A method for a shared risk alert interface based on data anonymization status
By constructing a data source dependency graph and a de-identification evidence chain, the effectiveness changes of the de-identification status during data sharing are identified, solving the problem of missing risk fields during data sharing and achieving more accurate risk assessment and handling.
Patent Information
- Application Number
- CN202611480139.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Priority Date
- 2026-08-28
- Filing Date
- 2026-09-22
- Publication Date
- 2026-11-10
AI Technical Summary
Existing technologies struggle to accurately identify the validity of data anonymization during data sharing, especially when data structures change or anonymization rules are not updated synchronously, leading to omissions of risk fields or inconsistencies between risk warnings and the actual situation.
By constructing a data source dependency graph and a de-identified evidence chain, de-identified evidence is obtained and compared to determine its validity. Invalid evidence nodes are identified, and the de-identification failure status is propagated along the data source dependency graph. The sharing risk level is determined by comprehensively considering the validity of the de-identified evidence, the impact of failure, and the sensitivity level of the fields, and a sharing risk warning interface is generated.
It improves the accuracy of desensitization status judgment, identifies the correlation between upstream changes and downstream fields, reduces risk omissions, improves the pertinence and reliability of risk assessment, and generates a detailed risk warning interface for users to handle.
Smart Images

Figure CN123069441A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of data security, and in particular to a method for a shared risk warning interface based on data anonymization status. Background Technology
[0002] As data resources are increasingly shared and transferred in scenarios such as government affairs, finance, enterprise management, and internet services, in order to reduce the risk of leakage of sensitive information during the sharing process, it is usually necessary to de-identify data fields involving personal information, business-sensitive information, or other protected information before data sharing, and decide whether to allow sharing or to provide risk warnings to users based on the de-identification status of the fields.
[0003] Existing data sharing risk management methods typically check the data to be shared based on whether data fields have undergone de-identification, whether de-identification rules have been configured, or whether they have corresponding de-identification identifiers. The checks then display corresponding risk levels or warning messages based on the results. However, after de-identification, data may undergo multiple processing stages such as copying, mapping, splicing, truncation, aggregation, and encoding conversion. Simultaneously, de-identification rules, data structures, field semantics, and data versions may change as business operations progress. Therefore, the "de-identified" status in historical records may not always accurately reflect the actual effective de-identification status of the data to be shared.
[0004] For example, when the data structure of a source field changes, the original de-identification rules are not updated synchronously, or the de-identified data undergoes a new processing path to form derived fields, the original de-identification result may no longer meet the protection requirements of the current data. Furthermore, if the de-identification status of an upstream field fails, it may further affect multiple downstream fields through the data processing relationships between fields. Existing technologies typically focus on statically judging the current de-identification flag of a single field, making it difficult to trace the evidence that led to the de-identification status and its processing path, and also difficult to identify the propagation impact of de-identification status failure between different fields. This can easily result in the omission of risk fields or inconsistencies between risk warnings and the actual data status.
[0005] To address this, a method for a shared risk alert interface based on data anonymization status is proposed. Summary of the Invention
[0006] The main objective of this invention is to provide a method for a shared risk warning interface based on data anonymization status, which can effectively solve the problems in the background art.
[0007] To achieve the above objectives, the technical solution adopted by the present invention is as follows:
[0008] A method for a shared risk alert interface based on data anonymization status includes the following steps:
[0009] S1. Obtain the data to be shared, the data sharing request, the sensitivity level of each target field, and the data processing record corresponding to the data to be shared. For each target field in the data to be shared, obtain the current de-identification status from the data processing record and extract the de-identification evidence supporting the current de-identification status. Obtain the current status information used to verify the de-identification evidence. The de-identification evidence includes at least de-identification rule information, data version information, source field structure information, field semantic information, field processing path information, and de-identification effect verification information.
[0010] S2. Using source fields, intermediate processing fields, and target fields as field nodes, establish directed connection edges based on the data processing relationships between fields to construct a data source dependency graph; for each target field, trace back upstream along the data source dependency graph, and associate the field nodes, data processing operations, and de-identified evidence on the backtracking path according to the data processing order to form a corresponding de-identified evidence chain;
[0011] S3. Compare the de-identified evidence in the de-identified evidence chain with the current status information to determine the validity of the de-identified evidence for each field node, and determine the invalid evidence node based on the validity of the de-identified evidence; wherein, the current status information at least represents the current de-identification rule, the current data structure, the current field semantics, the current field processing path, and the current de-identification effect verification status;
[0012] S4. Taking the failure evidence node as the propagation starting point, determine the failure inheritance coefficient according to the data processing relationship corresponding to the directed connection edge, propagate the desensitization failure status downstream along the data source dependency graph, and obtain the desensitization failure impact degree of each target field.
[0013] S5. Determine the sharing scenario risk based on the data sharing request, and determine the sharing risk level of each target field based on the validity of the de-identified evidence, the impact of de-identification failure, the sensitivity level of the field, and the sharing scenario risk.
