Data access control method and apparatus for trusted data space
By constructing a target call chain and assessing the risks of data operation behavior, the problem of unauthorized data use in existing technologies is solved, and more precise data access control and risk management are achieved.
Patent Information
- Application Number
- CN202611603014.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-10-16
- Publication Date
- 2026-11-17
AI Technical Summary
Existing data access control mechanisms are prone to the risk of unauthorized use of data and cannot effectively prevent the leakage of sensitive data, especially in scenarios involving cross-domain circulation and multi-party sharing, where precise control is difficult to achieve.
By constructing a target call chain and combining it with the operational context information of data operation behavior, risk assessment parameters are determined, and data access results are dynamically adjusted to identify risks arising from the superposition or association of multiple data operation behaviors.
It improves the dynamism and accuracy of data access control, reduces the risk of unauthorized or improper data use due to continuous data operation behavior, and enhances the ability to identify risks in data use behavior.
Smart Images

Figure CN123116342A_ABST
Abstract
Description
Technical Field
[0001] The embodiments in this specification relate to the field of data security technology, and in particular to a data access control method and apparatus for trusted data spaces. Background Technology
[0002] With the advancement of market-based allocation of data elements and the construction of a trusted data space, data access control has become a core technical aspect for ensuring data security and promoting compliant data circulation. In scenarios involving cross-domain data circulation, multi-party sharing, and collaborative computing, how to accurately control data access behavior and prevent unauthorized access and sensitive data leakage is crucial to the secure operation of a trusted data space.
[0003] Existing data access control mechanisms typically authenticate each access request as an independent event. However, this approach is prone to the risk of unauthorized data use. Therefore, there is an urgent need for a data access control method for trusted data spaces that can enhance the ability to manage data usage behavior. Summary of the Invention
[0004] In view of this, embodiments of this specification provide a data access control method for a trusted data space. One or more embodiments of this specification also relate to a data access control device for a trusted data space, a computing device, a computer-readable storage medium, and a computer program product, to address the technical deficiencies existing in the prior art.
[0005] According to a first aspect of the embodiments of this specification, a data access control method for a trusted data space is provided, comprising: In response to a target data access request initiated by the target object for the target task, the target call chain is determined. The target call chain is determined based on at least one data operation behavior involved in the execution of the target task and the corresponding operation context information. The operation context information is information related to the operation control of the corresponding data operation behavior. Based on the target call chain, determine at least one risk assessment parameter corresponding to the target data access request; Based on at least one risk assessment parameter, determine the risk assessment information corresponding to the target data access request, and based on the risk assessment information, determine the target access result corresponding to the target data access request.
[0006] According to a second aspect of the embodiments of this specification, a data access control device for a trusted data space is provided, comprising: The call chain determination module is configured to determine the target call chain in response to a target data access request initiated by the target object for the target task. The target call chain is determined based on at least one data operation behavior involved in the execution of the target task and the corresponding operation context information. The operation context information is information related to the operation control of the corresponding data operation behavior. The risk parameter determination module is configured to determine at least one risk assessment parameter corresponding to the target data access request based on the target call chain. The access result determination module is configured to determine the risk assessment information corresponding to the target data access request based on the at least one risk assessment parameter, and to determine the target access result corresponding to the target data access request based on the risk assessment information.
[0007] According to a third aspect of the embodiments of this specification, a computing device is provided, comprising: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the steps of the above-described data access control method for trusted data space are implemented.
[0008] According to a fourth aspect of the embodiments of this specification, a computer-readable storage medium is provided, the computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the steps of the data access control method for a trusted data space described above.
[0009] According to a fifth aspect of the embodiments of this specification, a computer program product is provided, including a computer program / instructions that, when executed by a processor, implement the steps of the data access control method for a trusted data space described above.
[0010] One embodiment of this specification implements a data access control method for a trusted data space. In response to a target object's data access request for a target task, the method determines the target call chain based on the data operation behaviors involved in the execution of the target task and the corresponding operation context information. Furthermore, it determines risk assessment parameters, risk assessment information, and the target access result based on the target call chain. This allows data access control to go beyond independently judging a single data access request, enabling a comprehensive evaluation of the target data access request by combining continuous data operation behaviors and their relationships formed during the execution of the target task. This allows for the identification of data usage risks arising from the superposition or association of multiple data operation behaviors, improving the ability to identify risks associated with the target object's data usage behavior during task execution. Simultaneously, determining the target access result based on the risk assessment information allows data access control to adjust accordingly to changes in the risk state during task execution, thereby improving the dynamism and accuracy of data access control and reducing the risk of unauthorized or improperly associated data use due to continuous data operation behaviors. Attached Figure Description
[0011] Figure 1 This is a flowchart illustrating a data access control method for a trusted data space, provided in one embodiment of this specification. Figure 2 This is a schematic diagram of an adaptive control system architecture for Agent data access risk provided in one embodiment of this specification; Figure 3 This is a flowchart illustrating the processing procedure of a data access control method for a trusted data space, provided in one embodiment of this specification. Figure 4 This is a schematic diagram of the structure of a data access control device for a trusted data space provided in one embodiment of this specification; Figure 5 This is a structural block diagram of a computing device provided in one embodiment of this specification. Detailed Implementation
[0012] Many specific details are set forth in the following description to provide a full understanding of this specification. However, this specification can be implemented in many other ways than those described herein, and those skilled in the art can make similar extensions without departing from the spirit of this specification. Therefore, this specification is not limited to the specific implementations disclosed below.
[0013] The terminology used in one or more embodiments of this specification is for the purpose of describing particular embodiments only and is not intended to be limiting of the one or more embodiments of this specification. The singular forms “a” and “the” as used in one or more embodiments of this specification and the appended claims are also intended to include the plural forms, unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in one or more embodiments of this specification refers to and includes any or all possible combinations of one or more associated listed items.
[0014] It should be understood that although the terms first, second, etc., may be used to describe various information in one or more embodiments of this specification, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, first may also be referred to as second without departing from the scope of one or more embodiments of this specification, and similarly, second may also be referred to as first. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to a determination."
[0015] Furthermore, it should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in one or more embodiments of this specification are all information and data authorized by the user or fully authorized by all parties. Moreover, the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and corresponding operation entry points are provided for users to choose to authorize or refuse.
[0016] This specification relates to a data access control method for a trusted data space, a data access control device for a trusted data space, a computing device, a computer-readable storage medium, and a computer program product, which are described in detail below with reference to specific embodiments.
[0017] See Figure 1 , Figure 1 A flowchart of a data access control method for a trusted data space according to an embodiment of this specification is shown, specifically including the following steps 102-106.
[0018] Step 102: In response to the target object's request for accessing target data for the target task, determine the target call chain. The target call chain is determined based on at least one data operation behavior involved in the execution of the target task and the corresponding operation context information. The operation context information is information related to the operation control of the corresponding data operation behavior.
[0019] It should be noted that the term "target object" in the embodiments of this specification refers to an access subject capable of initiating a data access request or triggering data operation behavior for data resources, and is not limited to a specific type of subject. The target object can be one or more of the following: user, application, business system, automation program, agent, data processing service, task scheduling system, workflow execution engine, or other data access subject capable of initiating or triggering data access.
[0020] In this context, an Agent can be understood as an intelligent execution entity capable of autonomously planning tasks, recognizing states, invoking tools, and processing data based on the target task. For example, an Agent can determine the data resources that need to be accessed based on the target task proposed by the user, and sequentially invoke multiple data resources or data processing services through API interfaces, service components, data product interfaces, Model Context Protocols (MCPs), etc. During the Agent's task execution, the data obtained or the processing results of the previous data access can also influence subsequent data access behaviors.
[0021] It should be noted that this specification does not require the target object to possess intelligent decision-making capabilities. The data access control methods for trusted data spaces provided in this specification can also be used for automated programs, business systems, or task scheduling systems that automatically execute data access according to preset procedures, rules, or task flows. For example, a business system sequentially queries user information, order information, and transaction information according to a preset business flow; or a data processing program retrieves data from multiple data sources and performs aggregated processing according to a preset task flow. In these cases, the corresponding business system or data processing program can also be taken as the target object, and multiple data operation behaviors during its execution can be associated to determine the corresponding target call chain.
[0022] Furthermore, in some implementations, the target object can also be the user initiating data access or other data access subjects. For example, a user initiates a data query task in a data analysis system, and the system accesses multiple data tables or data services based on the user's query operation. In this case, the target call chain can be determined based on the data operation behaviors generated during the execution of the data query task and their corresponding operation context information.
[0023] Therefore, the data access control method for trusted data spaces described in this specification can be applied to different data access methods, and is not limited to Agent scenarios. In particular, when the target object is an automated access subject capable of continuously executing multiple data access operations, the embodiments of this specification can utilize the correlation between multiple data operation behaviors on the target task dimension to perform overall analysis and control of the data access process.
[0024] In the embodiments of this specification, "target task" refers to the task objective or task process corresponding to the target object performing data access or data processing. Target tasks can be initiated by users or triggered by business systems, scheduling systems, automation programs, or other system components.
[0025] For example, the target task may include "generating a credit assessment report for a user", "analyzing sales in a certain region", "generating a business operation analysis report", "generating an operation report based on order data", or "completing a cross-data source query", etc.
[0026] During the execution of a target task, one or more data manipulation actions may be generated. Different data manipulation actions can be applied to different data resources, or different types of operations can be performed on the same data resource. Multiple data manipulation actions can occur in a certain execution order, and the data or processing result generated by the previous data manipulation action can serve as the input, reference information, or triggering condition for the subsequent data manipulation action.
[0027] Therefore, the target task can be used as a task-level associated object that links multiple data operation behaviors, so that when controlling the current target data access request, not only the current request itself can be considered, but also other data operation behaviors that have occurred during the execution of the current target task.
[0028] In this specification, data manipulation behavior is used to describe the operations performed by a target object on data resources during the execution of a target task, related to data acquisition, data processing, data transfer, or continued use of data. Data manipulation behavior may include one or more of data access, tool invocation, and result referencing.
[0029] Data access refers to the actions taken by a target object to retrieve data from data resources. For example, a target object can retrieve data from relevant data resources through database query interfaces, data service interfaces, file access interfaces, or other data access interfaces.
[0030] Tool invocation refers to the act of a target object calling an external tool, service, or functional component to process, analyze, transform, or perform other operations based on data. In some implementations, tool invocation can provide the invoked tool or service with data obtained from previous data operations, or it can obtain new data or processing results from the tool or service. For example, a target object can call a data analysis tool to perform aggregate analysis on already acquired order data and obtain order statistics results.
[0031] Result referencing refers to the act of a target object continuing to use data or processing results generated by previous data operations during the subsequent execution of a target task. Result referencing does not require the target object to initiate a new data access request to the data resource that generated the data or processing result. For example, after obtaining customer data in a previous task step, the target object can directly reference the customer data in the subsequent process of generating an analysis report; or, for example, after obtaining statistical results by calling a data analysis tool, the target object can directly reference the statistical results in subsequent steps.
[0032] Therefore, data access, tool invocation, and result referencing can be used to describe the acquisition, external processing or transmission, and subsequent use of data during the execution of the target task, respectively. By associating these different types of data operation behaviors, we can describe the continuous data operation behaviors of the target object in the target task from the perspective of the data usage process, rather than being limited to recording whether the target object has initiated a single access request for a certain data resource.
[0033] In some implementations, the data operation behaviors involved in the execution of the target task can be identified as nodes in the target call chain, and the execution relationship, call relationship, data flow relationship or result reference relationship between different data operation behaviors can be identified as the association relationship between nodes, thereby forming the target call chain.
[0034] For example, if the target object first obtains the first data from the first data resource through data access behavior, then provides the first data to the first data processing tool through tool invocation behavior to obtain the first processing result, and continues to use the first processing result through result referencing behavior in subsequent task steps, then the above data access behavior, tool invocation behavior and result referencing behavior can be associated with the target invocation chain respectively, and corresponding association relationships can be established based on the flow relationship of the first data between different operation behaviors.
[0035] It should be noted that data access, tool invocation, and result referencing are not mutually exclusive task execution phases, but rather different types of data manipulation behaviors used to describe the execution of a target task. A target task may contain only one or more of these data manipulation behaviors, which can be determined based on the task execution method and data usage method of the target object.
[0036] By incorporating the aforementioned data manipulation behaviors into the target call chain, the source, processing, transmission, and subsequent use of data during the execution of the target task can be tracked. This allows subsequent risk analysis to consider not only the data resources corresponding to the current data access request, but also the data manipulation behaviors that have already occurred in the target task and the flow and correlation of data between different manipulation behaviors. This enables the analysis of potential risks of data aggregation, related use, or information inference during continuous data use.
[0037] It should be noted that the target call chain can be created after receiving a target data access request, based on the data operations and corresponding operation context information that have already occurred in the target task; alternatively, it can be an existing call chain established before receiving the target data access request, based on the data operations and corresponding operation context information that have occurred previously during the execution of the target task. In the latter case, the existing call chain can be updated based on the data operations and operation context information corresponding to the target data access request. For example, the call node corresponding to the target data access request can be added to the existing call chain, and the association relationships between nodes can be updated according to the operation dependencies or data dependencies between the data operations corresponding to the target data access request and the existing call nodes, thereby obtaining the target call chain. Therefore, the target call chain can be either a call chain initially constructed for the target task or a call chain that is continuously updated as the target task continues to execute, to continuously reflect the data operations and their association relationships throughout the lifecycle of the target task.
[0038] In this specification, operation context information is used to describe information related to the execution of data operation behavior, data flow, or operation control, so as to characterize the task background, operation object, operation method, and relationship with other data operation behaviors in which the corresponding data operation behavior occurs.
[0039] For example, for data access behavior, the operation context information may include one or more of the following: the task identifier of the target task, the resource identifier of the data resource, the sensitivity of the data resource, the access time, the access method, the access subject, the access purpose, and the corresponding authorization information; for tool invocation behavior, the operation context information may include one or more of the following: the tool identifier of the invoked tool, the invocation parameters, the invocation time, the invocation result, and the association between the tool and the data resource; for result referencing behavior, the operation context information may include one or more of the following: the referenced operation result, the preceding data operation behavior that generated the operation result, the result referencing time, and the subsequent operation in which the operation result was used.
[0040] By obtaining the operation context information corresponding to each data operation behavior, the data flow relationship, call relationship or operation dependency relationship between different data operation behaviors can be determined, and a target call chain for the target task can be constructed based on the data operation behavior and the corresponding operation context information.
[0041] In one optional implementation of this specification, determining the target call chain includes: Based on at least one data operation behavior, identify at least one calling node in the target call chain; Based on the operation context information corresponding to at least one data operation behavior, determine the node attributes corresponding to at least one calling node; Based on the operation context information corresponding to at least one data operation behavior, determine the node dependency attributes of at least one calling node; The target call chain is determined based on at least one call node, the node attributes corresponding to at least one call node, and the node dependency attributes of at least one call node.
[0042] In the embodiments of this specification, at least one calling node in the target calling chain can be determined based on at least one data operation behavior involved in the execution of the target task, wherein one calling node corresponds to one data operation behavior instance.
[0043] For example, when a target object sequentially performs data access, tool invocation, and result referencing, corresponding call nodes can be generated for each of these three actions. Thus, different call nodes in the target call chain can correspond to different types of data operation behaviors, without requiring different call nodes to have the same attribute information.
[0044] Furthermore, based on the operation context information corresponding to each data operation behavior, the node attributes corresponding to each calling node can be determined. Node attributes are used to describe the basic characteristics of the corresponding calling node itself, to characterize what operation the calling node corresponds to, which task it belongs to, and when it occurred.
[0045] For example, for a data access node, node attributes may include one or more of the following: node identifier, task identifier, time information, operation type, data resource identifier, and the data sensitivity level corresponding to the data resource. The data resource identifier identifies the accessed data resource, and the data sensitivity level characterizes the degree of data sensitivity corresponding to the data resource.
[0046] For a tool invocation node, node attributes may include one or more of the following: node identifier, task identifier, time information, operation type, tool identifier, invocation parameters, input data identifier, and output result identifier. The input data identifier can be used to identify the data used in the tool invocation or the result of a preceding operation, while the output result identifier can be used to identify the operation result generated by the tool invocation. Optionally, the tool invocation node can also be associated with the sensitivity level of the input data to help determine the sensitivity of the data involved in the tool invocation later.
[0047] For a result referencing node, node attributes may include one or more of the following: node identifier, task identifier, time information, operation type, referenced result identifier, and the source node identifier corresponding to the referenced result. The referenced result identifier identifies the operation result used by the current result referencing behavior, and the source node identifier identifies the preceding calling node that generated the operation result.
[0048] In one implementation, different types of calling nodes can have the same basic node attributes and different extended node attributes. For example, different types of calling nodes can all have basic node attributes such as node identifier, task identifier, time information, and operation type, while extended node attributes such as data resource identifier, tool identifier, and referenced result identifier are set according to the different data operation behaviors corresponding to the calling node.
[0049] Furthermore, the node dependency attributes of each calling node can be determined based on the operation context information corresponding to each data operation behavior.
[0050] The node dependency attribute is used to describe the operational and data dependencies between a calling node and other calling nodes, to characterize which preceding calling nodes the data or operation results of the current calling node come from, and whether the data or operation results generated by the current calling node are used by subsequent calling nodes.
[0051] For example, if the target object first obtains order data through data access behavior and uses the order data as input to call a data analysis tool, it can be determined that there is a data dependency relationship between the first calling node corresponding to the data access behavior and the second calling node corresponding to the tool calling behavior, and the first calling node is the preceding dependent node of the second calling node.
[0052] For example, if the target object calls a data analysis tool to obtain statistical results and directly references the statistical results in subsequent operations, it can be determined that there is a result reference relationship between the calling node corresponding to the tool calling behavior and the calling node corresponding to the result referencing behavior.
[0053] In one implementation, the node dependency attribute may include one or more of the following: dependent node identifier, dependency relationship type, dependency direction, and dependency strength. The dependent node identifier identifies other calling nodes that have a dependency relationship with the current calling node; the dependency relationship type characterizes the data flow relationship, tool call relationship, result reference relationship, or operation dependency relationship between different calling nodes; the dependency direction characterizes the direction of data or operation results flow between different calling nodes; and the dependency strength characterizes the degree to which the current calling node depends on the data or operation results corresponding to the dependent node.
[0054] For example, when the input data of the second calling node includes data generated by the first calling node, a directed dependency relationship can be established between the first and second calling nodes, pointing from the first calling node to the second calling node. When a preset proportion of the input data of the second calling node originates from the output results of the first calling node, the corresponding dependency strength can be determined based on the preset proportion. The dependency strength can also be determined based on data lineage, the correspondence between input and output data, or preset association rules; this specification does not limit this.
[0055] Based on the above method, corresponding associations can be established between different calling nodes according to the node attributes and node dependency attributes of each calling node, and at least one calling node can be connected according to the association to obtain the target calling chain.
[0056] It should be noted that the target call chain is not simply arranged according to the occurrence time of each data operation, but rather determines the correlation between different data operation operations based on the operation context information corresponding to each data operation, and connects the corresponding call nodes according to the correlation. Therefore, the target call chain can simultaneously reflect the occurrence order of data operation operations during the execution of the target task, as well as the data source relationship, result reference relationship, and operation dependency relationship between different data operation operations, thereby forming a data usage path for the target task.
[0057] By applying the embodiments of this specification, call nodes are generated based on the data operation behaviors involved in the execution of the target task. Node attributes are used to describe the data operation behaviors corresponding to each call node, and node dependency attributes are used to describe the data dependencies, operation dependencies, and result references between different call nodes. This allows data access, tool calls, and result references that occur continuously within the lifecycle of the target task to be associated with the call chain corresponding to the same target task. This prevents data operation behaviors from being analyzed as independent operations, but rather reflects the process of data acquisition, flow, processing, and continued use during the execution of the same target task. Thus, the lifecycle of the target task is used as the scope of data use control. By comprehensively considering the correlation between continuous data operation behaviors, a more complete data use path is obtained, thereby improving the accuracy of judging data use risks during the execution of the target task.
[0058] In one optional implementation of this specification, node dependency attributes include node dependency relationships and node dependency strength; based on the operation context information corresponding to at least one data operation behavior, determining the node dependency attributes of at least one calling node includes: Based on the operation context information corresponding to at least one data operation behavior, determine the operation dependency relationship of at least one data operation behavior, and based on the operation dependency relationship of at least one data operation behavior, determine the node dependency relationship of at least one calling node. Based on the operation context information corresponding to at least one data operation behavior, determine the data dependency of at least one data operation behavior, and based on the data dependency of at least one data operation behavior, determine the node dependency strength of at least one calling node.
[0059] In the embodiments of this specification, node dependency attributes include node dependency relationships and node dependency strength. