Data desensitization control method and control system

Abstract

The invention discloses a data desensitization control method and system, and belongs to the technical field of information security, and the method comprises the steps: constructing a distributed probe matrix, collecting and recognizing sensitive information in data streams of each node, carrying out the preliminary screening and marking through a preset rule, and determining an environment risk score; establishing a dynamic risk assessment strategy and a dynamic desensitization rule, and adjusting the risk level according to rule chain type triggering; establishing a dual-channel dynamic desensitization execution mechanism; according to the environmental risk scores, decision matrixes of different data types are constructed, and different desensitization methods are distributed according to the decision matrixes; and a closed-loop feedback mechanism is established, and the desensitization effect is verified and self-optimized. In the implementation process of the technical scheme provided by the invention, the selection of the desensitization method is realized by establishing a dual-channel dynamic desensitization execution mechanism and establishing the desensitization method library, so that the encryption delay is reduced and the concurrency capability is improved while the data retention rate is improved.

Description

Technical Field

[0001] The present application relates to the field of information security technology, and specifically to a data desensitization control method and control system. Background Art

[0002] With the rapid development of information technology, data security has become the focus of attention for enterprises and individuals, and data desensitization technology is crucial in ensuring information security.

[0003] In existing data desensitization technologies, their application scope is relatively wide, such as individuals, enterprises, government agencies, etc., to ensure data security and prevent data leakage. Traditional data desensitization technologies include various types, such as using fixed rules (such as global field masking) to achieve data desensitization, which is usually applied to specific scenarios, such as financial statements, customer information, etc. This method is simple and convenient, does not require a large amount of computing power, and can quickly achieve the desensitization effect in some scenarios, and is the current mainstream data desensitization means.

[0004] However, fixed rules are difficult to cope with complex and changeable data environments, easily cause incomplete desensitization or over-desensitization, affect data availability, and at the same time lack real-time performance, unable to perceive dynamic factors such as user permissions, network environment, data flow, etc. in real time, and the desensitization decision delay is relatively high in high-concurrency scenarios, affecting the response speed. There is a lack of a desensitization mechanism that can perform dynamic perception, high-throughput processing, and intelligent decision-making, resulting in loopholes in data security protection.

[0005] Therefore, it is necessary to provide a data desensitization control method and control system to solve the above problems.

[0006] It should be noted that the above information disclosed in this background art section is only used to understand the background art of the concept of the present application, and therefore, it may include information that does not constitute the prior art. Summary of the Invention

[0007] Based on the above problems existing in the prior art, the problems to be solved by the present application are: to provide a data desensitization control method and control system, which can improve the data retention rate, reduce the encryption delay, and improve the concurrency ability by establishing a dual-channel dynamic desensitization execution mechanism.

[0008] The technical solution adopted by the present application to solve its technical problems is: A data desensitization control method, comprising: Construct a distributed probe matrix, collect and identify sensitive information in the data streams of each node, perform preliminary screening and marking through preset rules, and determine the environmental risk score; Establish a dynamic risk assessment strategy and dynamic desensitization rules, and adjust the risk level according to the rule chain trigger; Establish a dual-channel dynamic desensitization execution mechanism. The dual channels include a plaintext channel and a ciphertext channel, which can improve the data retention rate while reducing the encryption latency; According to the environmental risk score, construct a decision matrix for different data types, and allocate different desensitization methods according to this decision matrix. Among them, the desensitization methods are implemented by establishing a desensitization method library and are updated during each iteration; Establish a closed-loop feedback mechanism to verify and self-optimize the desensitization effect.

[0009] During the implementation of the technical solution of this application, by establishing a dual-channel dynamic desensitization execution mechanism and implementing the selection of desensitization methods by establishing a desensitization method library, the data retention rate is improved, the encryption latency is reduced, and the concurrency ability is enhanced.