[0014] S6. Generate a shared risk warning interface based on the shared risk level, failure evidence node, and propagation path of the desensitized failure status; wherein, the shared risk warning interface is used to display at least the risk field, the cause of the desensitized failure status, the failure propagation path, the affected field, and the corresponding risk handling information.
[0015] Furthermore, the de-identification rule information includes at least one of de-identification rule identifier, de-identification rule version, and de-identification parameters; the source field structure information includes at least one of field data type, field length, encoding format, and structure pattern; and the de-identification effect verification information includes verification result, the data version corresponding to the verification, and verification time.
[0016] The validity of the desensitized evidence is determined according to the following formula:
[0017]
[0018] in, Indicates the first The validity of de-identified evidence for each field node; , , , and These represent the consistency of de-identification rules, data structure consistency, field semantic consistency, field processing path consistency, and the validity of the de-identification effect verification status, respectively. , , , and Let be the corresponding non-negative weights, and satisfy:
[0019]
[0020] in, , , , and The range of values is .
[0021] Furthermore, the data processing relationships include at least one of copying, mapping, transformation, splicing, truncation, aggregation, encoding, decoding, hashing, replacement, generalization, perturbation, and referencing;
[0022] The directed connection edges are established according to the data processing direction from upstream fields to downstream fields, and the de-identification execution nodes, de-identification effect verification nodes, and data processing nodes that change the field data structure or field semantics are marked in the de-identification evidence chain.
[0023] Furthermore, determining invalid evidence nodes based on the validity of the de-identified evidence includes:
[0024] The validity of the desensitized evidence is compared with a first valid threshold and a second valid threshold, wherein the first valid threshold is greater than the second valid threshold;
[0025] When the validity of the de-identified evidence is not lower than the first valid threshold, the de-identification status of the corresponding field node is determined to be valid;
[0026] When the validity of the de-identified evidence is lower than the first valid threshold and not lower than the second valid threshold, the de-identification status of the corresponding field node is determined to be pending review;
[0027] When the validity of the de-identified evidence is lower than the second validity threshold, the corresponding field node will be identified as an invalid evidence node.
[0028] Based on the type of evidence that reduces the effectiveness of the de-identified evidence, determine at least one of the following failure causes: rule failure, structural failure, semantic failure, processing path failure, or verification status failure.
[0029] Furthermore, the initial desensitization failure impact of the failed evidence node is determined as follows:
[0030]
[0031] in, Indicates the first The initial desensitization failure impact of each failed evidence node. Indicates the first The validity of de-identified evidence at each invalid evidence node;
[0032] For field nodes located downstream of the failure evidence node, the impact of the de-identification failure is determined iteratively according to the data processing order:
[0033]
[0034] in, Indicates the first The impact of desensitization failure on individual field nodes; Indicates the relationship with the first The set of upstream field nodes directly connected to each field node; Indicates the first The impact of desensitization failure on upstream field nodes; Indicates the first The field node up to the first The failure inheritance coefficient between each field node, and .
[0035] Furthermore, the failure inheritance coefficient is configured according to the data processing relationship corresponding to the directed connection edge;
[0036] When a downstream field node re-executes the desensitization process and passes the desensitization effect verification, the corresponding failure propagation path passing through the downstream field node is cut off; or, when the desensitization failure impact is lower than a preset propagation threshold, the propagation along the corresponding failure propagation path is stopped.
[0037] Furthermore, the sensitivity levels of each target field are converted into normalized field sensitivity, and the shared scenario risk is converted into a normalized shared scenario risk value. The shared risk value for each target field is determined according to the following formula:
[0038]
[0039] in, Indicates the first Shared risk values for each target field; Indicates the first The validity of anonymized evidence for each target field; Indicates the first The impact of desensitization failure on individual target fields; Indicates the first Normalized field sensitivity of each target field; Indicates the first Normalized shared scenario risk values corresponding to each target field; , , and The weights are non-negative and satisfy:
[0040]
[0041] in, and The range of values is ;
[0042] The shared risk level of the corresponding target field is determined based on the comparison result between the shared risk value and the preset risk threshold.
[0043] Furthermore, the shared risk warning interface includes a field risk display area, a failure propagation path display area, and a risk handling area;
[0044] The risk processing area is used to receive processing instructions for risk fields. The processing instructions correspond to at least one of the following processing operations: re-de-identification, re-verification, removal of risk fields, adjustment of sharing scope, submission for manual review, or termination of sharing.
[0045] In response to the processing instruction, a processing operation is performed on the corresponding target field, the processed de-identified evidence is re-extracted, the validity of the de-identified evidence, the impact of de-identification failure, and the sharing risk level of the target field are updated, and the sharing risk warning interface is updated according to the update results.
[0046] A shared risk alert interface system based on data anonymization status, comprising:
[0047] The data acquisition module is used to acquire the data to be shared, the data sharing request, the field sensitivity level of each target field, and the data processing record, and to acquire the current de-identification status, de-identification evidence, and current status information for verifying the de-identification evidence of each target field.
[0048] The dependency construction module is used to build a data source dependency graph based on the data processing relationships between fields, and to form a de-identified evidence chain for each target field;
[0049] The evidence verification module is used to compare the de-identified evidence in the de-identified evidence chain with the current status information to determine the validity of the de-identified evidence and the invalid evidence nodes of each field node.
