Medical data sharing system and method based on security mechanism
By adopting desensitization processing, classification tag processing, user role and permission management, sandbox environment isolation and risk assessment in the medical data sharing system, the problem of low data security in the medical data sharing system is solved and efficient privacy is achieved Protect and secure sharing.
Patent Information
- Application Number
- CN202411227887.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-03
- Publication Date
- 2024-11-15
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
When existing medical data sharing systems face complex attack methods and the sharing needs of multiple parties, their data security is low, especially the keys are easily cracked, making it difficult to effectively improve the security of medical data during the sharing process.
Adopt a medical data sharing system based on security mechanisms, including collection modules, governance modules, access restriction modules, sandbox modules and risk assessment modules, through desensitization processing, classification tag processing, user role and permission management, sandbox environment isolation and risk Assessment to ensure the security and legality of the data.
Effectively protect patient privacy, improve data usage efficiency and quality, realize refined permission management, reduce data leakage risks, enhance the security and controllability of medical data, promptly discover and solve potential safety hazards, and reduce system maintenance costs.
Smart Images

Figure CN118965439A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of privacy protection technology, and in particular to a medical data sharing system and method based on a security mechanism. Background Art
[0002] With the rapid development of information technology and medical technology, the generation, storage and sharing of medical data have become more and more common. Medical data involves extremely sensitive personal privacy information. Therefore, how to achieve effective sharing of medical data while ensuring data privacy and security has become a major challenge facing the medical industry. Traditional medical data sharing mechanisms mainly rely on encryption technology, access control strategies, and data desensitization processing. However, these methods have many shortcomings when dealing with increasingly complex attack methods and sharing needs involving multiple parties. For example, although data encryption can protect the security of data during transmission to a certain extent, once the key is cracked.
[0003] Chinese Patent Publication No.: CN112836240A discloses a blockchain-based electronic medical data security sharing method, system and medium, the method includes: encrypting the patient's electronic medical data, obtaining the patient's electronic medical data ciphertext and storing it in the cloud service platform; initiating an encryption request to obtain the electronic medical data to the patient client; decrypting the encryption request and triggering the smart contract; notifying the patient client to generate a re-encryption conversion key; encrypting the re-encryption conversion key through the public key of the cloud service platform and transmitting it to the cloud service platform; obtaining the re-encryption conversion key to re-encrypt the patient's electronic medical data ciphertext, generating a re-encrypted ciphertext, and transmitting it to the medical institution client; decrypting the re-encrypted ciphertext to obtain the patient's electronic medical data; submitting the transaction between the patient client and the medical institution client to the verification node, and publishing the verification node to the blockchain. However, the key used in this scheme is easy to be cracked, and it is difficult to improve the security of medical data in the sharing process. Summary of the invention
[0004] To this end, the present invention provides a medical data sharing system based on a security mechanism to overcome the problem of low security of medical data in the sharing process in the prior art.
[0005] To achieve the above object, the present invention provides a medical data sharing system based on a security mechanism, comprising:
[0006] A collection module, used to collect medical data;
[0007] The governance module is used to build a medical desensitization model and a classification label model, and perform desensitization and classification label processing on medical data to obtain a classified data set;
[0008] Access restriction module, used to set user roles and basic user permissions, and also used to modify, adjust and optimize user roles and basic user permissions;
[0009] The sandbox module is used to establish a sandbox environment according to the user's basic permissions and import the classification data set corresponding to the sandbox environment into the sandbox environment;
[0010] A risk assessment module, used to analyze the security of the sandbox module based on the risk assessment model;
[0011] The environment destruction module is used to destroy the sandbox environment and recycle the resources in the sandbox environment.
[0012] Furthermore, the governance module includes:
[0013] The data desensitization unit is used to build a medical desensitization model and perform desensitization processing on medical data to obtain desensitized data;
[0014] The data labeling unit is used to build a classification labeling model and perform classification labeling on the desensitized data to obtain a classified data set.
[0015] Furthermore, when constructing a medical desensitization model, the data desensitization unit divides the preset desensitization database into a 75% desensitization training set and a 25% desensitization test set, inputs the desensitization training set into the convolutional neural network model for training, obtains the trained convolutional neural network model, and inputs the desensitization test set into the convolutional neural network model, tests the convolutional network neural model, obtains the desensitization accuracy rate Q, compares the desensitization accuracy rate Q with the preset desensitization accuracy rate Q0, judges the training compliance of the convolutional neural network model according to the comparison result, and outputs the convolutional neural network model according to the judgment result, wherein:
[0016] When Q≥Q0, the convolutional neural network model is judged to have met the training standards, and the convolutional neural network model is output as a medical desensitization model;
[0017] When Q < Q0, it is determined that the convolutional neural network model training does not meet the standards, and the convolutional neural network model parameters are iteratively optimized until the convolutional neural network model training meets the standards. The cross entropy loss function is set to the multi-classification cross entropy loss function. set up y is the actual measurement result in the preset desensitized database, is the predicted value output by the convolutional neural network model, yi is the true label of the i-th sample, is the probability that the model predicts that the sample is a positive class, N is the total number of samples, C is the total number of categories, yi, c are binary indicators, c is the probability that the model predicts that the i-th sample belongs to category c;
[0018] The data desensitization unit also inputs the medical data into the medical desensitization model for desensitization processing to obtain desensitized data;
[0019] The data labeling unit, when constructing the classification labeling model, trains the decision tree model according to the historical classification database, divides the historical classification database into 80% classification training set and 20% classification test set, inputs the classification training set into the decision tree model, and inputs the classification test set into the trained decision tree model, obtains the classification accuracy E, compares the classification accuracy E with the preset classification accuracy E0, judges the compliance of the decision tree model training according to the comparison result, and outputs the decision tree model according to the judgment result, wherein:
[0020] When E≥E0, the decision tree model training is judged to have met the standard, and the decision tree model is output as a classification label model;
[0021] When E<E0, it is determined that the decision tree model training does not meet the standard, and the decision tree model is repeatedly trained and the parameters are optimized until the decision tree model training meets the standard;
[0022] The data labeling module also inputs the desensitized data into the classification labeling model for classification labeling processing to obtain a classification data set.
