Intelligent grading and dynamic desensitization method, system and terminal for sensitive data

Abstract

The application relates to an intelligent grading and dynamic desensitization method, system and terminal of sensitive data, and belongs to the technical field of data processing. The intelligent grading and dynamic desensitization method comprises the following steps: receiving an original data field from a device; marking a basic sensitive label according to the field characteristics of the original data field, and outputting a metadata label table; receiving request information for accessing device data, and extracting request characteristics; calculating a combined sensitivity according to the metadata label table and the request characteristics; calculating an environment threat index according to the request characteristics, and determining a scene desensitization level; calculating a final desensitization level according to the combined sensitivity, the environment threat index and the scene desensitization level; and desensitizing the original data value and the device ID of the device according to the final desensitization level. The application has the beneficial effect of improving the data security protection level.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of data processing, in particular to a sensitive data intelligent grading and dynamic desensitization method, system and terminal. BACKGROUND

[0002] With the acceleration of digital transformation, governments, finance, medical and other industries have accumulated massive sensitive data (such as personal identity information, transaction records, device ID), and the risk of data leakage has increased significantly.

[0003] The current mainstream dynamic desensitization technology takes "rule engine + access control" as the core, and the typical implementation is as follows: role-based static rule desensitization: through predefining user roles (such as administrator, ordinary user) and desensitization rules (such as the display of the last four digits of a mobile phone number), fixed desensitization logic (such as MyBatis interceptor, database view desensitization) is triggered when data is queried.

[0004] Since the desensitization rules are statically bound to the roles, they cannot be dynamically adjusted according to the real-time access environment (such as abnormal IP), so there is a risk of being exploited by unauthorized access. SUMMARY

[0005] In order to improve the level of data security protection, the application provides a sensitive data intelligent grading and dynamic desensitization method, system and terminal.

[0006] In a first aspect, the application provides a sensitive data intelligent grading and dynamic desensitization method, which adopts the following technical scheme:

[0007] A sensitive data intelligent grading and dynamic desensitization method, comprising:

[0008] receiving an original data field from a device;

[0009] annotating a basic sensitive label according to the field characteristics of the original data field, and outputting a metadata label table;

[0010] receiving request information for accessing device data, and extracting request features;

[0011] calculating a combined sensitivity according to the metadata label table and the request features;

[0012] calculating an environment threat index according to the request features, and determining a scene desensitization level;

[0013] calculating a final desensitization level according to the combined sensitivity, the environment threat index and the scene desensitization level;

[0014] desensitizing the original data value and the device ID of the device according to the final desensitization level.

[0015] By adopting the technical scheme, the precision and dynamic of sensitive data protection are realized through the multi-dimensional fusion intelligent decision mechanism. Firstly, the basic sensitive label is marked through the original data field characteristic, and the combined sensitivity is calculated, which breaks through the limitation of traditional single field classification and can identify the risk caused by data correlation. Secondly, the environment threat index is calculated in real time combined with the request characteristics, and the scene desensitization level is matched, so that the desensitization strategy is upgraded from static "one size fits all" to dynamic adaptive mode. Finally, through the multi-dimensional decision model of combining the combined sensitivity, the environment threat index and the scene level, the system can intelligently output the optimal desensitization scheme, and the original data value and the device ID are linked to effectively prevent attackers from tracking data through device identification association. The method not only improves the data security protection level through fine-grained classification and dynamic adaptation, but also reduces the manual intervention cost through automatic label marking and rule engine.

[0016] Optionally, the step of marking the basic sensitive label according to the field characteristics of the original data field comprises:

[0017] Obtaining the field attribute of the original data field, the field attribute comprising an independent field and an identification field;

[0018] According to the field attribute, the single field sensitivity of the original data field is marked;

[0019] According to the field attribute, it is judged whether the original data field can be associated with other sensitive data;

[0020] If yes, the original data field is marked with an associated sensitive identification 1;

[0021] If not, the original data field is marked with an associated sensitive identification 0.

