative error risk assessment index threshold, normalizing the fitness values ​​of all chromosomes so that the fitness values ​​are in [0,1];

[0042] Step S44, chromosome selection;

[0043] Step S45, crossover recombination;

[0044] Step S46, mutation operation;

[0045] Step S47, iterative evolution: determine whether the iteration time reaches the preset iteration time threshold. If the iteration time is equal to the iteration time threshold, end the iteration process and output the optimal solution chromosome. If the iteration time is less than the iteration time threshold, repeat steps S43, S44, S45 and S46.

[0046] The optimal solution chromosome is decoded to configure the parameters of the final hash function, so that the cumulative error risk of the dynamic pseudonymization process is minimized.

[0047] Technical effects and advantages of the present invention:

[0048] 1. The present invention obtains data disturbance information, feature fuzzy information, precision loss information and data correlation information when medical institutions perform dynamic pseudonymization operations on health information, constructs a cumulative error risk assessment model, quantifies the cumulative error risk introduced by the dynamic pseudonymization operation, helps determine the potential impact of the pseudonymization operation on the overall analysis quality of health information data, compares the assessed cumulative error risk index with a preset risk threshold, and determines whether the health information after the dynamic pseudonymization operation has potential cumulative error risks. After detecting the cumulative error risk, a genetic algorithm is used to optimize the configuration of the hash function to minimize the errors introduced in the dynamic pseudonymization process, which can protect the privacy of health information while retaining the analytical value of the data to the greatest extent. By real-time risk assessment and optimization using a genetic algorithm, it is ensured that during the data sharing process, even in the face of frequent pseudonym updates and large-scale attacks, the integrity and availability of the data are still maintained, thereby improving the security and effectiveness of medical data and reducing the potential negative impact caused by frequent pseudonymization operations. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to facilitate understanding by those skilled in the art, the present invention is further described below in conjunction with the accompanying drawings;

[0050] Figure 1 Schematic diagram of the structure of the method of the embodiment of the present invention. DETAILED DESCRIPTION

[0051] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0052] Example: Figure 1 The present invention provides a risk assessment method for sharing health information of medical institutions based on big data, including the following steps:

[0053] Step S1, obtaining data disturbance information, feature fuzziness information, precision loss information, and data correlation information when a medical institution performs a dynamic pseudonymization operation on health information;

[0054] Step S2, constructing a cumulative error risk assessment model to comprehensively assess whether the health information after the dynamic pseudonymization operation has a cumulative error risk to obtain a cumulative error risk assessment index;

[0055] Step S3, comparing the cumulative error risk assessment index with a preset cumulative error risk assessment index threshold to determine whether the health information after the dynamic pseudonymization operation has a cumulative error risk;

[0056] Step S4, when there is a cumulative error risk in the health information after the dynamic pseudonymization operation, a genetic algorithm is introduced to optimize the hash function configuration so as to minimize the cumulative error risk of the dynamic pseudonymization process;

[0057] Medical institutions usually store personal basic data, medical history data, diagnostic data and other health information data, and these personal health information are usually shared in different usage scenarios or time periods. In order to reduce the probability of data being re-identified and reduce the risk of sensitive information exposure, medical institutions usually perform dynamic pseudonymization on health information. However, when medical institutions are subjected to large-scale and high-frequency unknown attacks and cannot stop sharing health information, they can only passively minimize the risk of sensitive information leakage by frequently updating pseudonyms. In order to accurately identify whether the health information after dynamic pseudonymization has potential cumulative error problems such as disturbance, feature fuzziness, precision loss, and destruction of correlation between health data, therefore, in step S1, by obtaining data disturbance information, feature fuzziness information, precision loss information and data correlation information when the medical institution performs dynamic pseudonymization on health information;

[0058] The data disturbance information includes a data disturbance coefficient, the feature fuzziness information includes a feature fuzziness coefficient, the precision loss information includes a precision loss coefficient, and the correlation information between data includes a correlation coefficient;

[0059] The acquired data disturbance coefficient, feature fuzzy coefficient, precision loss coefficient, and correlation coefficient are marked as sjr, tzm, jdx, and xgx respectively;

[0060] The data perturbation coefficient is an important indicator used to measure the data perturbation that may be introduced after the dynamic pseudonymization operation. It reflects the changes in the health information after the perturbation compared with the original health information data. A larger data perturbation coefficient indicates that the greater the degree of data perturbation introduced after the dynamic pseudonymization operation, the greater the risk of potential cumulative error in the health information. Conversely, a smaller data perturbation coefficient indicates that the smaller the degree of data perturbation introduced after the dynamic pseudonymization operation, the smaller the probability of potential cumulative error risk in the health information.