[0023] Furthermore, the access restriction module includes:
[0024] User definition unit, used to set user roles according to user attributes;
[0025] User authority setting unit, used to set basic user authority according to user roles;
[0026] A role modification unit is used to judge the validity of the user role according to the user's access time and geographical location, modify the user role according to the judgment result, and adjust the user's basic permissions according to the modification result;
[0027] The cycle optimization unit is used to obtain the number of invalid user roles within a preset cycle, and compare the number of invalid user roles with the preset invalid number, judge the invalid frequency of the user role according to the comparison result, and optimize the process of correcting the user role according to the judgment result.
[0028] Furthermore, the user definition unit obtains user attributes and sets the user role according to the user attributes, wherein:
[0029] When the user attribute is a target administrator, the user role is set to a medical data administrator;
[0030] When the user attribute is a target physician, the user role is set as a medical data editor;
[0031] When the user attribute is a target patient, the user role is set to a medical data reader;
[0032] When the user attribute is untargeted, the user role is set as a malicious user, and the user attribute and user role setting window are pushed to the medical data administrator, who sets the user role. The user definition unit obtains the setting data of the user role setting window and sets the user role according to the setting data.
[0033] Furthermore, the user authority setting unit sets the user basic authority according to the user role, wherein:
[0034] When the user role is a medical data administrator, the basic permission of the user is set as the first permission;
[0035] When the user role is a medical data editor, the basic permission of the user is set as the second permission;
[0036] When the user role is a medical data reader, the basic permission of the user is set to the third permission;
[0037] When the user role is a malicious user, the basic permissions of the user are set to no permissions.
[0038] Furthermore, the role modification unit obtains the user's access time and geographic location, determines the validity of the user role according to the user's access time and geographic location, modifies the user role according to the determination result, and adjusts the user's basic permissions according to the modification result, wherein:
[0039] When the access time is working hours, the user role is determined to be valid, and the user role is not modified, and the user's basic permissions are not adjusted;
[0040] When the access time is a non-working day, the user role is determined to be invalid, the user role is modified to a malicious user, and the user's basic permissions are adjusted to no permissions;
[0041] When the geographical location is consistent with the preset geographical location, the user role is determined to be valid, and the user role is not modified, and the user's basic permissions are not adjusted;
[0042] When the geographic location is inconsistent with the preset geographic location, the user role is determined to be invalid, the user role is modified to a malicious user, and the user's basic permissions are adjusted to no permissions.
[0043] Furthermore, in the cycle optimization unit, the number of invalid user roles T within a preset period is obtained, and the number of invalid user roles T is compared with the preset invalid number T0, and the frequency of invalid user roles is judged according to the comparison result, and the process of correcting the user roles is optimized according to the judgment result, wherein:
[0044] When T≤T0, the invalidity of the user role is judged as low frequency, and the process of correcting the user role is not optimized;
[0045] When T≥T0, the user role invalidity is determined to be high frequency, and the process of correcting the user role is optimized. The user is marked as a high-frequency invalid user, and the user role is set to a malicious user. The user's basic permissions are set to no permissions, and a danger warning window is pushed to the administrator for maintenance and processing.
[0046] Furthermore, the sandbox module establishes a sandbox environment according to the user's basic permissions and imports the classified data set into the sandbox environment, wherein:
[0047] When the user basic authority is the first authority, a first sandbox environment is established, and a first classification data set corresponding to the first sandbox environment is imported;
[0048] When the user basic authority is the second authority, a second sandbox environment is established, and a second classification data set corresponding to the second sandbox environment is imported;
[0049] When the user's basic authority is the third authority, a third sandbox environment is established, and a third classification data set corresponding to the third sandbox environment is imported;
[0050] When the user's basic permissions are no permissions, the sandbox environment is not established;
[0051] The sandbox module uses Docker in the virtualization technology to create the sandbox environment when establishing the sandbox environment, and uses VMware in the virtualization technology to create a virtual machine in the sandbox environment, imports the classified data set corresponding to the sandbox environment into the virtual machine, and sets the isolation level of the virtual machine to complete isolation;
[0052] The risk assessment module obtains real-time log data, inputs the real-time log data into the risk assessment model to perform risk assessment, obtains an assessment result, judges the security of the sandbox module according to the assessment result, and outputs the judgment result, wherein:
[0053] When the assessment result is no risk, the sandbox module is judged to be safe, and the judgment result is output to the environment destruction module;
[0054] When the assessment result is risky, the sandbox module is determined to be unsafe, the sandbox module is closed, and a risk alert is pushed to the administrator for maintenance.
[0055] On the other hand, the present invention also provides a medical data sharing method based on a security mechanism, comprising:
[0056] Step S1, collecting medical data;
[0057] Step S2, constructing a medical desensitization model and desensitizing the medical data to obtain desensitized data;
[0058] Step S3, constructing a classification label model, and performing classification label processing on the desensitized data to obtain a classification data set;
[0059] Step S4, setting the user role according to the user attributes;
[0060] Step S5, setting basic user permissions according to user roles;
[0061] Step S6, judging the validity of the user role according to the user's access time and geographical location, modifying the user role according to the judgment result, and adjusting the user's basic permissions according to the modification result;
[0062] Step S7, obtaining the number of invalid user roles within a preset period, and comparing the number of invalid user roles with the preset number of invalid user roles, judging the invalid frequency of the user roles according to the comparison result, and optimizing the process of correcting the user roles according to the judgment result.
[0063] Step S8, establishing a sandbox environment according to the user's basic permissions, and importing the classification data set corresponding to the sandbox environment into the sandbox environment;
[0064] Step S9, analyzing the security of the sandbox module according to the risk assessment model;
[0065] Step S10, destroying the sandbox environment and reclaiming the resources in the sandbox environment.
[0066] Compared with the prior art, the beneficial effects of the present invention are that the system collects medical data through the collection module to provide a solid foundation for subsequent data processing and analysis; the system effectively protects patient privacy by desensitizing and classifying medical data through the governance module, while improving the efficiency and quality of data use; the system sets user roles and basic user permissions through the access restriction module, and corrects, adjusts and optimizes user roles and basic user permissions to achieve refined permission management and dynamic adjustment mechanism, ensure the legality and security of data access, and prevent data leakage and abuse; the system constructs an independent sandbox environment through the sandbox module to achieve data isolation and on-demand access, reduce the risk of data leakage, and improve the flexibility and efficiency of data processing; the system performs security assessment on the sandbox environment through the risk assessment module, timely discovers and solves potential safety hazards, and ensures the safety and controllability of medical data during the sharing process; the system destroys the sandbox environment that is no longer needed and recycles resources through the environment destruction module, thereby achieving effective utilization of resources and cleanliness of the sandbox environment and reducing system maintenance costs.
[0067] In particular, the data desensitization unit can more accurately identify and desensitize sensitive information in medical data through learning and optimization of the convolutional neural network model, reduce the risk of misjudgment and missed judgment, thereby effectively removing sensitive information in medical data, reducing the risk of data leakage and abuse, and enhancing the security of medical data.
[0068] In particular, the data labeling unit can more accurately assign classification labels to medical data through the training and optimization of the decision tree model, reduce the classification error rate, reduce the need for manual intervention, improve the efficiency and accuracy of data processing, and provide a structured basis for the analysis, sharing and utilization of medical data.
[0069] In particular, the user definition unit sets user roles by combining automatic and manual methods, which can more effectively identify and isolate potential malicious users, reduce unauthorized data access and operations, and thus improve the security of medical data. Clear user role division helps to achieve refined authority management and effectively prevent data leakage and abuse.
[0070] In particular, the user authority setting unit can ensure that only authorized users can access and operate medical data through role-based authority setting, thereby preventing data leakage and abuse, and simplifying the complexity of authority management based on the mapping relationship between roles and authorities. Administrators do not need to set authorities for each user individually, but can indirectly manage user authorities by managing roles.
[0071] In particular, the role modification unit can reduce the possibility of misjudgment, improve the accuracy of permission control, more effectively identify and prevent potential malicious access behaviors, and thus enhance the security of the system by combining access time and geographical location to judge the validity of the user role.
[0072] In particular, the cycle optimization unit can significantly reduce the risk of malicious access and improve the overall security of the system by automatically identifying and processing high-frequency invalid users. It can also enhance administrators' awareness of potential security threats by pushing danger warning windows to administrators, allowing them to take timely measures to deal with them. For low-frequency invalid users, the system does not intervene excessively to avoid affecting the experience of normal users due to misjudgment.
[0073] In particular, the sandbox module effectively prevents user operations from potentially affecting other parts of the system by creating a completely isolated sandbox environment and virtual machine, thereby enhancing the overall security of the system. It also supports the dynamic creation of different sandbox environments and virtual machines based on user permission levels, providing a flexible medical data sharing environment.
[0074] In particular, the environment destruction module can avoid waste of system resources, reduce the burden on the system, improve the overall performance and response speed of the system by promptly destroying the sandbox environment that is no longer needed, and reduce the risk of data leakage and malicious exploitation by promptly destroying the sandbox environment. BRIEF DESCRIPTION OF THE DRAWINGS
[0075] Figure 1 This is a schematic diagram of the structure of the medical data sharing system based on the security mechanism of this example;
[0076] Figure 2 This is a schematic diagram of the structure of the governance module of this embodiment;
[0077] Figure 3 This is a schematic diagram of the structure of the access restriction module in this embodiment;
[0078] Figure 4 Schematic diagram of the process of the medical data sharing method based on the security mechanism of this embodiment. DETAILED DESCRIPTION
[0079] In order to make the objects and advantages of the present invention more clearly understood, the present invention is further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.
[0080] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the protection scope of the present invention.
[0081] It should be noted that in the description of the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected" and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, or it can be the internal communication of two components. For those skilled in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.
[0082] See also Figure 1 As shown, it is a schematic diagram of the structure of the medical data sharing system based on the security mechanism of this example, and the system includes:
[0083] A collection module, used to collect medical data;
[0084] The governance module is used to build a medical desensitization model and a classification label model, and perform desensitization and classification label processing on medical data to obtain a classified data set. The governance module is connected to the acquisition module;
[0085] The access restriction module is used to set user roles and basic user permissions, and is also used to modify, adjust and optimize user roles and basic user permissions. The access restriction module is connected to the governance module;
[0086] A sandbox module is used to establish a sandbox environment according to the user's basic permissions and import the classified data set corresponding to the sandbox environment into the sandbox environment. The sandbox module is connected to the access restriction module;
[0087] A risk assessment module, used to analyze the security of the sandbox module according to the risk assessment model, and the risk assessment module is connected to the sandbox module;
[0088] The environment destruction module is used to destroy the sandbox environment and recycle the resources in the sandbox environment. The environment destruction module is connected to the risk assessment module.
[0089] Specifically, the system is applied to a medical data sharing platform. The system collects and governs medical data and sets user roles and basic user permissions to build a sandbox environment to achieve secure sharing of medical data. The system collects medical data through a collection module to provide a solid foundation for subsequent data processing and analysis. The system desensitizes and classifies and labels medical data through a governance module to effectively protect patient privacy while improving data utilization efficiency and quality. The system sets user roles and basic user permissions through an access restriction module, and modifies user roles and basic user permissions. Adjustment and optimization are carried out to achieve refined permission management and dynamic adjustment mechanism, ensure the legality and security of data access, and prevent data leakage and abuse. The system builds an independent sandbox environment through the sandbox module to achieve data isolation and on-demand access, reduce the risk of data leakage, and improve the flexibility and efficiency of data processing. The system conducts security assessment on the sandbox environment through the risk assessment module, timely discovers and resolves potential security risks, and ensures the security and controllability of medical data during the sharing process. The system destroys the sandbox environment that is no longer needed and recycles resources through the environment destruction module, thereby achieving effective utilization of resources and cleanliness of the sandbox environment and reducing system maintenance costs.
[0090] Specifically, this embodiment does not limit the method of collecting medical data. Relevant technical personnel in this field can freely set it according to actual needs, and only need to meet the needs of real-time collection of medical data. For example, it can be set to collect medical data in real time through the application program interface of the medical device. The medical device refers to instruments, equipment, appliances and materials used on the human body, such as X-ray diagnostic equipment, ultrasonic diagnostic equipment, surgical equipment and radiotherapy equipment. The application program interface refers to a set of definitions, programs and protocols, through which different computer software systems can communicate with each other and share data and functions. The medical data refers to information related to health conditions collected, analyzed and applied in the medical field, such as patients' personal information, medical records, diagnostic results and treatment plans.
[0091] Specifically, the acquisition module collects medical data in real time, reduces data errors caused by human factors, and improves the accuracy of medical data.
[0092] See also Figure 2 As shown, it is a schematic diagram of the structure of the governance module of this embodiment, and the governance module includes:
[0093] The data desensitization unit is used to build a medical desensitization model and perform desensitization processing on medical data to obtain desensitized data;
[0094] The data labeling unit is used to build a classification labeling model and perform classification labeling on the desensitized data to obtain a classified data set. The data labeling unit is connected to the data desensitizing unit.
[0095] Specifically, when constructing a medical desensitization model, the data desensitization unit divides the preset desensitization database into a 75% desensitization training set and a 25% desensitization test set, inputs the desensitization training set into the convolutional neural network model for training, obtains the trained convolutional neural network model, and inputs the desensitization test set into the convolutional neural network model, tests the convolutional network neural model, obtains the desensitization accuracy rate Q, compares the desensitization accuracy rate Q with the preset desensitization accuracy rate Q0, judges the training compliance of the convolutional neural network model according to the comparison result, and outputs the convolutional neural network model according to the judgment result, wherein:
[0096] When Q≥Q0, the convolutional neural network model is judged to have met the training standards, and the convolutional neural network model is output as a medical desensitization model;
[0097] When Q < Q0, it is determined that the convolutional neural network model training does not meet the standards, and the convolutional neural network model parameters are iteratively optimized until the convolutional neural network model training meets the standards. The cross entropy loss function is set to the multi-classification cross entropy loss function. set up y is the actual measurement result in the preset desensitized database, is the predicted value output by the convolutional neural network model, yi is the true label of the i-th sample, is the probability that the model predicts that the sample is a positive class, N is the total number of samples, C is the total number of categories, yi, c are binary indicators, c is the probability that the model predicts that the i-th sample belongs to category c;
[0098] The data desensitization unit also inputs the medical data into the medical desensitization model for desensitization processing to obtain desensitized data.
[0099] Specifically, the preset desensitized database refers to a database composed of preset un-desensitized medical data and preset desensitized medical data. The preset undesensitized medical data refers to medical data that has not undergone any form of desensitization processing and directly contains sensitive information of patients. The preset desensitized medical data refers to medical data that has been desensitized by preset desensitization rules and methods. The preset desensitization rules and methods refer to data processing rules and operating procedures formulated for medical data, such as replacing real names with randomly generated pseudonyms and aliases, retaining the first and last few digits of the ID number, and replacing the middle part with "*" and "X". The desensitized training set refers to a data set used to train a convolutional neural network model. The desensitized test set refers to a data set used to evaluate the performance of a convolutional neural network model. The convolutional neural network model refers to a deep feedforward neural network used to process data with a grid structure. The desensitization accuracy rate Q refers to the proportion of the number of samples W correctly output by the convolutional neural network model to the total number of samples W0 in the desensitization test set, that is, Q=W / W0. The preset desensitization accuracy rate Q0 refers to the preset desensitization accuracy rate Q0 for The value for judging whether the training of the convolutional neural network model meets the standard is used. This embodiment does not limit the value of the preset desensitization accuracy rate Q0. Relevant technical personnel in this field can freely set it according to actual needs, and only need to meet the needs of judging whether the training of the convolutional neural network model meets the standard. For example, the preset desensitization accuracy rate Q0 can be set to 0.97. The cross entropy loss function refers to a measurement method for measuring the difference between two probability distributions. The multi-classification cross entropy loss function refers to a cross entropy loss function for measuring the prediction accuracy of the model on multiple categories. The true label of the i-th sample refers to the true category to which the i-th sample belongs. The probability that the model predicts that the sample is a positive class refers to the probability value of the sample belonging to the positive class calculated by the model according to the input data. The total number of samples refers to the total number of samples in the preset desensitization database. The total number of categories refers to the number of categories appearing in the classification problem. The binary indicator refers to a binary vector and value used to represent the sample category. The probability that the model predicts that the i-th sample belongs to category c refers to the probability value of the sample belonging to category c calculated by the model according to the input data.
[0100] Specifically, the data desensitization unit can more accurately identify and desensitize sensitive information in medical data through learning and optimization of the convolutional neural network model, reduce the risk of misjudgment and missed judgment, thereby effectively removing sensitive information in medical data, reducing the risk of data leakage and abuse, and enhancing the security of medical data.
[0101] Specifically, when constructing the classification label model, the data label unit trains the decision tree model according to the historical classification database, divides the historical classification database into 80% classification training set and 20% classification test set, inputs the classification training set into the decision tree model, and inputs the classification test set into the trained decision tree model, obtains the classification accuracy E, compares the classification accuracy E with the preset classification accuracy E0, judges the compliance of the decision tree model training according to the comparison result, and outputs the decision tree model according to the judgment result, wherein:
[0102] When E≥E0, the decision tree model training is judged to have met the standard, and the decision tree model is output as a classification label model;
[0103] When E<E0, it is determined that the decision tree model training does not meet the standard, and the decision tree model is repeatedly trained and the parameters are optimized until the decision tree model training meets the standard;
[0104] The data labeling module also inputs the desensitized data into the classification labeling model for classification labeling processing to obtain a classification data set.
[0105] Specifically, the historical classification database refers to a database for storing historical desensitized data that has not been processed by classification labels and its corresponding historical desensitized data that has been processed by classification labels. The decision tree model refers to a classification and regression method based on a tree structure. The classification training set refers to a data set used to train a decision tree model. The classification test set refers to a data set used to evaluate the performance of a decision tree model. The classification accuracy E refers to the ratio of the number of samples R correctly classified by the model to the total number of samples R0 in the classification test set. Set E=R / R0. The preset classification accuracy E0 refers to the preset accuracy used to reflect whether the decision tree model training has met the standards. This embodiment does not limit the value of the preset classification accuracy. Relevant technical personnel in this field can freely set it according to actual needs. It only needs to meet the need to judge whether the decision tree model training has met the standards. For example, the preset classification accuracy E0=0.99 can be set. This embodiment does not limit the content of the classification data set. Relevant technical personnel in this field can freely set it according to actual needs. It only needs to meet the need to classify and label the desensitized data. For example, the content of the classification data set can be set to research data only, data for use by medical institutions only, publicly available data, and data for query by patients themselves only.
[0106] Specifically, through the training and optimization of the decision tree model, the data labeling unit can more accurately assign classification labels to medical data, reduce the classification error rate, reduce the need for manual intervention, improve the efficiency and accuracy of data processing, and provide a structured basis for the analysis, sharing and utilization of medical data.
[0107] See also Figure 3As shown, it is a schematic diagram of the structure of the access restriction module of this embodiment, and the access restriction module includes:
[0108] User definition unit, used to set user roles according to user attributes;
[0109] A user authority setting unit is used to set basic user authority according to user roles, and the user authority setting unit is connected to the user definition unit;
[0110] A role modification unit is used to judge the validity of the user role according to the user's access time and geographical location, modify the user role according to the judgment result, and adjust the user's basic authority according to the modification result. The role modification unit is connected to the user authority setting unit;
[0111] The cycle optimization unit is used to obtain the number of invalid user roles within a preset period, and compare the number of invalid user roles with the preset invalid number, judge the invalid frequency of the user role according to the comparison result, and optimize the process of correcting the user role according to the judgment result. The cycle optimization unit is connected to the role correction unit.
[0112] Specifically, the user definition unit obtains user attributes and sets the user role according to the user attributes, wherein:
[0113] When the user attribute is a target administrator, the user role is set to a medical data administrator;
[0114] When the user attribute is a target physician, the user role is set as a medical data editor;
[0115] When the user attribute is a target patient, the user role is set to a medical data reader;
[0116] When the user attribute is untargeted, the user role is set as a malicious user, and the user attribute and user role setting window are pushed to the medical data administrator, who sets the user role. The user definition unit obtains the setting data of the user role setting window and sets the user role according to the setting data.
[0117] Specifically, the user attributes refer to labels used to describe user characteristics and identities, the user roles refer to a collection of users with similar responsibilities, permissions and behaviors, the target administrator refers to the administrator role responsible for managing and maintaining the medical information system, the target physician refers to the user with the physician role in the medical information system, the target patient refers to the service object of the medical information system, that is, the patient who receives medical services, the untargeted refers to the user who is not within the service scope of the medical information system, the medical data administrator refers to the professional responsible for data management of the medical information system, the medical data editor refers to the user with the permission to modify and update data in the medical information system, the medical data reader refers to the user with the permission to view data in the medical information system, and the malicious user refers to the user who intentionally violates the system security policy, abuses the permission and conducts illegal activities for improper purposes.
[0118] Specifically, the user definition unit sets user roles by combining automatic and manual methods, which can more effectively identify and isolate potential malicious users, reduce unauthorized data access and operations, and thus improve the security of medical data. Clear user role division helps to achieve refined authority management and effectively prevent data leakage and abuse.
[0119] Specifically, the user authority setting unit sets the user basic authority according to the user role, wherein:
[0120] When the user role is a medical data administrator, the basic permission of the user is set as the first permission;
[0121] When the user role is a medical data editor, the basic permission of the user is set as the second permission;
[0122] When the user role is a medical data reader, the basic permission of the user is set to the third permission;
[0123] When the user role is a malicious user, the basic permissions of the user are set to no permissions.
[0124] Specifically, the basic user permissions refer to the set of permissions that the user has and is necessary to perform operations. The first permissions include managing users, setting roles, managing the system, maintaining the system, and log management. Managing users includes creating, editing, and deleting users. The second permissions include data entry and editing, data query and retrieval, data review and verification. The third permissions include data access and data reading.
[0125] Specifically, the user authority setting unit can ensure that only authorized users can access and operate medical data through role-based authority setting, thereby preventing data leakage and abuse, and simplifying the complexity of authority management according to the mapping relationship between roles and authorities. Administrators do not need to set authorities for each user separately, but can indirectly manage user authorities by managing roles.
[0126] Specifically, the role modification unit obtains the user's access time and geographical location, determines the validity of the user role according to the user's access time and geographical location, modifies the user role according to the determination result, and adjusts the user's basic permissions according to the modification result, wherein:
[0127] When the access time is working hours, the user role is determined to be valid, and the user role is not modified, and the user's basic permissions are not adjusted;
[0128] When the access time is a non-working day, the user role is determined to be invalid, the user role is modified to a malicious user, and the user's basic permissions are adjusted to no permissions;
[0129] When the geographical location is consistent with the preset geographical location, the user role is determined to be valid, and the user role is not modified, and the user's basic permissions are not adjusted;
[0130] When the geographic location is inconsistent with the preset geographic location, the user role is determined to be invalid, the user role is modified to a malicious user, and the user's basic permissions are adjusted to no permissions.
[0131] Specifically, the access time refers to the specific time point when the user accesses the medical data sharing system based on the security mechanism, the working time refers to the time period when the medical data sharing system based on the security mechanism performs normal work and business activities, and the non-working time refers to the time period when the medical data sharing system based on the security mechanism does not perform normal work and business activities. The geographical location refers to the specific location of the user when accessing the medical data sharing system based on the security mechanism, and the preset geographical location refers to a pre-set geographical location. This embodiment does not limit the specific content of the preset geographical location. Relevant technical personnel in this field can freely set it according to actual needs, and only need to meet the need to judge the validity of the user role. For example, the preset geographical location can be set to the geographical location of the medical institution and the home address filled in by the user.
[0132] Specifically, the role modification unit determines the validity of the user role by combining the access time and geographical location, which can reduce the possibility of misjudgment, improve the accuracy of permission control, and more effectively identify and prevent potential malicious access behaviors, thereby enhancing the security of the system.
[0133] Specifically, in the cycle optimization unit, the number of invalid user roles T in a preset cycle is obtained, and the number of invalid user roles T is compared with the preset invalid number T0, and the frequency of invalid user roles is judged according to the comparison result, and the process of correcting the user roles is optimized according to the judgment result, wherein:
[0134] When T≤T0, the invalidity of the user role is judged as low frequency, and the process of correcting the user role is not optimized;
[0135] When T≥T0, the user role invalidity is determined to be high frequency, and the process of correcting the user role is optimized. The user is marked as a high-frequency invalid user, and the user role is set to a malicious user. The user's basic permissions are set to no permissions, and a danger warning window is pushed to the administrator for maintenance and processing.
[0136] Specifically, the cycle optimization unit can significantly reduce the risk of malicious access and improve the overall security of the system by automatically identifying and processing high-frequency invalid users. It can also enhance administrators' awareness of potential security threats by pushing danger warning windows to administrators, allowing them to take timely measures to deal with them. For low-frequency invalid users, the system does not intervene excessively to avoid affecting the experience of normal users due to misjudgment.
[0137] Specifically, the sandbox module establishes a sandbox environment according to the user's basic permissions and imports the classified data set into the sandbox environment, wherein:
[0138] When the user basic authority is the first authority, a first sandbox environment is established, and a first classification data set corresponding to the first sandbox environment is imported;
[0139] When the user basic authority is the second authority, a second sandbox environment is established, and a second classification data set corresponding to the second sandbox environment is imported;
[0140] When the user's basic authority is the third authority, a third sandbox environment is established, and a third classification data set corresponding to the third sandbox environment is imported;
[0141] When the user's basic permissions are no permissions, the sandbox environment is not established;
[0142] The sandbox module uses Docker in virtualization technology to create a sandbox environment when establishing the sandbox environment, and uses VMware in virtualization technology to create a virtual machine in the sandbox environment, imports the classified data set corresponding to the sandbox environment into the virtual machine, and sets the virtual machine isolation level to complete isolation.
[0143] Specifically, the first classified data set corresponding to the first sandbox environment refers to the first classified data set prepared by the sandbox module for the first sandbox environment when the user has the first permission, the second classified data set corresponding to the second sandbox environment refers to the second classified data set prepared by the sandbox module for the second sandbox environment when the user has the second permission, the third classified data set corresponding to the third sandbox environment refers to the third classified data set prepared by the sandbox module for the third sandbox environment when the user has the third permission, the virtualization technology refers to the technology of simulating multiple independent logical computing environments on a single physical computer, the Docker refers to a containerization platform that allows applications and their dependencies to be packaged into a lightweight, portable container and run on a machine that supports Docker, the VMware refers to the technology that allows users to create and manage virtual machines on physical computers, the virtual machine refers to an independent virtual environment simulated on a physical computer using virtualization technology, the virtual machine isolation level refers to the degree of demarcation between different computing environments of the virtual machine, including complete isolation, partial isolation and no isolation, and the complete isolation means that each virtual machine runs in an independent sandbox without any interaction with each other.
[0144] Specifically, the sandbox module effectively prevents user operations from potentially affecting other parts of the system by creating a completely isolated sandbox environment and virtual machine, thereby enhancing the overall security of the system. It also supports the dynamic creation of different sandbox environments and virtual machines based on user permission levels, providing a flexible medical data sharing environment.
[0145] Specifically, the risk assessment module obtains real-time log data, inputs the real-time log data into the risk assessment model to perform risk assessment, obtains an assessment result, judges the security of the sandbox module according to the assessment result, and outputs according to the judgment result, wherein:
[0146] When the assessment result is no risk, the sandbox module is judged to be safe, and the judgment result is output to the environment destruction module;
[0147] When the assessment result is risky, the sandbox module is determined to be unsafe, the sandbox module is closed, and a risk alert is pushed to the administrator for maintenance.
[0148] Specifically, the risk assessment model refers to an LSTM neural network model that is trained with preset log data and meets a preset accuracy rate. This embodiment does not limit the training method of the LSTM neural network model. Relevant technical personnel in this field can freely set it according to actual conditions, as long as the demand for risk assessment of real-time log data is met. For example, the training method of the LSTM neural network model can be set to divide the preset log data into 70% risk training set, 20% risk verification set and 10% risk test set, and the risk training set is input into the LSTM neural network model for training, and the parameters of the trained LSTM neural network model are iteratively optimized according to the risk verification set until the risk test set is input into the LSTM neural network model for risk assessment. During risk assessment, if the LSTM neural network model meets the preset risk assessment accuracy, the LSTM neural network model is output as a risk assessment model, the preset log data refers to the log data obtained in advance for training the risk assessment model, the risk training set refers to the data set used to train the LSTM neural network model, the risk verification set refers to the data set used to evaluate and adjust the parameters of the LSTM neural network model during the training process, the risk test set refers to the data set used to evaluate the final performance of the LSTM neural network model, the real-time log data refers to the log information generated and recorded in real time during the operation process, and the risk alarm refers to the process of issuing warnings to administrators and relevant personnel when risks and abnormal conditions are detected during the operation process.
[0149] Specifically, the risk assessment module ensures that the real-time status information of the system can be obtained and analyzed in a timely manner through the collection and processing of real-time log data, and uses the risk assessment model to identify potential security risks. At the same time, it automatically executes corresponding security measures based on the assessment results to improve the security and response speed of the system.
[0150] Specifically, the environment destruction module destroys the sandbox environment and recycles the resources in the sandbox environment when the user interaction ends and the determination result input by the risk assessment module is received.
[0151] Specifically, this embodiment does not limit the method of destroying the sandbox environment. Relevant technical personnel in the field can freely set it according to actual needs, and only need to meet the need to destroy the sandbox environment. For example, the method of destroying the sandbox environment can be set to destroy the sandbox environment through a virtualization management tool. The resources in the sandbox environment include computing resources, storage resources, and network resources.
[0152] Specifically, the environment destruction module can avoid waste of system resources, reduce the burden on the system, improve the overall performance and response speed of the system by promptly destroying the sandbox environment that is no longer needed, and reduce the risk of data leakage and malicious exploitation by promptly destroying the sandbox environment.
[0153] See also Figure 4 As shown, it is a flow chart of the medical data sharing method based on the security mechanism of this embodiment, and the method includes:
[0154] Step S1, collecting medical data;
[0155] Step S2, constructing a medical desensitization model and desensitizing the medical data to obtain desensitized data;
[0156] Step S3, constructing a classification label model, and performing classification label processing on the desensitized data to obtain a classification data set;
[0157] Step S4, setting the user role according to the user attributes;
[0158] Step S5, setting basic user permissions according to user roles;
[0159] Step S6, judging the validity of the user role according to the user's access time and geographical location, modifying the user role according to the judgment result, and adjusting the user's basic permissions according to the modification result;
[0160] Step S7, obtaining the number of invalid user roles within a preset period, and comparing the number of invalid user roles with the preset number of invalid user roles, judging the invalid frequency of the user roles according to the comparison result, and optimizing the process of correcting the user roles according to the judgment result.
[0161] Step S8, establishing a sandbox environment according to the user's basic permissions, and importing the classification data set corresponding to the sandbox environment into the sandbox environment;
[0162] Step S9, analyzing the security of the sandbox module according to the risk assessment model;
[0163] Step S10, destroying the sandbox environment and reclaiming the resources in the sandbox environment.
[0164] So far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present invention.
Claims
1. A medical data sharing system based on security mechanism, characterized in that: include: A collection module, used to collect medical data; The governance module is used to build a medical desensitization model and a classification label model, and perform desensitization and classification label processing on medical data to obtain a classified data set; Access restriction module, used to set user roles and basic user permissions, and also used to modify, adjust and optimize user roles and basic user permissions; The sandbox module is used to establish a sandbox environment according to the user's basic permissions and import the classification data set corresponding to the sandbox environment into the sandbox environment; A risk assessment module, used to analyze the security of the sandbox module based on the risk assessment model; The environment destruction module is used to destroy the sandbox environment and recycle the resources in the sandbox environment.
2. The medical data sharing system based on security mechanism according to claim 1 is characterized in that: The governance module includes: The data desensitization unit is used to build a medical desensitization model and perform desensitization processing on medical data to obtain desensitized data; The data labeling unit is used to build a classification labeling model and perform classification labeling on the desensitized data to obtain a classified data set.
3. The medical data sharing system based on security mechanism according to claim 2 is characterized in that: The data desensitization unit, when constructing a medical desensitization model, divides a preset desensitization database into a 75% desensitization training set and a 25% desensitization test set, inputs the desensitization training set into the convolutional neural network model for training, obtains a trained convolutional neural network model, and inputs the desensitization test set into the convolutional neural network model, tests the convolutional network neural model, obtains a desensitization accuracy rate Q, compares the desensitization accuracy rate Q with a preset desensitization accuracy rate Q0, judges the training compliance of the convolutional neural network model according to the comparison result, and outputs the convolutional neural network model according to the judgment result, wherein: When Q≥Q0, the convolutional neural network model is judged to have met the training standards, and the convolutional neural network model is output as a medical desensitization model; When Q < Q0, it is determined that the convolutional neural network model training does not meet the standards, and the convolutional neural network model parameters are iteratively optimized until the convolutional neural network model training meets the standards. The cross entropy loss function is set to the multi-classification cross entropy loss function. set up y is the actual measurement result in the preset desensitized database, is the predicted value output by the convolutional neural network model, yi is the true label of the i-th sample, is the probability that the model predicts that the sample is a positive class, N is the total number of samples, C is the total number of categories, yi, c are binary indicators, c is the probability that the model predicts that the i-th sample belongs to category c; The data desensitization unit also inputs the medical data into the medical desensitization model for desensitization processing to obtain desensitized data; The data labeling unit, when constructing the classification labeling model, trains the decision tree model according to the historical classification database, divides the historical classification database into 80% classification training set and 20% classification test set, inputs the classification training set into the decision tree model, and inputs the classification test set into the trained decision tree model, obtains the classification accuracy E, compares the classification accuracy E with the preset classification accuracy E0, judges the compliance of the decision tree model training according to the comparison result, and outputs the decision tree model according to the judgment result, wherein: When E≥E0, the decision tree model training is judged to have met the standard, and the decision tree model is output as a classification label model; When E<E0, it is determined that the decision tree model training does not meet the standard, and the decision tree model is repeatedly trained and the parameters are optimized until the decision tree model training meets the standard; The data labeling module also inputs the desensitized data into the classification labeling model for classification labeling processing to obtain a classification data set.
4. The medical data sharing system based on security mechanism according to claim 1 is characterized in that: The access restriction module includes: User definition unit, used to set user roles according to user attributes; User authority setting unit, used to set basic user authority according to user roles; A role modification unit is used to judge the validity of the user role according to the user's access time and geographical location, modify the user role according to the judgment result, and adjust the user's basic permissions according to the modification result; The cycle optimization unit is used to obtain the number of invalid user roles within a preset cycle, and compare the number of invalid user roles with the preset invalid number, judge the invalid frequency of the user role according to the comparison result, and optimize the process of correcting the user role according to the judgment result.
5. The medical data sharing system based on security mechanism according to claim 4 is characterized in that: The user definition unit obtains user attributes and sets the user role according to the user attributes, wherein: When the user attribute is a target administrator, the user role is set to a medical data administrator; When the user attribute is a target physician, the user role is set as a medical data editor; When the user attribute is a target patient, the user role is set to a medical data reader; When the user attribute is untargeted, the user role is set as a malicious user, and the user attribute and user role setting window are pushed to the medical data administrator, who sets the user role. The user definition unit obtains the setting data of the user role setting window and sets the user role according to the setting data.
6. The medical data sharing system based on security mechanism according to claim 4 is characterized in that: The user authority setting unit sets the user basic authority according to the user role, wherein: When the user role is a medical data administrator, the basic permission of the user is set as the first permission; When the user role is a medical data editor, the basic permission of the user is set as the second permission; When the user role is a medical data reader, the basic permission of the user is set to the third permission; When the user role is a malicious user, the basic permissions of the user are set to no permissions.
7. The medical data sharing system based on security mechanism according to claim 4 is characterized in that: The role modification unit obtains the user's access time and geographical location, determines the validity of the user role according to the user's access time and geographical location, modifies the user role according to the determination result, and adjusts the user's basic permissions according to the modification result, wherein: When the access time is working hours, the user role is determined to be valid, and the user role is not modified, and the user's basic permissions are not adjusted; When the access time is a non-working day, the user role is determined to be invalid, the user role is modified to a malicious user, and the user's basic permissions are adjusted to no permissions; When the geographical location is consistent with the preset geographical location, the user role is determined to be valid, and the user role is not modified, and the user's basic permissions are not adjusted; When the geographic location is inconsistent with the preset geographic location, the user role is determined to be invalid, the user role is modified to a malicious user, and the user's basic permissions are adjusted to no permissions.
8. The medical data sharing system based on security mechanism according to claim 4 is characterized in that: In the cycle optimization unit, the number of invalid user roles T in a preset period is obtained, and the number of invalid user roles T is compared with the preset invalid number T0, and the frequency of invalid user roles is judged according to the comparison result, and the process of correcting the user roles is optimized according to the judgment result, wherein: When T≤T0, the invalidity of the user role is judged as low frequency, and the process of correcting the user role is not optimized; When T≥T0, the user role invalidity is determined to be high frequency, and the process of correcting the user role is optimized. The user is marked as a high-frequency invalid user, and the user role is set to a malicious user. The user's basic permissions are set to no permissions, and a danger warning window is pushed to the administrator for maintenance and processing.
9. The medical data sharing system based on security mechanism according to claim 1 is characterized in that: The sandbox module establishes a sandbox environment according to the user's basic permissions and imports the classification data set into the sandbox environment, wherein: When the user basic authority is the first authority, a first sandbox environment is established, and a first classification data set corresponding to the first sandbox environment is imported; When the user basic authority is the second authority, a second sandbox environment is established, and a second classification data set corresponding to the second sandbox environment is imported; When the user's basic authority is the third authority, a third sandbox environment is established, and a third classification data set corresponding to the third sandbox environment is imported; When the user's basic permissions are no permissions, the sandbox environment is not established; The sandbox module uses Docker in the virtualization technology to create the sandbox environment when establishing the sandbox environment, and uses VMware in the virtualization technology to create a virtual machine in the sandbox environment, imports the classified data set corresponding to the sandbox environment into the virtual machine, and sets the isolation level of the virtual machine to complete isolation; The risk assessment module obtains real-time log data, inputs the real-time log data into the risk assessment model to perform risk assessment, obtains an assessment result, judges the security of the sandbox module according to the assessment result, and outputs the judgment result, wherein: When the assessment result is no risk, the sandbox module is judged to be safe, and the judgment result is output to the environment destruction module; When the assessment result is risky, the sandbox module is determined to be unsafe, the sandbox module is closed, and a risk alert is pushed to the administrator for maintenance.
10. A method applied to the medical data sharing system based on security mechanism as claimed in any one of claims 1 to 9, characterized in that: include: Step S1, collecting medical data; Step S2, constructing a medical desensitization model and desensitizing the medical data to obtain desensitized data; Step S3, constructing a classification label model, and performing classification label processing on the desensitized data to obtain a classification data set; Step S4, setting the user role according to the user attributes; Step S5, setting basic user permissions according to user roles; Step S6, judging the validity of the user role according to the user's access time and geographical location, modifying the user role according to the judgment result, and adjusting the user's basic permissions according to the modification result; Step S7, obtaining the number of invalid user roles within a preset period, and comparing the number of invalid user roles with the preset number of invalid user roles, judging the invalid frequency of the user roles according to the comparison result, and optimizing the process of correcting the user roles according to the judgment result. Step S8, establishing a sandbox environment according to the user's basic permissions, and importing the classification data set corresponding to the sandbox environment into the sandbox environment; Step S9, analyzing the security of the sandbox module according to the risk assessment model; Step S10, destroying the sandbox environment and reclaiming the resources in the sandbox environment.
Citation Information
Patent Citations
A secure and credible exchange and sharing method and system for medical data
CN109670340A
Information security collection and interaction system based on cloud server
CN111818083A
Medical data sharing method and system based on block chain and data desensitization
CN116910799A
Intelligent encryption USB flash disk based on fingerprint technology
CN117828573A
Cited By
Data security management method and device, electronic equipment and storage medium
CN120387185A
Hospital information integration platform management method and system based on big data
CN121144325A
Software access control method and device based on data security and medium
CN121786803A