[0022] Optionally, the step of calculating the combined sensitivity according to the metadata label table and the request characteristics comprises:

[0023] The request characteristics comprise a query field;

[0024] According to the metadata label table, the highest single field sensitivity in the query field is queried;

[0025] The number of associated sensitive fields of the query field is queried;

[0026] According to the highest single field sensitivity and the number of associated sensitive fields, the combined sensitivity is calculated.

[0027] By adopting the technical scheme, the most sensitive field can be taken as a basic threshold for combined sensitive evaluation, ensuring that the combined evaluation result will not be underestimated due to the existence of a single high-sensitive field, and providing a key anchor point for overall sensitivity calculation. The number of associated sensitive fields is introduced as a calculation parameter to effectively identify the information leakage risk formed by the combination of multiple fields. The metadata label table provides the inherent sensitive attribute basis of the field, and the query field in the request feature reflects the specific scene demand of current data access. The combination of the two makes the combined sensitivity calculation rely on both static data attributes and dynamic access behavior, avoiding the problem of disconnection between traditional static classification and actual scene.

[0028] Optionally, according to the request feature, the step of calculating an environment threat index and determining a scene desensitization level comprises:

[0029] The request feature further comprises an HTTP request header and a device real-time state;

[0030] According to the HTTP request header, it is judged whether it is an intranet request;

[0031] If yes, the scene desensitization level is determined to be 0;

[0032] If no, the scene desensitization level is determined to be 1;

[0033] According to the device real-time state, a device state risk is determined;

[0034] According to the scene desensitization level and the device state risk, an environment threat index is calculated.

[0035] By adopting the technical scheme, the request feature dimension is refined, the scene desensitization level determination logic is optimized, and a multi-factor fusion environment threat index calculation model is constructed, which significantly improves the accuracy and dynamic adaptability of the sensitive data desensitization strategy. Specifically, the HTTP request header and the device real-time state are included in the request feature analysis category, combined with intranet / internet environment judgment and device risk assessment, the multi-dimensional risk of the data access scene is realized, which provides a more comprehensive decision basis for the dynamic adjustment of the final desensitization level, effectively avoiding the imbalance between security and usability of the traditional static desensitization rule in complex access scene.

[0036] Optionally, according to the combined sensitivity, the environment threat index and the scene desensitization level, the step of calculating the final desensitization level comprises:

[0037] It is judged whether the environment threat index is not greater than a first threshold value and the combined sensitivity is not greater than a second threshold value;

[0038] If yes, the final desensitization level is determined to be 0;

[0039] If not, the final desensitization level is calculated according to the set desensitization model; the desensitization model is , is the final desensitization level, is the scene desensitization level, is the combined sensitivity.

[0040] Optionally, the step of desensitizing the original data value of the device and the device ID before the final desensitization level includes:

[0041] Access the device ID access log and count the access times;

[0042] Determine whether the access times are less than the access threshold value, and the access threshold value is , is the threshold coefficient, is the historical average access times;

[0043] If not, the final desensitization level is updated according to the fuse formula; the fuse formula is , is the updated desensitization level, is the fuse enhancement coefficient.

[0044] By adopting the above technical solution, by accessing the device ID access log and counting the access times, the system can monitor the access behavior pattern of a single device in real time, and the access frequency is used as a supplementary dimension of desensitization decision. When the access times exceed the threshold value, the fuse formula is triggered to update the final desensitization level, realizing the upgrade from "static rule matching" to "dynamic behavior risk control", and effectively dealing with the potential data leakage risk brought by high-frequency access. The setting of the access times threshold value in combination with the fuse mechanism enables the desensitization strategy to be dynamically adjusted according to the actual interaction intensity of the device. The introduction of the fuse formula adds an "adaptive defense layer" to data security. When the access behavior of the device ID deviates from the normal baseline, the protection strength can be automatically upgraded without manual intervention, and the attack benefit is reduced by dynamically increasing the desensitization level.

[0045] Optionally, the step of desensitizing the original data value of the device and the device ID according to the final desensitization level includes:

[0046] According to the final desensitization level, the precision attenuation step is obtained;

[0047] According to the precision attenuation step, the original data value of the device is desensitized to generate a desensitized data value;

[0048] According to the final desensitization level, the prefix length of the device ID is calculated;

[0049] According to the prefix length, the device ID is masked to generate a desensitized device ID.

[0050] By adopting the technical scheme, the system can perform hierarchical precision control on the numerical original data by "obtaining the precision attenuation step length according to the final desensitization level". The dynamic adjustment mechanism meets the requirements of different scenes on data availability, and realizes gradient protection of sensitive information through precision attenuation. The scheme of "dynamic calculation of prefix length + mask processing" for the device ID can effectively avoid the risk of association leakage caused by the fixed desensitization rule. The processing manner maintains the format characteristics of the device ID to support normal interaction of the system, and cuts off the association path between the device identifier and the sensitive data by dynamically adjusting the mask length, thereby preventing the attacker from tracking the data across scenes through the device fingerprint.

[0051] In a second aspect, the application provides an intelligent hierarchical and dynamic desensitization system for sensitive data, which adopts the following technical scheme:

[0052] An intelligent hierarchical and dynamic desensitization system for sensitive data comprises:

[0053] An information receiving module is configured to receive original data fields from a device and request information for accessing device data;

[0054] An information processing module is configured to label a basic sensitive label according to the field characteristics of the original data fields, and output a metadata label table, to calculate a combined sensitivity according to the metadata label table and the request characteristics, and to calculate a final desensitization level according to the combined sensitivity, the environment threat index and the scene desensitization level;

[0055] A desensitization module is configured to desensitize the original data values and the device ID of the device according to the final desensitization level.

[0056] In a third aspect, the application provides a terminal, which adopts the following technical scheme:

[0057] A terminal comprises:

[0058] A memory stores an intelligent hierarchical and dynamic desensitization program for sensitive data;

[0059] A processor is configured to execute the program stored on the memory to implement the steps of the above-mentioned intelligent hierarchical and dynamic desensitization method for sensitive data.

[0060] In summary, the application has at least the following beneficial effects:

[0061] Through a multi-dimensional integrated intelligent decision-making mechanism, the protection of sensitive data is made more precise and dynamic. First, by labeling basic sensitive tags with the characteristics of raw data fields and calculating combined sensitivity, the limitations of traditional single-field classification are overcome, enabling the identification of risks arising from data correlation. Second, by combining request characteristics to calculate the environmental threat index in real time and matching the scenario-based de-identification level, the de-identification strategy is upgraded from a static "one-size-fits-all" approach to a dynamic adaptation mode. Finally, by integrating combined sensitivity, environmental threat index, and scenario-level multi-dimensional decision-making models, the system can intelligently output the optimal de-identification solution, while simultaneously linking the original data values ​​with the device ID, effectively preventing attackers from tracking data through device identifiers. This method not only improves data security protection through fine-grained classification and dynamic adaptation but also reduces the cost of manual intervention through automated tagging and a rule engine. Attached Figure Description

[0062] Figure 1 This is a first flowchart of an embodiment of the method of this application;

[0063] Figure 2 This is a second flowchart of an embodiment of the method of this application;

[0064] Figure 3 This is a third flowchart of an embodiment of the method of this application;

[0065] Figure 4 This is the fourth flowchart of an embodiment of the method of this application;

[0066] Figure 5 This is the fifth flowchart of an embodiment of the method of this application;

[0067] Figure 6 This is a schematic diagram of one implementation method of the desensitization scoring model. Detailed Implementation

[0068] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the appendices in the embodiments of the present invention will be described below. Figure 1 - Appendix Figure 6 The technical solutions in the embodiments of the present invention are clearly and completely described herein. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0069] The first embodiment of this application discloses a method for intelligent classification and dynamic desensitization of sensitive data. (Refer to...) Figure 1 The intelligent grading and dynamic desensitization method may include S110-S170:

[0070] S110, receiving a raw data field from a device;

[0071] S120, labeling a basic sensitive tag according to the field characteristics of the raw data field, and outputting a metadata tag table;

[0072] S130, receiving request information for accessing device data, and extracting request features;

[0073] S140, calculating a combined sensitivity according to the metadata tag table and the request features;

[0074] S150, calculating an environmental threat index according to the request features, and determining a scene desensitization level;

[0075] S160, calculating a final desensitization level according to the combined sensitivity, the environmental threat index, and the scene desensitization level;

[0076] S170, desensitizing the raw data value of the device and the device ID according to the final desensitization level.

[0077] Specifically, the data field of the device end is collected through an API interface or a device SDK; the raw data field can include a device ID, a sensor type, a data field name, etc.

[0078] Then the field attributes of the raw data field are obtained, the field attributes include independent fields and identification fields, and the raw data field is marked with a single field sensitivity according to the field attributes. Then it is judged whether the raw data field can be associated with other sensitive data according to the field attributes; if yes, the raw data field is marked with an associated sensitive identification 1; if not, the raw data field is marked with an associated sensitive identification 0.

[0079] For example, the single field sensitivity : the independent field (such as temperature, frequency) is set to 0.2, and the identification field (such as device ID, location) is set to 0.8.

[0080] Associated sensitive identification : if the field can be associated with other sensitive data (such as device ID can be associated with location), the associated sensitive identification of the device ID is set to 1.

[0081] The output metadata tag table includes the field name and .

[0082] The received API request or user behavior instruction is parsed to extract request features, including query fields, device real-time state, HTTP request headers, etc.

[0083] Referring to Figure 2, S140, calculating the combined sensitivity according to the metadata label table and the request characteristics, the step S210-S230:

[0084] S210, querying the highest single-field sensitivity in the query field according to the metadata label table;

[0085] S220, querying the number of associated sensitive fields of the query field;

[0086] S230, calculating the combined sensitivity according to the highest single-field sensitivity and the number of associated sensitive fields.

[0087] Specifically, the sensitivity label of the target field can be filtered through a structured query (such as SQL), and the highest value can be obtained through the MAX() function. Then, by using the table association metadata in the metadata label table, by querying the "associated table", "associated field" and "association type" information of the table to which the query field belongs, all association paths are traversed, and the total number of fields marked as "1" on the association chain is counted.

[0088] Combined sensitivity is the highest single-field sensitivity; is the number of associated sensitive fields, and a is the association influence coefficient, which can be calibrated through business scenarios.

[0089] S150, calculating the environment threat index according to the request characteristics, and determining the scene desensitization level, the step can specifically include:

[0090] According to the HTTP request header, it is judged whether it is an intranet request; if yes, the scene desensitization level is determined to be 0; if not, the scene desensitization level is determined to be 1; then, according to the device real-time state, the device state risk is determined; and then, according to the scene desensitization level and the device state risk, the environment threat index is calculated.

[0091] Specifically, the Host field in the HTTP request header is extracted, and it is judged whether it conforms to the intranet IP address specification (such as the 10.0.0.0 / 8, 172.16.0.0 / 12, 192.168.0.0 / 16 network segment in IPv4, or the domain name suffix is.local,.lan, etc. Intranet identifier). For example, if the Host value is 192.168.1.100:8080 or server.internal, it is determined to be an intranet request; if it is www.example.com or a public IP (such as 203.0.113.5), it is determined to be an extranet request.

[0092] If it is an intranet request, the scene desensitization level , indicating plaintext; if it is an extranet request, the scene desensitization level , indicating basic desensitization.

[0093] For S340, for example, the device is at shutdown, then the corresponding device state risk is 0; the device is at standby, then the corresponding device state risk is 0.5; the device is in operation, then the corresponding device state risk is 1.

[0094] environment threat index , is the scene desensitization level, is the device state risk; , is the weight of the corresponding parameter, is greater than , and , and

[0095] Referring to Figure 3 , S160, according to the combined sensitivity, the environment threat index and the scene desensitization level, the step of calculating the final desensitization level includes S310-S330:

[0096] S310, judging whether the environment threat index is not greater than a first threshold value and the combined sensitivity is not greater than a second threshold value;

[0097] S320, if yes, determining that the final desensitization level is 0;

[0098] S330, if no, calculating the final desensitization level according to a set desensitization model.

[0099] Specifically, the desensitization model: final desensitization level .

[0100] For example, if and , then (clear text), otherwise ; if , then the desensitization is enhanced.

[0101] In addition, if , then the clear text is represented, but the audit log needs to be recorded. If , then the light desensitization is enabled: for example, only the high-risk field (such as the device ID) is masked, and the numerical field retains the original precision.

[0102] Referring to Figure 4 , S170, before the step of desensitizing the original data value and the device ID of the device according to the final desensitization level includes S410-S430:

[0103] S410, calling the device ID access log and counting the access times;

[0104] S420, judging whether the access times are less than an access threshold value;

[0105] S430, if not, update the final desensitization level according to the fuse formula.

[0106] Specifically, from the Nginx / Apache access log or database audit log of the application server, filter the records containing the device ID, and extract the key fields including device ID, access timestamp, request path, etc. The statistical period can be set to the last 24 hours, and then the access frequency can be counted by device ID using the pandas library of Python, or high-efficiency query can be realized through SQL statements.

[0107] Access threshold = , is the threshold coefficient, which can be defined according to the business scenario, is the historical average access frequency; the fuse formula is , is the final desensitization level after updating, is the fuse enhancement coefficient.

[0108] Referring to Figure 5 , S180, according to the final desensitization level, the steps of desensitizing the original data value of the device and the device ID include S510-S540:

[0109] S510, according to the final desensitization level, obtain the precision attenuation step;

[0110] S520, according to the precision attenuation step, desensitize the original data value of the device to generate a desensitized data value;

[0111] S530, according to the final desensitization level, calculate the prefix length of the device ID;

[0112] S540, according to the prefix length, mask the device ID to generate a desensitized device ID.

[0113] Specifically, obtain the original data value (such as vibration frequency) and , then according to the formula: ; , is the desensitized data value, is the precision attenuation coefficient, which can be defined according to the actual situation; to reduce the precision on the premise of preserving the trend of the original data.

[0114] Obtain the device ID (such as "WT-01") and , and then segment mask processing:

[0115] Formula: ; the prefix length ;

[0116] Example: Prefix Length Then the de-sensitization device ID is "WT-**".

[0117] In addition, after receiving the original request, the final de-sensitization level, the de-sensitized data value, and the de-sensitization device ID, the key audit fields, the requester IP and the scene ID, the data comparison value before and after de-sensitization, the fuse trigger flag are recorded, and the standardized audit log is output according to the audit format requirements or template.

[0118] In addition, the de-sensitization score and the number of false positives can be periodically (such as monthly, semi-monthly) updated according to the audit log and business feedback; then the parameters involved in de-sensitization are modified and updated according to the de-sensitization score and the number of false positives.

[0119] Specifically, through the collector module of the comprehensive log audit platform, the audit logs of network devices, servers, application systems and de-sensitization systems themselves are collected, covering user operation records, de-sensitization rule execution logs, data access behaviors, etc. Using the log analysis function of the platform, unstructured logs (such as free text) are converted into a unified format (including fields: operation time, user ID, device IP, de-sensitization rule ID, original data type, de-sensitization result, whether to trigger alarm, etc.). Configure the log collection task to be triggered automatically on the 1st and 16th of each month to generate periodic log data packages. Build an online feedback platform (such as an internal system form) to support business personnel to submit de-sensitization false positives (such as "mobile phone de-sensitization shows not full, resulting in inability to contact customers") and de-sensitization deficiencies (such as "ID card number only hides the middle 6 digits, still has a risk of leakage") and other issues. Feedback information includes feedback ID, business scenario (such as customer management system), original data sample, de-sensitization result, false positive type (missing de-sensitization / over de-sensitization), occurrence time, processing priority, etc.

[0120] An embodiment of the de-sensitization score model can refer to the following figure. Figure 6

[0121] Comprehensive score formula: de-sensitization score = Σ (dimension score x weight), result rounded to 1 decimal place (such as 4.2 points).

[0122] False positive definition:

[0123] Missing de-sensitization false positive: the de-sensitization field is not de-sensitized (such as the full display of the ID card number).

[0124] Over de-sensitization false positive: non-sensitive field is de-sensitized by mistake (such as the username only displays the first letter).

[0125] Statistical method:

[0126] ​Filter the records marked as "warning" from the audit log, correlate the false positive cases in the business feedback, and count the total number by period after deduplication.

[0127] When the desensitization score in the period is less than 3.5 or the number of false positives is greater than or equal to 10, the system automatically generates a parameter correction task.

[0128] When correcting parameters, the formula can be: n is the parameter before correction, is the parameter after correction, and z is a custom optimization coefficient value.

[0129] The implementation principle of the embodiment is:

[0130] Receive the original data field from the device, then obtain the field attribute of the original data field, and then mark the single field sensitivity of the original data field according to the field attribute, and determine whether the original data field can be associated with sensitive data, if yes, mark the original data field with an associated sensitive identifier 1, if not, mark the original data field with an associated sensitive identifier 0; after completing the basic sensitive tag of the field characteristics of the original data field, output the metadata tag table;

[0131] After receiving the request information for accessing the device data, extract the request characteristics, and then query the highest single field sensitivity and the number of associated sensitive fields in the query field in the request characteristics according to the metadata tag table; then calculate the combined sensitivity according to the highest single field sensitivity and the number of associated sensitive fields;

[0132] According to the HTTP request header, determine whether it is an intranet request, if yes, determine that the scene desensitization level is 0; if not, determine that the scene desensitization level is 1; then determine the device state risk according to the real-time state of the device; then calculate the environment threat index according to the scene desensitization level and the device state risk;

[0133] Determine whether the environment threat index is not greater than a first threshold value, and whether the combined sensitivity is not greater than a second threshold value; if yes, determine that the final desensitization level is 0, if not, calculate the final desensitization level according to the set desensitization model;

[0134] Call the device ID access log and count the access times, and determine whether the access times are less than the access threshold; if not, update the final desensitization level according to the fuse formula;

[0135] According to the final desensitization level, obtain the precision decay step, then desensitize the original data value of the device according to the precision decay step to generate a desensitized data value, then calculate the prefix length of the device ID according to the final desensitization level, and then mask the device ID according to the prefix length to generate a desensitized device ID.

[0136] Based on the above method embodiments, the second embodiment of the present application discloses a sensitive data intelligent grading and dynamic desensitization system. The sensitive data intelligent grading and dynamic desensitization system of the embodiment of the present application can implement any one of the above sensitive data intelligent grading and dynamic desensitization methods, and the specific working process of each module in the sensitive data intelligent grading and dynamic desensitization system can refer to the corresponding process in the above method embodiments.

[0137] For ease of understanding, the following is an example: a sensitive data intelligent grading and dynamic desensitization system, comprising:

[0138] An information receiving module is configured to receive raw data fields from a device and request information for accessing device data;

[0139] An information processing module is configured to label a basic sensitive tag according to a field characteristic of the raw data field, and output a metadata tag table, to calculate a combined sensitivity according to the metadata tag table and a request characteristic, and to calculate a final desensitization level according to the combined sensitivity, an environmental threat index, and a scene desensitization level;

[0140] A desensitization module is configured to desensitize raw data values and a device ID of the device according to the final desensitization level.

[0141] The third embodiment of the present application provides a terminal, which can include a memory and a processor as an implementation manner of the terminal, wherein

[0142] The memory is configured to store a sensitive data intelligent grading and dynamic desensitization program;

[0143] The processor is configured to execute the program stored on the memory to implement the steps of the above sensitive data intelligent grading and dynamic desensitization method.

[0144] The memory can be in communication connection with the processor through a communication bus, and the communication bus can be an address bus, a data bus, a control bus, etc.

[0145] In addition, the memory can include a random access memory (RAM) and can also include a non-volatile memory (NVM), such as at least one disk memory.

[0146] The processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; and can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic device, a discrete gate or transistor logic device, a discrete hardware component, etc.

[0147] The above are only preferred embodiments of the present application and are not used to limit the protection scope of the present application. Any feature disclosed in the specification (including the abstract and drawings) can be replaced by other equivalent or similar features unless otherwise stated. That is, each feature is only an example of a series of equivalent or similar features unless otherwise stated.

Claims

1. A method for intelligent grading and dynamic desensitization of sensitive data, characterized in that, The method comprises the following steps: receiving a raw data field from a device; annotating a basic sensitive label according to the field characteristics of the raw data field, and outputting a metadata label table; receiving request information for accessing device data, and extracting request features; calculating a combined sensitivity according to the metadata label table and the request features; calculating an environment threat index according to the request features, and determining a scene desensitization level; calculating a final desensitization level according to the combined sensitivity, the environment threat index, and the scene desensitization level; desensitizing the raw data value and the device ID of the device according to the final desensitization level; The step of calculating a combined sensitivity according to the metadata label table and the request features comprises: The request features include a query field; querying the highest single-field sensitivity in the query field according to the metadata label table; querying the number of associated sensitive fields of the query field; calculating a combined sensitivity according to the highest single-field sensitivity and the number of associated sensitive fields; The step of calculating an environment threat index according to the request features, and determining a scene desensitization level comprises: The request features further include HTTP request headers and device real-time status; determining whether it is an intranet request according to the HTTP request headers; if yes, determining that the scene desensitization level is 0; if no, determining that the scene desensitization level is 1; determining the device state risk according to the device real-time status; calculating an environment threat index according to the scene desensitization level and the device state risk; The step of calculating a final desensitization level according to the combined sensitivity, the environment threat index, and the scene desensitization level comprises: determining whether the environment threat index is not greater than a first threshold value, and whether the combined sensitivity is not greater than a second threshold value; if yes, determining that the final desensitization level is 0; If not, the final desensitization level is calculated according to the set desensitization model; the desensitization model is , is the final desensitization level, is the scene desensitization level, is the combined sensitivity.

2. The method of claim 1, wherein, The step of annotating a basic sensitive label according to the field characteristics of the raw data field comprises: obtaining field attributes of the raw data field, the field attributes including independent fields and identification fields; labeling a single-field sensitivity of the raw data field according to the field attributes; determining whether the raw data field can be associated with other sensitive data according to the field attributes; if yes, labeling the raw data field with an associated sensitive identification 1; if no, labeling the raw data field with an associated sensitive identification 0. 3.The method of claim 1, wherein, The step before desensitizing the raw data value and the device ID of the device according to the final desensitization level comprises: calling a device ID access log and counting the number of accesses; determining whether the number of accesses is less than an access threshold, the access threshold being , is a threshold coefficient, is a historical average number of accesses; If not, update the final desensitization level according to the fuse formula; the fuse formula is , is the updated desensitization level, is a fuse enhancement coefficient.

4. The method of claim 3, wherein the method further comprises: The step of desensitizing the raw data value and the device ID of the device according to the final desensitization level comprises: obtaining a precision attenuation step according to the final desensitization level; desensitizing the raw data value of the device according to the precision attenuation step to generate a desensitized data value; calculating the prefix length of the device ID according to the final desensitization level; masking the device ID according to the prefix length to generate a desensitized device ID.

5. An intelligent grading and dynamic desensitization system for sensitive data, characterized in that, The method for intelligently grading and dynamically desensitizing sensitive data according to any one of claims 1-4 comprises: an information receiving module configured to receive raw data fields and request information for accessing device data from a device; an information processing module configured to label a base sensitivity tag according to a field characteristic of the raw data fields, and output a metadata tag table, to calculate a combined sensitivity according to the metadata tag table and the request characteristic, and to calculate a final desensitization level according to the combined sensitivity, the environmental threat index, and the scenario desensitization level; a desensitization module configured to desensitize raw data values and device IDs of the device according to the final desensitization level.

6. A terminal, characterized by comprising: comprising: a memory storing an intelligent grading and dynamic desensitization program of sensitive data; a processor configured to execute the program stored on the memory to implement the steps of the intelligent grading and dynamic desensitization method of sensitive data according to any one of claims 1-4.

Images