[0061] The logic for obtaining the data perturbation coefficient is as follows:

[0062] Get the user's original health information dataset D orig , calculate the information entropy H of the original health information data set orig , the expression is as follows Where x i represents the value of the i-th data point in the original health information dataset, p(x i ) represents x i The probability of occurrence in the original health information dataset;

[0063] Get the health information dataset D after dynamic pseudonymization operation pseudo , calculate the information entropy H of the health information dataset after the dynamic pseudonym operation pseudo , the expression is as follows In the formula represents the value of the i-th data point in the health information dataset after the dynamic pseudonymization operation, Represents x i ～ The probability of occurrence in the original health information dataset;

[0064] Calculate the disturbance deviation value ΔH, the expression is as follows

[0065] Calculate the data disturbance coefficient sjr, the expression is as follows Where ΔH t,f represents the disturbance deviation value of the fth scene in the tth period;

[0066] It should be noted that, since personal health information is usually shared in different usage scenarios or time periods, the impact of dynamic pseudonymization on data disturbance is measured by calculating the disturbance deviation value of personal health information in different scenarios within the same time period and accumulating different time periods. The time period can be set according to the actual situation, for example, 0.5h, 1h, etc., and the time period can be appropriately adjusted according to the scale and frequency of attacks on medical institutions. For example, when the scale and frequency of attacks on medical institutions are large, the time period can be appropriately shortened. The usage scenarios include cross-institutional data sharing, cross-regional data sharing, multi-department collaboration and other practical usage scenarios;

[0067] The feature fuzziness coefficient is an indicator used to measure the degree to which the features of health information data are gradually blurred or distorted after multiple dynamic pseudonymization operations. This indicator can help quantify the risk of health information data losing its original features during multiple pseudonymization processes, thereby evaluating the availability and accuracy of health information data. A higher feature fuzziness coefficient indicates that the original feature information of the health information data is significantly blurred or distorted, which may lead to potential cumulative error risks in health information, while a lower feature fuzziness coefficient indicates that the features of the health information data remain basically unchanged after pseudonymization, and the probability of potential cumulative error risks in health information is lower;

[0068] The logic for obtaining the feature fuzzy coefficient is as follows:

[0069] The user's original health information dataset D orig Perform feature extraction to obtain the original feature vector Vorig, and perform feature extraction on the health information data set after each dynamic pseudonymization operation to obtain the pseudonymized feature vector V;

[0070] The similarity xs between the original feature vector Vorig and the pseudonymized feature vector V is calculated according to the cosine similarity calculation method. The expression is as follows Where Vorig·V represents the dot product operation of the original eigenvector Vorig and the eigenvector V after pseudonymization, ||Vorig|| represents the norm of the original eigenvector Vorig, and ||V|| represents the norm of the eigenvector V after pseudonymization;

[0071] Calculate the characteristic fuzzy value mh, the expression is as follows: mh = 1-xs;

[0072] It should be noted that the closer the feature fuzziness value mh is to 0, the higher the feature similarity between the pseudonymized health information dataset and the original health information dataset, the lower the feature fuzziness, and the better the data fidelity; the closer the feature fuzziness value mh is to 1, the lower the feature similarity between the pseudonymized health information dataset and the original health information dataset, the higher the feature fuzziness, and the more serious the data distortion;

[0073] Calculate the characteristic fuzzy coefficient tzm, the expression is as follows Where j represents the order number of the dynamic pseudonymization operation, j = {1, 2, ..., m}, m is a positive integer;

[0074] The precision loss coefficient is used to measure the precision loss of health information data after dynamic pseudonymization. Due to the precision limitations of the hardware computer equipment on the system when processing floating-point numbers, there may be slight differences. The precision loss caused by a single dynamic pseudonymization operation may be negligible, but frequent dynamic pseudonymization operations may cause the precision loss to gradually accumulate, thereby affecting the final precision of the health information data, increasing the value of shared health information analysis, reducing accuracy and other consequences. A larger precision loss coefficient indicates that the precision loss of health information data after multiple dynamic pseudonymization operations is more serious, which may lead to potential cumulative error risks in health information. Conversely, a smaller precision loss coefficient indicates that the precision loss of health information data after multiple dynamic pseudonymization operations is in line with the expected loss, and the probability of potential cumulative error risks in health information is smaller.

[0075] The logic for obtaining the precision loss coefficient is as follows:

[0076] Get the health information dataset D after dynamic pseudonymization operation pseudo , construct the accuracy loss function L(μ,σ 2 |D pseudo ), the expression is as follows Where μ represents the expected mean initial accuracy loss, σ 2 represents the expected initial accuracy loss variance;

[0077] For the accuracy loss function L(μ,σ 2 |D pseudo ) Take the logarithm operation and get the logarithmic accuracy loss function, the expression is as follows

[0078] For μ and σ 2 Derivatives are taken separately and their derivatives are set equal to 0 to obtain the final mean accuracy loss And the final accuracy loss variance

[0079] Calculate the precision loss coefficient jdx, the expression is as follows

[0080] In medical data analysis, there is usually a significant correlation between multiple health indicators of patients. For example, there may be a strong correlation between blood sugar levels and insulin levels. If these data are pseudonymized multiple times, this correlation may be weakened or disappear, making it impossible for doctors to accurately grasp the overall trend of the disease during analysis and diagnosis. In this case, the results of data analysis may no longer be reliable and may even affect the patient's treatment plan.

[0081] The correlation coefficient is used to measure whether the correlation between the user's health information indicator data and other indicators has changed and the degree of change after the pseudonymization operation. A smaller correlation coefficient means that the correlation between the user's health information indicator data has changed significantly after multiple dynamic pseudonymization operations, which may lead to potential cumulative error risks in health information. Conversely, a larger correlation coefficient means that the correlation between the user's health information indicator data has changed less or even not changed after multiple dynamic pseudonymization operations, and the probability of potential cumulative error risks in health information is smaller.

[0082] The logic for obtaining the correlation coefficient is as follows:

[0083] Get the user's original health information dataset D orig and the health information dataset D after dynamic pseudonymization pseudo , respectively for the original health information dataset D orig and the health information dataset D after dynamic pseudonymization pseudo The data points in are sorted from small to large to get the corresponding ranking;

[0084] It should be noted that if the data points have the same value, the average rank is used for processing;

[0085] Calculate ranking difference Where Rx i represents the ranking of the i-th data point in the original health information dataset, represents the ranking of the i-th data point in the health information dataset after the dynamic pseudonymization operation;

[0086] Calculate the correlation coefficient xgx, the expression is as follows

[0087] In step S2, a cumulative error risk assessment model is constructed to comprehensively assess whether the health information after the dynamic pseudonymization operation has a cumulative error risk and obtain a cumulative error risk assessment index, which is as follows:

[0088] The data disturbance coefficient, feature fuzzy coefficient, precision loss coefficient, and correlation coefficient are normalized, and a cumulative error risk assessment model is constructed based on the normalized data disturbance coefficient, feature fuzzy coefficient, precision loss coefficient, and correlation coefficient to generate a cumulative error risk assessment index ljw. The model is based on the following formula: Wherein a1, a2, a3, and a4 represent the preset proportional coefficients of the data disturbance coefficient, feature fuzziness coefficient, precision loss coefficient, and correlation coefficient, respectively, and a1, a2, a3, and a4 are all greater than 0;

[0089] It can be seen from the above calculation expression that the larger the data disturbance coefficient, the larger the feature fuzziness coefficient, the larger the precision loss coefficient, and the smaller the correlation coefficient, the larger the cumulative error risk assessment index, indicating that the probability of potential cumulative error risk in health information is greater; conversely, the smaller the data disturbance coefficient, the smaller the feature fuzziness coefficient, the smaller the precision loss coefficient, and the larger the correlation coefficient, the smaller the cumulative error risk assessment index, indicating that the probability of potential cumulative error risk in health information is smaller;

[0090] In step S3, the cumulative error risk assessment index is compared with a preset cumulative error risk assessment index threshold to determine whether the health information after the dynamic pseudonymization operation has a cumulative error risk, as follows:

[0091] If the cumulative error risk assessment index is greater than the cumulative error risk assessment index threshold, it indicates that the health information has a potential cumulative error risk, and a cumulative error risk signal is generated. A genetic algorithm can be introduced to optimize the hash function configuration to minimize the cumulative error risk of the dynamic pseudonymization process;

[0092] If the cumulative error risk assessment index is less than or equal to the cumulative error risk assessment index threshold, it indicates that the health information does not have a potential cumulative error risk, and a cumulative error escape signal is generated. There is no need to further optimize the hash function configuration, and the existing data processing configuration can be maintained;

[0093] In the dynamic pseudonymization operation, hash functions play a key role, mainly used to disguise and protect sensitive information. Hash functions can dynamically generate different hash values ​​according to different times, locations, users or other conditions. This dynamic nature means that even if the same data is processed multiple times in different occasions, the generated hash values ​​are different, further enhancing the protection strength of health information data when it is shared. However, when there is a potential risk of cumulative errors in health information, the hash function may have the following impacts:

[0094] The configuration of the hash function (such as the selected hash algorithm, the use of salt values, the output length, etc.) directly affects the degree of disturbance of the health information data. An overly aggressive or inappropriate hash configuration may introduce too much noise, resulting in increased disturbance of the health information data, which gradually destroys the correlation and structure between the health information data, and increases the risk of cumulative errors;

[0095] Too short a hash output may cause hash conflicts between different health information data items, resulting in loss of distinction. With multiple dynamic pseudonymization operations, this fuzzification will gradually accumulate, resulting in a decrease in the accuracy and effectiveness of health information;

[0096] The various indicators of health information data often have a certain correlation, but improper hash function configuration may weaken this correlation. For example, an overly dynamic or overly complex hash function configuration may break the original correlation or even introduce random noise that is irrelevant to the characteristics of health information data, causing the relationship between health indicators to be misleading.

[0097] Therefore, in step S4, when there is a cumulative error risk in the health information after the dynamic pseudonymization operation, a genetic algorithm is introduced to optimize the hash function configuration so as to minimize the cumulative error risk of the dynamic pseudonymization process, as follows:

[0098] Step S41, defining chromosomes of the genetic algorithm, encoding the configuration parameters of the hash function and using them as the encoding representation of the chromosome, where one chromosome represents a specific combination of configuration parameters of the hash function;

[0099] The configuration parameters of the hash function include but are not limited to salt value, hash algorithm selection, and output length;

[0100] The encoding method includes binary encoding, real number encoding, etc., and the binary encoding method is used for explanation in this application. Binary encoding: convert each configuration parameter into binary form. For example, the salt value size is represented by 8 bits of binary, such as `00101010`, the hash algorithm selection is represented by 2 bits of binary, such as `01`, indicating the hash algorithm SHA-256, and the output length is represented by 4 bits of binary, such as `1010`, indicating 16 bytes;

[0101] The chromosome generated according to the binary encoding method can be expressed as `00101010-01-1010`;

[0102] Step S42, randomly generating a number of chromosomes based on a pseudo-random number generator, and forming an initial population with the number of randomly generated chromosomes;

[0103] Step S43, calculate the fitness value F, the expression is as follows In the formula, F represents the fitness value F, ljw represents the cumulative error risk assessment index, and ljw z represents the cumulative error risk assessment index threshold, normalizing the fitness values ​​of all chromosomes so that the fitness values ​​are in [0,1];

[0104] Step S44, chromosome selection, using a selection strategy to select a chromosome with a higher fitness value, and using the chromosome with a higher fitness value as a parent;

[0105] It should be noted that selection strategies include roulette selection strategies, tournament selection strategies, etc.;

[0106] Step S45, crossover recombination, exploring new solution space through crossover operation between parents, and performing crossover recombination operation on the population after chromosome selection. Crossover recombination refers to randomly selecting two parent chromosomes according to a certain recombination probability, exchanging them at multiple random positions of the chromosomes, and generating new offspring chromosomes, thereby increasing the genetic diversity of the population and providing new possibilities, which is helpful to find the global optimal solution;

[0107] Step S46, mutation operation: the mutation operation also increases the diversity of the population and introduces randomness, thereby helping to avoid falling into a local optimal solution. The mutation operation causes the chromosome to mutate with a certain probability, that is, to perform bit flipping at a random position of the chromosome, for example, changing the third bit of the salt value part of the chromosome `00101010-01-1010` to 0;

[0108] Step S47, iterative evolution: determine whether the iteration time reaches the preset iteration time threshold. If the iteration time is equal to the iteration time threshold, end the iteration process and output the optimal solution chromosome. If the iteration time is less than the iteration time threshold, repeat steps S43, S44, S45 and S46.

[0109] The optimal solution chromosome is decoded to configure the parameters of the final hash function, so as to minimize the cumulative error risk of the dynamic pseudonymization process;

[0110] The present invention obtains data disturbance information, feature fuzzy information, precision loss information and data correlation information when medical institutions perform dynamic pseudonymization operations on health information, constructs a cumulative error risk assessment model, quantifies the cumulative error risk introduced by the dynamic pseudonymization operation, helps determine the potential impact of the pseudonymization operation on the overall analysis quality of health information data, compares the assessed cumulative error risk index with a preset risk threshold, and determines whether the health information after the dynamic pseudonymization operation has potential cumulative error risks. After detecting the cumulative error risk, a genetic algorithm is used to optimize the configuration of the hash function to minimize the errors introduced in the dynamic pseudonymization process, which can protect the privacy of health information while retaining the analytical value of the data to the greatest extent. By real-time risk assessment and optimization using a genetic algorithm, it is ensured that during the data sharing process, even in the face of frequent pseudonym updates and large-scale attacks, the integrity and availability of the data are still maintained, thereby improving the security and effectiveness of medical data and reducing the potential negative impact caused by frequent pseudonymization operations.

[0111] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0112] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0113] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A risk assessment method for sharing health information in medical institutions based on big data, characterized by: The steps include: Step S1, obtaining data disturbance information, feature fuzziness information, precision loss information, and data correlation information when a medical institution performs a dynamic pseudonymization operation on health information; Step S2, constructing a cumulative error risk assessment model to comprehensively assess whether the health information after the dynamic pseudonymization operation has a cumulative error risk to obtain a cumulative error risk assessment index; Step S3, comparing the cumulative error risk assessment index with a preset cumulative error risk assessment index threshold to determine whether the health information after the dynamic pseudonymization operation has a cumulative error risk; Step S4: When there is a risk of cumulative error in the health information after the dynamic pseudonymization operation, a genetic algorithm is introduced to optimize the hash function configuration to minimize the cumulative error risk of the dynamic pseudonymization process.

2. The risk assessment method for sharing health information of medical institutions based on big data development according to claim 1 is characterized by: The data disturbance information includes a data disturbance coefficient, the feature fuzziness information includes a feature fuzziness coefficient, the precision loss information includes a precision loss coefficient, and the correlation information between data includes a correlation coefficient.

3. The risk assessment method for sharing health information of medical institutions based on big data development according to claim 2 is characterized by: The logic for obtaining the data perturbation coefficient is as follows: Get the user's original health information dataset D orig , calculate the information entropy H of the original health information data set orig , the expression is as follows Where x i represents the value of the i-th data point in the original health information dataset, p(x i ) represents x i The probability of occurrence in the original health information dataset; Get the health information dataset D after dynamic pseudonymization operation pseudo , calculate the information entropy H of the health information dataset after the dynamic pseudonym operation pseudo , the expression is as follows In the formula represents the value of the i-th data point in the health information dataset after the dynamic pseudonymization operation, express The probability of occurrence in the original health information dataset; Calculate the disturbance deviation value ΔH, the expression is as follows Calculate the data disturbance coefficient sjr, the expression is as follows Where ΔH t,f Represents the disturbance deviation value of the fth scene in the tth period of time.

4. The risk assessment method for sharing health information of medical institutions based on big data development according to claim 2 is characterized by: The logic for obtaining the feature fuzzy coefficient is as follows: The user's original health information dataset D orig Perform feature extraction to obtain the original feature vector Vorig, and perform feature extraction on the health information data set after each dynamic pseudonymization operation to obtain the pseudonymized feature vector V; The similarity xs between the original feature vector Vorig and the pseudonymized feature vector V is calculated according to the cosine similarity calculation method. The expression is as follows Where Vorig·V represents the dot product operation of the original eigenvector Vorig and the eigenvector V after pseudonymization, ||Vorig|| represents the norm of the original eigenvector Vorig, and ||V|| represents the norm of the eigenvector V after pseudonymization; Calculate the characteristic fuzzy value mh, the expression is as follows: mh = 1-xs; Calculate the characteristic fuzzy coefficient tzm, the expression is as follows Wherein j represents the sequence number of the dynamic pseudonymization operation, j={1,2,...,m}, and m is a positive integer.

5. The risk assessment method for sharing health information of medical institutions based on big data development according to claim 2 is characterized by: The logic for obtaining the precision loss coefficient is as follows: Get the health information dataset D after dynamic pseudonymization operation pseudo , construct the accuracy loss function L(μ,σ 2 |D pseudo ), the expression is as follows Where μ represents the expected mean initial accuracy loss, σ 2 represents the expected initial accuracy loss variance; For the accuracy loss function L(μ,σ 2 |D pseudo ) Take the logarithmic operation and get the logarithmic accuracy loss function, the expression is as follows For μ and σ 2 Derivatives are taken separately and their derivatives are set equal to 0 to obtain the final mean accuracy loss And the final accuracy loss variance Calculate the precision loss coefficient jdx, the expression is as follows 6. The risk assessment method for sharing health information of medical institutions based on big data development according to claim 2 is characterized by: The logic for obtaining the correlation coefficient is as follows: Get the user's original health information dataset D orig and the health information dataset D after dynamic pseudonymization pseudo , respectively for the original health information dataset D orig and the health information dataset D after dynamic pseudonymization pseudo The data points in are sorted from small to large to get the corresponding ranking; Calculate ranking difference Where Rx i represents the ranking of the i-th data point in the original health information dataset, represents the ranking of the i-th data point in the health information dataset after the dynamic pseudonymization operation; Calculate the correlation coefficient xgx, the expression is as follows 7. The risk assessment method for sharing health information of medical institutions based on big data development according to claim 2 is characterized by: In step S2, a cumulative error risk assessment model is constructed to comprehensively assess whether the health information after the dynamic pseudonymization operation has a cumulative error risk and obtain a cumulative error risk assessment index, which is as follows: The data disturbance coefficient, feature fuzzy coefficient, precision loss coefficient, and correlation coefficient are normalized, and a cumulative error risk assessment model is constructed based on the normalized data disturbance coefficient, feature fuzzy coefficient, precision loss coefficient, and correlation coefficient to generate a cumulative error risk assessment index ljw. The model is based on the following formula: Wherein a1, a2, a3, and a4 represent the preset proportional coefficients of the data disturbance coefficient, feature fuzzy coefficient, precision loss coefficient, and correlation coefficient, respectively, and a1, a2, a3, and a4 are all greater than 0.

8. The risk assessment method for sharing health information of medical institutions based on big data development according to claim 7 is characterized by: In step S3, the cumulative error risk assessment index is compared with a preset cumulative error risk assessment index threshold to determine whether the health information after the dynamic pseudonymization operation has a cumulative error risk, as follows: If the cumulative error risk assessment index is greater than the cumulative error risk assessment index threshold, a cumulative error risk signal is generated; If the cumulative error risk assessment index is less than or equal to the cumulative error risk assessment index threshold, a cumulative error escape signal is generated without further optimizing the hash function configuration.

9. The risk assessment method for sharing health information of medical institutions based on big data development according to claim 8 is characterized by: Step S4, when there is a cumulative error risk in the health information after the dynamic pseudonymization operation, a genetic algorithm is introduced to optimize the hash function configuration so as to minimize the cumulative error risk of the dynamic pseudonymization process, as follows: Step S41, defining chromosomes of the genetic algorithm, encoding the configuration parameters of the hash function and using them as the encoding representation of the chromosome, where one chromosome represents a specific combination of configuration parameters of the hash function; The configuration parameters of the hash function include but are not limited to salt value, hash algorithm selection, and output length; Step S42, randomly generating a number of chromosomes based on a pseudo-random number generator, and forming an initial population with the number of randomly generated chromosomes; Step S43, calculate the fitness value F, the expression is as follows In the formula, F represents the fitness value F, ljw represents the cumulative error risk assessment index, and ljw z represents the cumulative error risk assessment index threshold, normalizing the fitness values ​​of all chromosomes so that the fitness values ​​are in [0,1]; Step S44, chromosome selection; Step S45, crossover recombination; Step S46, mutation operation; Step S47, iterative evolution: determine whether the iteration time reaches the preset iteration time threshold. If the iteration time is equal to the iteration time threshold, end the iteration process and output the optimal solution chromosome. If the iteration time is less than the iteration time threshold, repeat steps S43, S44, S45 and S46. The optimal solution chromosome is decoded to configure the parameters of the final hash function, so that the cumulative error risk of the dynamic pseudonymization process is minimized.

Images