[0010] Furthermore, the distributed probe matrix includes three levels: Level a, network layer capture. Deploy data probes at the API gateway to intercept HTTP request messages in real time and implement traffic mirroring collection based on the eBPF technology. The whole process is compatible with TCP / IP / HTTP / HTTPS protocols; Level b, database layer capture. Grab and parse SQL query statements through a JDBC proxy, extract the SELECT field list and the WHERE condition. The SELECT field list refers to the set of fields queried from the database, and the WHERE condition is the limiting condition for the query; Level c, terminal layer capture. Identify the access mode according to the integrated terminal information and judge the compliance of data access based on user behavior analysis.

[0011] Furthermore, in Level c, it also includes real-time analysis of user behavior logs to determine whether the user violates the regulations. Whether it violates the regulations is judged according to the access density. The access density refers to the access frequency of the user to sensitive data within a unit time. The higher the access density, the greater the possibility of violation.

[0012] Furthermore, after collecting and identifying sensitive information in the data streams of each node, it is also necessary to perform preliminary screening and marking through preset rules. The preset rules include keyword matching, data type identification, access permission verification, and environmental risk assessment.

[0013] Furthermore, the environmental risk assessment uses the multi-index weight analysis method to calculate the environmental risk score. The evaluation indicators include network encryption strength, terminal credibility, and geographical location risk. Among them, the detection method of network encryption strength uses the TLS protocol version + key length detection, the detection method of terminal credibility uses the device certificate + device status, and the evaluation of geographical location risk is based on the geographical location information parsed from the IP address and combines the historical security event database to comprehensively judge the risk level.

[0014] Furthermore, establishing a dynamic risk assessment strategy and dynamic desensitization rules includes: building a relational data graph to store field-level association relationships, automatically assigning risk levels to each field, marking fields with a high risk level, triggering the desensitization mechanism automatically when high-risk fields appear in the data stream, and automatically triggering chained desensitization rules when multiple high-risk fields intersect to enhance the data protection layer.

[0015] Furthermore, the plaintext channel includes SM4 national cryptographic algorithm block encryption and dynamic key management, and the ciphertext channel uses Paillier homomorphic encryption to achieve statistical analysis. The SM4 national cryptographic algorithm adopts the NIST standard, and the dynamic key management is based on the key rotation mechanism of HSM, and a fixed key life cycle is set.

[0016] Furthermore, before constructing a decision matrix for different data types, it is necessary to first assign a risk interval, which is determined based on the environmental risk score, and then match corresponding desensitization methods according to the data sensitivity and usage scenarios within the risk interval.

[0017] Furthermore, the closed-loop feedback mechanism includes performing reversibility tests and constructing adversarial samples. The process of the adversarial samples is to go from the original data to the desensitized data, then reverse-derive the attack, and calculate the information loss rate. Among them, the calculation formula of the information loss rate is: Information loss rate = 1 - (H(X|Y) / H(X)), where H(X) is the information entropy of the original data X, representing the uncertainty or amount of information of the data itself, and H(X|Y) is the conditional entropy of the original data X under the condition of knowing the desensitized data Y.

[0018] A data desensitization control system, which includes: An information collection module, used to build a distributed probe matrix, collect and identify sensitive information in the data streams of each node, perform preliminary screening and marking through preset rules, and determine the environmental risk score; A risk level adjustment module, used to establish a dynamic risk assessment strategy and dynamic desensitization rules, and adjust the risk level by chain-triggering according to the rules; A desensitization execution module, used to establish a dual-channel dynamic desensitization execution mechanism. The dual channels include a plaintext channel and a ciphertext channel, reducing the encryption delay while increasing the data retention rate; A desensitization method selection module, used to construct a decision matrix for different data types according to the environmental risk score, and assign different desensitization methods according to the decision matrix. Among them, the desensitization methods are implemented by establishing a desensitization method library and are updated during each iteration; A verification and optimization module, used to establish a closed-loop feedback mechanism to verify and self-optimize the desensitization effect.

[0019] The beneficial effects of the present application are as follows: A data desensitization control method and control system provided by the present application establish a dual-channel dynamic desensitization execution mechanism and select desensitization methods by establishing a desensitization method library, which improves the data retention rate, reduces encryption latency, and enhances concurrency capabilities.

[0020] In addition to the purposes, features, and advantages described above, the present application has other purposes, features, and advantages. The following will refer to the drawings for a further detailed description of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The accompanying drawings forming a part of this application are used to provide a further understanding of the present application. The schematic embodiments and descriptions thereof of the present application are used to explain the present application and do not constitute an improper limitation to the present application. In the drawings: Figure 1 is a schematic diagram of the overall process of a data desensitization control method in the present application; Figure 2 is a schematic diagram of the module composition of a data desensitization control system in the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0022] It should be noted that, without conflict, the embodiments in the present application and the features in the embodiments may be combined with each other. The following will refer to the drawings and combine with the embodiments to detail the present application.

[0023] In order to enable those skilled in the art to better understand the solution of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0024] Embodiment 1: As Figure 1 shown, the present application provides a data desensitization control method, which is used to protect sensitive data in different scenarios, improve data security, reduce the risk of privacy leakage, and reduce acquisition latency, and support multiple data sources. The method includes the following steps: Step A: Construct a distributed probe matrix, collect and identify sensitive information in the data streams of each node, perform preliminary screening and marking through preset rules, and determine the environmental risk score; During the process of data collection, by constructing a distributed probe matrix, it is possible to monitor and capture sensitive information in the data stream in real time, ensuring that data is not leaked during transmission. In the prior art, the probe matrix is an efficient data monitoring tool. Through distributed deployment, multi-point collaborative monitoring is achieved. Among them, the distributed probe matrix includes three levels: Level a, network layer capture. Data probes are deployed at the API gateway to intercept HTTP request messages in real time and implement traffic mirroring collection based on eBPF technology. The entire process is compatible with TCP / IP / HTTP / HTTPS protocols; Level b, database layer capture. SQL query statements are captured and parsed through a JDBC proxy to extract the SELECT field list and WHERE conditions. The SELECT field list refers to the set of fields queried from the database, and the WHERE condition is the limiting condition for the query; Level c, terminal layer capture. The access mode is identified based on the integrated terminal information, and the compliance of data access is judged based on user behavior analysis. Among them, the terminal information includes but is not limited to iOS / Android, SDK, IMEI of the collection device, GPS, biometric features, etc. Through the multi-level collaborative effect of the distributed probe matrix, comprehensive monitoring of sensitive data in all links of transmission, storage, and access is ensured, achieving millisecond-level data capture latency and improving the accuracy of field-level operation recognition; In Level c, it also includes real-time analysis of user behavior logs to judge whether the user violates the regulations. Whether it violates the regulations is judged based on the access density. The access density refers to the access frequency of the user to sensitive data within a unit time. The higher the access density, the greater the possibility of violation. The access records of users with high access density are merged and notified to the administrator. The administrator conducts further verification based on the records and takes corresponding measures after confirming the violation behavior; After collecting and identifying sensitive information in the data streams of each node, preliminary screening and marking are also required through preset rules. The preset rules include keyword matching, data type identification, access permission verification, and environmental risk assessment. Among them, keyword matching, data type identification, and access permission verification can refer to existing technologies. The environmental risk assessment calculates the environmental risk score using the multi-index weight analysis method. Specifically, the assessment indicators include network encryption strength, terminal credibility, and geographical location risk. Among them, the detection method for network encryption strength uses the TLS protocol version + key length detection, the detection method for terminal credibility uses device certificates + device status (such as whether it is jailbroken, rooted, etc.), and the assessment of geographical location risk is based on the geographical location information parsed from the IP address, combined with the historical security event database, to comprehensively judge the risk level to ensure the comprehensiveness and accuracy of data security protection. Then, after setting the weights respectively, the environmental risk score is output. For example, the weight value of network encryption strength is set to 0.3, the weight value of terminal credibility is set to 0.4, and the weight value of geographical location risk is set to 0.3. The final score is obtained through weighted calculation. It should be noted that the data range of the environmental risk score is from 0 to 1, and the higher the score, the greater the risk.

[0025] Step B: Establish a dynamic risk assessment strategy and dynamic data masking rules, and adjust the risk level according to the rule chain trigger. Since the data masking in existing technologies is usually based on fixed rules and cannot adapt to complex and ever-changing actual scenarios, resulting in an increased risk of exposure of sensitive data, and when the data volume surges, fixed rules are difficult to cope with. Therefore, a dynamic risk assessment strategy is introduced to dynamically adjust the data masking rules according to real-time data flow to ensure the effective protection of sensitive data at different risk levels and reduce the risk of data leakage. Specifically, establishing a dynamic risk assessment strategy and dynamic data masking rules includes: Establish a relational data graph to store field-level association relationships, automatically assign risk levels to each field, mark the fields with high risk levels, and when high-risk fields appear in the data stream, the system automatically triggers the data masking mechanism. When multiple high-risk fields cross, the chained data masking rules are automatically triggered to improve the data protection level. For example, in a certain data stream, cross-border payment, biometrics, and high-risk regions are all fields with high risk levels. When the three conditions of "cross-border payment" + "biometrics" + "high-risk region" are all met, it is automatically upgraded to the highest protection level, and the system will start multiple data masking measures, such as data encryption, access control, and behavior auditing, to ensure data security. At the same time, it can also prevent the reverse derivation of sensitive data information through non-sensitive fields, thereby constructing a multi-level and three-dimensional data protection system to achieve the dynamic balance between risk and protection.

[0026] Step C: Establish a dual-channel dynamic desensitization execution mechanism. The dual channels include a plaintext channel and a ciphertext channel, which can improve the data retention rate while reducing the encryption latency. In the traditional single-channel desensitization mechanism, the data encryption and decryption processes are cumbersome, resulting in an increase in processing latency and affecting system performance. The dual-channel mechanism optimizes the data transfer efficiency by processing plaintext and ciphertext data in parallel. The plaintext channel includes the SM4 national cryptography algorithm block encryption and dynamic key management, while the ciphertext channel uses Paillier homomorphic encryption to achieve statistical analysis. The SM4 national cryptography algorithm adopts the NIST standard, and the dynamic key management is based on the key rotation mechanism of HSM, and a fixed key lifecycle is set to ensure key security. Paillier homomorphic encryption supports data operations in the ciphertext state while protecting data privacy, reducing decryption operations, lowering latency, and improving the overall data processing efficiency. Through the dual-channel mechanism, the plaintext data is transferred under the protection of SM4 encryption, and the ciphertext data is efficiently operated under Paillier encryption. The two work together to ensure data security and processing speed, achieving a balanced optimization of risk prevention and control and system performance. After actual application, the actual effects are shown in Table 1 below. The data encryption latency using the dual-channel mechanism is reduced by approximately 46%, the concurrent capacity is increased by 14 times, and the data retention rate is increased to 293%. Step D: According to the environmental risk score, construct a decision matrix for different data types, and allocate different desensitization methods according to this decision matrix. Among them, the desensitization methods are implemented by establishing a desensitization method library and are updated during each iteration. Specifically, before constructing the decision matrix for different data types, it is necessary to first allocate a risk interval, which is determined based on the environmental risk score in Step A. Then, according to the data sensitivity and usage scenarios within the risk interval, corresponding desensitization methods are matched. The desensitization methods include, but are not limited to, AES encryption, differential privacy, noise addition, generalization processing, etc. In actual applications, one or more desensitization methods can be combined and used for each data type. Table 2 below is an example of one of the decision matrices: After constructing the decision matrix, automated desensitization processing can be carried out according to the data type and desensitization method, ensuring that the data meets the privacy protection requirements during the transfer process without affecting the business analysis needs, achieving a balance between data security and business efficiency. And by continuously iterating and optimizing the desensitization method library, the system can adapt to the changing data environment and security requirements, further improving the data governance level.

[0027] Step E: Establish a closed-loop feedback mechanism to verify and self-optimize the desensitization effect.

[0028] After completing the above steps, it is also necessary to perform closed-loop feedback on the desensitization effect to ensure the effectiveness of desensitization. Specifically, the closed-loop feedback mechanism includes performing a reversibility test and constructing adversarial samples. The process of the adversarial samples is to go from the original data to the desensitized data, then reverse-derive the attack, and calculate the information loss rate. The calculation formula for the information loss rate is: Information loss rate = 1 - (H(X|Y) / H(X)), where H(X) is the information entropy of the original data X, representing the uncertainty or amount of information of the data itself, and H(X|Y) is the conditional entropy of the original data X under the condition that the desensitized data Y is known; In this embodiment, it is assumed that there are N equally probable values in a certain field for the convenience of example, then H(X) = log2(N), H(X|Y) = log2(N / M), where M is the number of values after desensitization. Assuming the ID number desensitization scenario: Original data X: 18-digit ID number (each digit 0-9), Desensitized data Y: The first 6 digits are shown + the last 12 digits are masked, then H(X) = log2(10^18) = 59.76 bits, H(X|Y) = log2(10^12) = 39.84 bits, Information loss rate = 1 - (39.84 / 59.76) ≈ 0.333, which means that this desensitization method only eliminates 33% of the original information; Based on the above example, the desensitization effect can be verified and the desensitization strategy can be adjusted, such as increasing the number of masked digits or using a more complex encryption algorithm, to reduce the information loss rate and ensure the balance between privacy protection and data availability. Through multiple iterations of optimization, an efficient and secure desensitization scheme is finally achieved.

[0029] Embodiment 2: As Figure 2 shown, the present application also proposes a data desensitization control system, which runs the desensitization control method in Embodiment 1. The system includes: An information collection module, used to construct a distributed probe matrix, collect and identify sensitive information in the data streams of each node, perform preliminary screening and marking through preset rules, and determine the environmental risk score; A risk level adjustment module, used to establish a dynamic risk assessment strategy and dynamic desensitization rules, and adjust the risk level according to the rule chain trigger; A desensitization execution module, used to establish a dual-channel dynamic desensitization execution mechanism. The dual channels include a plaintext channel and a ciphertext channel, which can improve the data retention rate while reducing the encryption delay; A desensitization method selection module, used to construct a decision matrix for different data types according to the environmental risk score, and allocate different desensitization methods according to the decision matrix. Among them, the desensitization method is implemented by establishing a desensitization method library and is updated during each iteration; The verification and optimization module is used to establish a closed-loop feedback mechanism to verify and self-optimize the desensitization effect.

[0030] The above are only the preferred embodiments of the present application and are not intended to limit the present application. For those skilled in the art, various changes and modifications can be made to the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included within the protection scope of the present application.

Claims

1. A data desensitization control method, characterized in that: Including: Construct a distributed probe matrix to collect and identify sensitive information in the data streams of each node, perform preliminary screening and marking through preset rules, and determine the environmental risk score; Establish a dynamic risk assessment strategy and dynamic desensitization rules, and adjust the risk level according to the chained trigger of the rules; Establish a dual-channel dynamic desensitization execution mechanism. The dual channels include a plaintext channel and a ciphertext channel, which can improve the data retention rate while reducing the encryption delay; According to the environmental risk score, construct a decision matrix for different data types, and allocate different desensitization methods according to this decision matrix. Among them, the desensitization methods are implemented by establishing a desensitization method library and are updated during each iteration; Establish a closed-loop feedback mechanism to verify and self-optimize the desensitization effect.

2. The data desensitization control method according to claim 1, wherein: The distributed probe matrix includes three levels: Level a, network layer capture. Deploy data probes at the API gateway to intercept HTTP request messages in real time, and implement traffic mirroring collection based on eBPF technology. The whole process is compatible with TCP / IP / HTTP / HTTPS protocols; Level b, database layer capture. Grab and parse SQL query statements through a JDBC proxy, extract the SELECT field list and WHERE conditions. The SELECT field list refers to the set of fields queried from the database, and the WHERE condition is the limiting condition for the query; Level c, terminal layer capture. Identify the access mode according to the integrated terminal information, and judge the compliance of data access based on user behavior analysis.

3. The data desensitization control method according to claim 2, wherein: In Level c, it also includes real-time analysis of user behavior logs to judge whether the user violates the regulations. Whether the user violates the regulations is judged according to the access density. The access density refers to the access frequency of the user to sensitive data within a unit time. The higher the access density, the greater the possibility of violation.

4. A data desensitization control method according to claim 1, characterized in that: After collecting and identifying sensitive information in the data streams of each node, it is also necessary to perform preliminary screening and marking through preset rules. The preset rules include keyword matching, data type identification, access permission verification, and environmental risk assessment.

5. A data desensitization control method according to claim 4, characterized in that: The environmental risk assessment calculates the environmental risk score using the multi-index weight analysis method. The assessment indicators include network encryption strength, terminal credibility, and geographical location risk. Among them, the detection method of network encryption strength uses the TLS protocol version + key length detection, the detection method of terminal credibility uses the device certificate + device status, and the assessment of geographical location risk is based on the geographical location information parsed from the IP address, combined with the historical security event database, to comprehensively judge the risk level.

6. A data desensitization control method according to claim 1, characterized in that: Establishing a dynamic risk assessment strategy and dynamic desensitization rules includes: establishing a relational data graph, storing field-level association relationships, and automatically assigning risk levels to each field, marking fields with a high risk level. When high-risk fields appear in the data stream, the system automatically triggers the desensitization mechanism, and when multiple high-risk fields cross, the chained desensitization rules are automatically triggered to improve the data protection layer.

7. A data desensitization control method according to claim 1, characterized in that: The plaintext channel includes SM4 national cryptographic algorithm block encryption and dynamic key management. The ciphertext channel uses Paillier homomorphic encryption to implement statistical analysis. The SM4 national cryptographic algorithm adopts the NIST standard, and the dynamic key management is based on the key rotation mechanism of HSM, and a fixed key life cycle is set.

8. A data desensitization control method according to claim 1, characterized in that: Before constructing the decision matrix for different data types, it is necessary to first allocate the risk interval, which is determined based on the environmental risk score, and then match the corresponding desensitization method according to the data sensitivity and usage scenario within the risk interval.

9. A data desensitization control method according to claim 1, characterized in that: The closed-loop feedback mechanism includes performing a reversibility test and constructing adversarial samples. The process of the adversarial samples is to go from the original data to the desensitized data, then reverse-derive the attack, and calculate the information loss rate. Among them, the calculation formula of the information loss rate is: Information loss rate = 1 - (H(X|Y) / H(X)), where H(X) is the information entropy of the original data X, representing the uncertainty or information amount of the data itself, and H(X|Y) is the conditional entropy of the original data X under the condition that the desensitized data Y is known.

10. A data desensitization control system, characterized in that: For implementing the data desensitization control method according to any one of claims 1 to 9, the system includes: An information collection module, configured to construct a distributed probe matrix, collect and identify sensitive information in the data streams of each node, perform preliminary screening and marking through preset rules, and determine the environmental risk score; A risk level adjustment module, configured to establish a dynamic risk assessment strategy and dynamic desensitization rules, and adjust the risk level according to the rule chain trigger; A desensitization execution module, configured to establish a dual-channel dynamic desensitization execution mechanism. The dual channels include a plaintext channel and a ciphertext channel, which can improve the data retention rate while reducing the encryption delay; A desensitization method selection module, configured to construct a decision matrix for different data types according to the environmental risk score, and allocate different desensitization methods according to the decision matrix. Among them, the desensitization method is implemented by establishing a desensitization method library and is updated during each iteration; A verification and optimization module, configured to establish a closed-loop feedback mechanism to verify and self-optimize the desensitization effect.

Images