[0050] The failure propagation module is used to determine the failure inheritance coefficient based on the data processing relationship, and propagate the desensitization failure status downstream along the data source dependency graph to obtain the desensitization failure impact degree of each target field.
[0051] The risk determination module is used to determine the sharing risk level of each target field based on the validity of the de-identified evidence, the impact of de-identification failure, the sensitivity level of the field, and the risk of the sharing scenario corresponding to the data sharing request.
[0052] The interface generation module is used to generate a shared risk warning interface based on the shared risk level, failure evidence nodes, and failure propagation path.
[0053] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described.
[0054] Compared with the prior art, the present invention has the following beneficial effects:
[0055] 1. This invention extracts de-identification evidence supporting the current de-identification status of target fields in the data to be shared, and combines the current de-identification rules, data structure, field semantics, field processing path, and de-identification effect verification status to determine the validity of the de-identification evidence. This makes the data sharing risk judgment no longer solely dependent on the static label of "de-identified" or "not de-identified", and can identify situations where the historical de-identification status has become invalid due to changes in data or processing environment, thereby improving the accuracy of de-identification status judgment.
[0056] 2. This invention constructs a data source dependency graph and a desensitization evidence chain to associate source fields, intermediate processing fields, target fields, and corresponding data processing operations. Starting from the failure evidence node, it propagates the desensitization failure status along the data source dependency relationship. This can identify the associated impact of upstream desensitization status changes on downstream derived fields, reducing the possibility of missing risks due to only checking the desensitization status of the target field itself.
[0057] 3. This invention comprehensively considers the validity of desensitized evidence, the impact of desensitization failure, the sensitivity level of the field, and the risk of the sharing scenario to determine the sharing risk level of the target field. This allows the risk assessment to simultaneously reflect the field's own desensitization status, the impact of upstream failure propagation, and the actual sharing scenario. Compared with a single desensitization status assessment method, this can improve the pertinence and reliability of the sharing risk assessment.
[0058] 4. This invention generates a shared risk warning interface based on the shared risk level, failure evidence nodes, and failure propagation path, enabling users to obtain information such as risk fields, reasons for failure in the desensitized state, failure propagation path, and affected fields, and to take corresponding risk handling measures accordingly, thereby improving the traceability and processing efficiency of data sharing risks. Attached Figure Description
[0059] Figure 1 This is a schematic diagram of the method flow of the present invention;
[0060] Figure 2 This is a schematic diagram of the system modules of the present invention. Detailed Implementation
[0061] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.
[0062] Example 1
[0063] like Figure 1-2 As shown, a method for a shared risk warning interface based on data anonymization status is applicable to enterprise data platforms, government data sharing platforms, data middleware, data exchange platforms, or other computer systems that need to perform sensitive data risk checks before data output or sharing.
[0064] In this embodiment, the data to be shared can be a database table, dataset, query results, API returned data, report data, or result data processed from multiple data sources. The target field refers to the data field in the data to be shared that requires de-identification and sharing risk assessment.
[0065] Specifically, this embodiment includes the following steps.
[0066] S1. Obtain the data to be shared and anonymized evidence.
[0067] Retrieve the data to be shared, the data sharing request, the sensitivity level of each target field, and the data processing record corresponding to the data to be shared.
[0068] The data sharing request may include one or more of the following: the identifier of the data to be shared, the recipient of the data, the purpose of sharing, the scope of sharing, the data output method, and the validity period of sharing.
[0069] The sensitivity level of a field can be determined based on pre-established data classification and grading rules. For example, data can be classified into ordinary data, low-sensitivity data, medium-sensitivity data, and high-sensitivity data according to its sensitivity level, or a numerical sensitivity level can be used. The specific method of classifying the sensitivity level of a field does not constitute a limitation of this invention.
[0070] For each target field in the data to be shared, the current de-identification status is obtained from the historical data processing records, and de-identification evidence supporting the current de-identification status is extracted.
[0071] For example, for the target field "contact number", if the historical processing record shows that the field has been processed by masking the phone number, then its current de-identification status can be recorded as "de-identified". The de-identification evidence corresponding to this status can include the de-identification rule version, de-identification algorithm type, de-identification processing time, data version at the time of processing, field length, character encoding, field semantic type, and de-identification effect verification result.
[0072] Simultaneously, the current status information used to verify the aforementioned de-identified evidence is obtained. This current status information can originate from the current data object, the current de-identification rule base, and the current data processing configuration, including the current de-identification rule, the current data structure, the current field semantics, the current field processing path, and the current de-identification effect verification status.
[0073] Therefore, it is possible to distinguish between the de-identified evidence from the historical de-identification state and the current actual data state.
[0074] For example, a mobile phone number field might have historically been an 11-digit numeric string when it underwent data masking, but now the field contains mobile phone numbers with the international country code "+86". Although historical data processing records still mark this field as "masked", the current data structure is different from the original data masking structure. Therefore, the current data cannot be directly determined to meet the sharing requirements based on the historical data masking flag.
[0075] In this embodiment, the de-identification evidence includes at least de-identification rule information, data version information, source field structure information, field semantic information, field processing path information, and de-identification effect verification information.
[0076] The information includes: the de-identification rule information (which may include the de-identification rule identifier, de-identification rule version, de-identification algorithm type, and de-identification parameters); the source field structure information (which may include the field type, field length, character encoding, value format, and structure pattern); the field semantic information (which indicates the data meaning of the field, such as mobile phone number, ID card number, name, address, account number, or device identifier); the field processing path information (which indicates the data processing process that the field undergoes to form the current target field from the source data); and the de-identification effect verification information (which may include the verification time, verification data version, verification result, and corresponding validity period).
[0077] S2. Construct a data source dependency graph and anonymized evidence chain.
[0078] Using source fields, intermediate processing fields, and target fields as field nodes, and establishing directed connections based on the data processing relationships between fields, a data source dependency graph is constructed:
[0079]
[0080] in, Represents a collection of field nodes. This represents the set of directed edges connecting field nodes.
[0081] A directed connection edge points from an upstream data field to a downstream data field, indicating that the data in the downstream field originates at least partially from the corresponding upstream field.
[0082] Data processing relationships between fields can include copying, mapping, transformation, concatenation, truncation, aggregation, encoding, decoding, hashing, replacement, generalization, perturbation, or referencing.
[0083] For example, the original mobile phone number field A is masked to obtain field B, field B is intercepted to obtain the last four digits field C, and field C is concatenated with the area code field D to form the customer auxiliary identifier field E. Then, directed connections are established from A to B, B to C, C to E, and D to E.
[0084] For each target field, trace back upstream along the data source dependency graph from that target field until one or more source fields are reached. Associate the field nodes, data processing operations, de-identification execution records, and de-identification effect verification records in the backtracking path according to the data processing order to form a corresponding de-identification evidence chain.
[0085] For example, the de-identified evidence chain corresponding to the target field E can include, in sequence, the source field A, the de-identified node B, the truncated node C, the splicing node E, and the shared output node.
[0086] In the de-identification evidence chain, de-identification execution nodes, de-identification effect verification nodes, and data processing nodes that change the field data structure or field semantics can be further marked.
[0087] Through the above processing, it is possible to determine which source fields, desensitization operations, and subsequent processing operations together form the current desensitization status of the target field, rather than judging solely based on the desensitization status identifier stored by the target field itself.
[0088] S3. Determine the validity of desensitized evidence and the points where evidence becomes invalid.
[0089] For field nodes in the de-identified evidence chain, the de-identified evidence saved when the historical de-identification status was formed is compared with the current status information.
[0090] In this embodiment, the validity of de-identification evidence can be determined from five aspects: consistency of de-identification rules, consistency of data structure, consistency of field semantics, consistency of field processing path, and verification status of de-identification effect.
[0091] No. Validity of anonymized evidence for each field node Determined according to the following formula:
[0092]
[0093] in, Indicates the first The validity of de-identified evidence for each field node; , , , and These correspond to the consistency of de-identification rules, data structure consistency, field semantic consistency, field processing path consistency, and the validity of the de-identification effect verification status, respectively. , , , and Let represent the corresponding non-negative weights, and satisfy:
[0094]
[0095] Each consistency score has a value range of [0,1], therefore:
[0096]
[0097] The consistency degree of the de-identification rules can be determined based on the version relationship and content differences between the de-identification rules used when the historical de-identification status was formed and the currently applicable de-identification rules. When the two are completely consistent, the consistency degree of the de-identification rules can be set to 1; when the rule versions are different but the de-identified objects, de-identification scope, and output constraints remain compatible, an intermediate value between 0 and 1 can be assigned; when the original de-identification rule is no longer applicable to the current field, the consistency degree should be reduced.
[0098] Data structure consistency can be determined based on information such as field data type, field length, encoding method, format template, and value distribution. For example, if a historical mobile phone number field is a fixed 11-digit numeric string, and an international country code or concatenation character is added to the current field, the data structure consistency can be reduced.
[0099] Field semantic consistency is used to determine whether the current actual semantics of a field are consistent with the semantics of the field recorded during the historical de-identification process. For example, if the name of a field remains unchanged, but its actual business meaning changes from an internal serial number to a personnel identity association number, then the field semantic consistency will be reduced accordingly.
[0100] Field processing path consistency can be compared between the data processing path during historical de-identification effect verification and the current actual processing path of the target field. For example, if the historical processing path was "de-identification - direct output", while the current processing path has become "de-identification - decoding - concatenation - output", then the field processing path consistency is reduced.
[0101] The validity of the desensitization effect verification status can be determined based on the verification time, the data version corresponding to the verification result, and the verification validity period. When the verification result has expired, or when the data version corresponding to the verification is significantly different from the current data version, the validity of the desensitization effect verification status can be reduced.
[0102] After obtaining the validity of the desensitized evidence, it is compared with a first validity threshold and a second validity threshold. Let the first validity threshold be... The second effective threshold is And satisfy:
[0103]
[0104] The de-identification status of the corresponding field node is determined to be valid when the following formula is met:
[0105]
[0106] When the following formula is met, the de-identification status of the corresponding field node is determined and awaits review:
[0107]
[0108] The corresponding field node is identified as a failed evidence node when the following condition is met:
[0109]
[0110] For example, the first effective threshold can be set to 0.8 and the second effective threshold can be set to 0.5. However, the above values are only one implementation method and can be adjusted according to the actual business scenario, data type and historical risk identification results.
[0111] At the same time, the cause of failure is determined based on the specific type of evidence that leads to a decrease in the validity of the de-identified evidence. For example, if the consistency of the de-identification rules is significantly reduced, it can be determined that the rules are invalid; if the consistency of the data structure is abnormal, it can be determined that the structure is invalid; if the field processing path changes significantly, it can be determined that the processing path is invalid; if the validity of the verification status of the de-identification effect is reduced, it can be determined that the verification status is invalid.
[0112] S4, Desensitization failure status
[0113] For the identified failure evidence nodes, the initial desensitization failure impact of the node is first determined.
[0114] In this embodiment, the first The initial desensitization failure impact of each failed evidence node is determined according to the following formula:
[0115]
[0116] in, Indicates the first The initial desensitization failure impact of each failed evidence node. This indicates the validity of the de-identified evidence for the invalid evidence node. Therefore, the lower the validity of the de-identified evidence for a field node, the greater the impact of its initial de-identification failure.
[0117] Starting from the invalid evidence node, the desensitization failure status is propagated to downstream field nodes along the directed connection edges in the data source dependency graph.
[0118] Since different data processing relationships retain upstream field information to varying degrees, this embodiment configures corresponding failure inheritance coefficients for different directed connection edges. ,satisfy:
[0119]
[0120] in, Indicates the first The desensitization failure status of the first field node is transmitted to the second field node through the corresponding data processing relationship. The degree of inheritance when propagating individual field nodes.
[0121] For example, for direct copying or equivalent mapping relationships, downstream fields basically retain the information characteristics of upstream fields, so a higher failure inheritance coefficient can be configured; for data processing relationships such as truncation and splicing that still retain some upstream characteristics, the corresponding failure inheritance coefficient can be configured according to the degree of original information retention; for statistical aggregation, strong random perturbation, or irreversible anonymization processing, a lower failure inheritance coefficient can be configured.
[0122] In a specific example, the failure inheritance coefficient for direct replication can be set to 0.95, for field truncation to 0.80, for field concatenation to 0.70, and for statistical aggregation to 0.30. The above values are only used to explain the parameter configuration methods under different data processing relationships and do not constitute a limitation on the scope of protection of this invention.
[0123] For the first with multiple direct upstream fields The impact of desensitization failure on individual field nodes. It can be determined according to the following formula:
[0124]
[0125] in, Indicates the relationship with the first The set of upstream field nodes directly connected to each field node; Indicates the first The impact of desensitization failure on upstream field nodes; Indicates the first The field node up to the first Failure inheritance coefficient between field nodes.
[0126] For field nodes that are not affected by the propagation of the desensitization failure status, their desensitization failure impact degree is set to:
[0127]
[0128] For example, if target field E is generated from both field C and field D, where field C is affected by the desensitization failure status, while field D is not affected by the desensitization failure status, then the desensitization failure impact degree corresponding to field D is set to 0, so that the desensitization failure impact degree of target field E will not be increased.
[0129] For downstream field nodes, the impact of desensitization failure is calculated iteratively layer by layer according to the data processing order in the data source dependency graph until the target field is reached or the preset propagation termination condition is met.
[0130] When a downstream field node re-executes effective desensitization processing and passes the desensitization effect verification, that node can be used as a failure propagation cutoff node to stop the upstream desensitization failure state from continuing to propagate along that path.
[0131] Furthermore, when the calculated impact of the desensitization failure is lower than a preset propagation threshold, propagation along the corresponding path can be stopped. For example, the propagation threshold can be set to 0.05 to reduce the amount of unnecessary calculations caused by the continued propagation of extremely weak failures.
[0132] S5. Determine the risk level of sharing.
[0133] After obtaining the validity of the de-identification evidence for each target field and the impact of de-identification failure, the risk of the sharing scenario is determined based on the data sharing request.
[0134] Risks associated with data sharing scenarios can be determined based on information such as the recipient of the data, the purpose of sharing, the scope of sharing, the data output method, whether downloading is allowed, the validity period of sharing, and whether secondary forwarding is allowed.
[0135] The same data sharing request can first form an overall sharing scenario risk, and then further combine the field attributes of each target field to determine the sharing scenario risk corresponding to each target field.
[0136] For example, the risk of a data sharing scenario can be relatively low when the data recipient only has viewing rights but cannot download the data; however, the risk can increase when the data can be downloaded and further forwarded to other entities.
[0137] Convert the field sensitivity level of each target field into a normalized field sensitivity. The corresponding shared scenario risks are converted into normalized shared scenario risk values. ,in:
[0138]
[0139]
[0140] No. Shared risk value of each target field It can be determined according to the following formula:
[0141]
[0142] in, Indicates the first The validity of anonymized evidence for each target field; Indicates the first The impact of desensitization failure on individual target fields; Indicates the first Normalized field sensitivity of each target field; Indicates the first Normalized shared scenario risk values corresponding to each target field; , , and These are the corresponding non-negative weights, and they satisfy:
[0143]
[0144] Based on the aforementioned definition, it is possible to:
[0145]
[0146] In one implementation, the following can be set:
[0147]
[0148] The above parameters give relatively high weight to the degree of failure of the field's own desensitization status and the impact of upstream failure propagation.
[0149] Furthermore, risk levels can be categorized based on shared risk values. For example, when A value less than 0.35 is considered low risk; when A value greater than or equal to 0.35 and less than 0.65 is defined as medium risk; when A value greater than or equal to 0.65 is considered high risk. These risk classification thresholds can be adjusted based on actual data security management requirements.
[0150] S6. Generate a shared risk warning interface
[0151] A shared risk warning interface is generated based on the shared risk level of each target field, the failure evidence node, and the propagation path of the de-identified failure status.
[0152] The shared risk warning interface can include a field risk display area, a failure propagation path display area, and a risk handling area.
[0153] The field risk display area is used to show the target field name, current de-identification status, validity of de-identification evidence, impact of de-identification failure, field sensitivity level, and sharing risk level.
[0154] The failure propagation path display area is used to show the upstream failure evidence nodes that cause the target field to be at risk, as well as the propagation process of the de-identified failure status.
[0155] For example, the data processing path can be displayed as follows: Original mobile number field → Historical anonymized field → Last digits of the phone number field → Customer auxiliary identifier field → Field to be shared.
[0156] Among them, invalid evidence nodes can be displayed separately, and the corresponding reasons for invalidation can be further displayed, such as changes in data structure, incompatibility of desensitization rule version, changes in field processing path, or expiration of desensitization effect verification results.
[0157] The risk handling area can receive processing instructions for risky fields. These instructions can correspond to one or more of the following processing operations: re-de-identification, re-verification, removal of risky fields, adjustment of sharing scope, submission for manual review, or termination of sharing.
[0158] For example, when a user performs a re-desensitization operation on a high-risk field, the system reprocesses the corresponding field based on the currently applicable desensitization rules and generates a new desensitization processing record and desensitization effect verification results.
[0159] Subsequently, the processed de-identified evidence was retrieved again, and the validity of the de-identified evidence for the corresponding fields was recalculated.
[0160] If the re-de-identified field passes the de-identification effect verification, it will be used as the failure propagation cutoff node, and the de-identification failure impact and sharing risk level of the field nodes located downstream of this node will be recalculated.
[0161] Update the shared risk warning interface based on the recalculation results, ensuring that the risk status, failure propagation path, and risk level displayed on the interface are consistent with the current data processing status.
[0162] Therefore, the shared risk warning interface in this embodiment is not only used to statically display risk information, but can also be linked with the verification of de-identified evidence, failure propagation analysis, and the process of recalculating shared risks.
[0163] Specific application examples
[0164] The following example, using a customer data sharing scenario, further illustrates the processing procedure of this invention.
[0165] The data to be shared includes fields such as customer ID, contact number, region, and customer level.
[0166] The contact number field was processed using data masking rule R1 in historical data version D1. This data masking rule applies to 11-digit mobile phone numbers, replacing certain numbers in the middle of the number with characters, while retaining a preset number of prefix and suffix characters.
[0167] After the contact number field is anonymized, it becomes intermediate field B. Intermediate field B is further truncated to form the last four digits field C. The last four digits field C is concatenated with the area code field D to form the customer auxiliary identifier field E. The customer auxiliary identifier field E is finally included as the target field in the data to be shared.
[0168] In the current data version D2, due to the addition of overseas customers to the business system, phone numbers with country codes have begun to appear in the contact phone number field, such as "+86-13812345678".
[0169] Since the historical desensitization rule R1 mainly processed based on a fixed 11-digit number format, although the system still records the contact number field as "desensitized", there are already differences between the original desensitization rule and the current field structure.
[0170] After receiving a data sharing request, the system retrieves historical de-identification evidence for the contact number field from the data processing records, including de-identification rule R1, historical field length, field structure pattern, data version D1, and historical de-identification effect verification results. It also retrieves the data structure of the current field and the currently applicable de-identification rule.
[0171] Assuming that the comparison yields the following:
[0172]
[0173]
[0174]
[0175]
[0176]
[0177] These represent the consistency of the current de-identification rules, data structure, field semantics, field processing path, and the validity of the de-identification effect verification status, respectively. Settings:
[0178]
[0179] The validity of de-identified evidence in the contact number field is:
[0180]
[0181] Right now:
[0182]
[0183] If the first valid threshold is set to 0.8 and the second valid threshold is set to 0.5, then the field is in a state of pending review.
[0184] If further testing reveals that the current data masking rules are completely ineffective at masking certain international number formats, the consistency or structural consistency of the data masking rules will further decrease, leading to... If the value is less than 0.5, the contact number field will be identified as an invalid evidence node.
[0185] Assuming the validity of the desensitized evidence decreases to:
[0186]
[0187] The initial desensitization failure impact is:
[0188]
[0189] The desensitization failure status then propagates along the path from the contact number field to the last three digits field, and then to the customer auxiliary identifier field.
[0190] If the relationship between the contact number field and the last four digits field is truncation, the corresponding failure inheritance coefficient can be set to 0.8; if the relationship between the last four digits field and the customer auxiliary identifier field is concatenation, the corresponding failure inheritance coefficient can be set to 0.7. The desensitization failure impact of each field is calculated layer by layer accordingly.
[0191] For the customer auxiliary identifier field, if there is also a normal area code field D as an upstream field, the desensitization failure impact of field D will be set to 0. Therefore, this normal field will not add any additional failure impact to the customer auxiliary identifier field.
[0192] Furthermore, by combining the validity of the de-identified evidence of the customer auxiliary identification field, the impact of de-identification failure, the sensitivity of the field, and the risk of the sharing scenario, the sharing risk value is calculated and the sharing risk level is determined.
[0193] If the Customer Auxiliary Identifier field is ultimately determined to be a high-risk field, the shared risk warning interface can display the failure propagation path formed by "Contact Number Field → Last Digit of Phone Number Field → Customer Auxiliary Identifier Field", and show the Contact Number Field as a failure evidence node, with the failure reason being "the current field data structure is inconsistent with the data structure during the original desensitization process, and the historical desensitization rules do not cover the newly added number format sufficiently".
[0194] The customer is also notified that the auxiliary identification field is affected by the aforementioned failure status propagation.
[0195] Users can select to re-identify the data in the risk handling area. The system reprocesses the contact number field using the de-identification rules applicable to the current international number format and re-verifies the de-identification effect.
[0196] Once the new de-identification results are verified, the re-de-identification node will be used as the failure propagation cutoff node, and the impact of the de-identification failure of the telephone number tail field and the customer auxiliary identification field, as well as the sharing risk level, will be recalculated.
[0197] If the recalculated risk level of sharing drops to the permissible sharing range, the sharing risk warning interface will be updated, and the original high-risk warning will be adjusted to the corresponding low-risk or permissible sharing status.
[0198] Through the above methods, the present invention can re-verify the historical de-identification status when the data structure, de-identification rules, field semantics, or data processing path changes, and identify the impact of the de-identification status failure on downstream fields through data source dependency relationships, thereby improving the accuracy, traceability, and consistency between risk warning results and the current data status of data sharing risk identification.
[0199] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.
Claims
1. A method for a shared risk warning interface based on data anonymization status, characterized in that, Includes the following steps: S1. Obtain the data to be shared, the data sharing request, the sensitivity level of each target field, and the data processing record corresponding to the data to be shared. For each target field in the data to be shared, obtain the current de-identification status from the data processing record and extract the de-identification evidence supporting the current de-identification status. Obtain the current status information used to verify the de-identification evidence. The de-identification evidence includes at least de-identification rule information, data version information, source field structure information, field semantic information, field processing path information, and de-identification effect verification information. S2. Using source fields, intermediate processing fields, and target fields as field nodes, establish directed connection edges based on the data processing relationships between fields to construct a data source dependency graph; for each target field, trace back upstream along the data source dependency graph, and associate the field nodes, data processing operations, and de-identified evidence on the backtracking path according to the data processing order to form a corresponding de-identified evidence chain; S3. Compare the de-identified evidence in the de-identified evidence chain with the current status information to determine the validity of the de-identified evidence for each field node, and determine the invalid evidence node based on the validity of the de-identified evidence; wherein, the current status information at least represents the current de-identification rule, the current data structure, the current field semantics, the current field processing path, and the current de-identification effect verification status; S4. Taking the failure evidence node as the propagation starting point, determine the failure inheritance coefficient according to the data processing relationship corresponding to the directed connection edge, propagate the desensitization failure status downstream along the data source dependency graph, and obtain the desensitization failure impact degree of each target field. S5. Determine the sharing scenario risk based on the data sharing request, and determine the sharing risk level of each target field based on the validity of the de-identified evidence, the impact of de-identification failure, the sensitivity level of the field, and the sharing scenario risk. S6. Generate a shared risk warning interface based on the shared risk level, failure evidence node, and propagation path of the desensitized failure status; wherein, the shared risk warning interface is used to display at least the risk field, the cause of the desensitized failure status, the failure propagation path, the affected field, and the corresponding risk handling information.
2. The method for a shared risk warning interface based on data anonymization status according to claim 1, characterized in that, The de-identification rule information includes at least one of the following: de-identification rule identifier, de-identification rule version, and de-identification parameters; the source field structure information includes at least one of the following: field data type, field length, encoding format, and structure pattern; the de-identification effect verification information includes the verification result, the data version corresponding to the verification, and the verification time. The validity of the desensitized evidence is determined according to the following formula: in, Indicates the first The validity of de-identified evidence for each field node; , , , and These represent the consistency of de-identification rules, data structure consistency, field semantic consistency, field processing path consistency, and the validity of the de-identification effect verification status, respectively. , , , and Let be the corresponding non-negative weights, and satisfy: in, , , , and The range of values is .
3. The method for a shared risk warning interface based on data anonymization status according to claim 1, characterized in that, The data processing relationships include at least one of the following: copying, mapping, transformation, splicing, truncation, aggregation, encoding, decoding, hashing, replacement, generalization, perturbation, and referencing; The directed connection edges are established according to the data processing direction from upstream fields to downstream fields, and the de-identification execution nodes, de-identification effect verification nodes, and data processing nodes that change the field data structure or field semantics are marked in the de-identification evidence chain.
4. The method for a shared risk warning interface based on data anonymization status according to claim 2, characterized in that, Determining invalid evidence nodes based on the validity of the de-identified evidence includes: The validity of the desensitized evidence is compared with a first valid threshold and a second valid threshold, wherein the first valid threshold is greater than the second valid threshold; When the validity of the de-identified evidence is not lower than the first valid threshold, the de-identification status of the corresponding field node is determined to be valid; When the validity of the de-identified evidence is lower than the first valid threshold and not lower than the second valid threshold, the de-identification status of the corresponding field node is determined to be pending review; When the validity of the de-identified evidence is lower than the second validity threshold, the corresponding field node will be identified as an invalid evidence node. Based on the type of evidence that reduces the effectiveness of the de-identified evidence, determine at least one of the following failure causes: rule failure, structural failure, semantic failure, processing path failure, or verification status failure.
5. A method for a shared risk warning interface based on data anonymization status according to claim 1, characterized in that, The initial desensitization failure impact of the failed evidence node is determined as follows: in, Indicates the first The initial desensitization failure impact of each failed evidence node. Indicates the first The validity of de-identified evidence at each invalid evidence node; For field nodes located downstream of the failure evidence node, the impact of the de-identification failure is determined iteratively according to the data processing order: in, Indicates the first The impact of desensitization failure on individual field nodes; Indicates the relationship with the first The set of upstream field nodes directly connected to each field node; Indicates the first The impact of desensitization failure on upstream field nodes; Indicates the first The field node up to the first The failure inheritance coefficient between each field node, and .
6. A method for a shared risk warning interface based on data anonymization status according to claim 5, characterized in that, Configure the failure inheritance coefficient according to the data processing relationship corresponding to the directed connection edge; When a downstream field node re-executes the desensitization process and passes the desensitization effect verification, the corresponding failure propagation path passing through the downstream field node is cut off; or, when the desensitization failure impact is lower than a preset propagation threshold, the propagation along the corresponding failure propagation path is stopped.
7. A method for a shared risk warning interface based on data anonymization status according to claim 1, characterized in that, The sensitivity levels of each target field are converted into normalized field sensitivity, and the shared scenario risk is converted into a normalized shared scenario risk value. The shared risk value for each target field is determined according to the following formula: in, Indicates the first Shared risk values for each target field; Indicates the first The validity of anonymized evidence for each target field; Indicates the first The impact of desensitization failure on individual target fields; Indicates the first Normalized field sensitivity of each target field; Indicates the first Normalized shared scenario risk values corresponding to each target field; , , and The weights are non-negative and satisfy: in, and The range of values is ; The shared risk level of the corresponding target field is determined based on the comparison result between the shared risk value and the preset risk threshold.
8. A method for a shared risk warning interface based on data anonymization status according to claim 1, characterized in that, The shared risk warning interface includes a field risk display area, a failure propagation path display area, and a risk handling area; The risk processing area is used to receive processing instructions for risk fields. The processing instructions correspond to at least one of the following processing operations: re-de-identification, re-verification, removal of risk fields, adjustment of sharing scope, submission for manual review, or termination of sharing. In response to the processing instruction, a processing operation is performed on the corresponding target field, the processed de-identified evidence is re-extracted, the validity of the de-identified evidence, the impact of de-identification failure, and the sharing risk level of the target field are updated, and the sharing risk warning interface is updated according to the update results.
9. A shared risk warning interface system based on data anonymization status, characterized in that, include: The data acquisition module is used to acquire the data to be shared, the data sharing request, the field sensitivity level of each target field, and the data processing record, and to acquire the current de-identification status, de-identification evidence, and current status information for verifying the de-identification evidence of each target field. The dependency construction module is used to build a data source dependency graph based on the data processing relationships between fields, and to form a de-identified evidence chain for each target field; The evidence verification module is used to compare the de-identified evidence in the de-identified evidence chain with the current status information to determine the validity of the de-identified evidence and the invalid evidence nodes of each field node. The failure propagation module is used to determine the failure inheritance coefficient based on the data processing relationship, and propagate the desensitization failure status downstream along the data source dependency graph to obtain the desensitization failure impact degree of each target field. The risk determination module is used to determine the sharing risk level of each target field based on the validity of the de-identified evidence, the impact of de-identification failure, the sensitivity level of the field, and the risk of the sharing scenario corresponding to the data sharing request. The interface generation module is used to generate a shared risk warning interface based on the shared risk level, failure evidence nodes, and failure propagation path.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method described in any one of claims 1 to 8.
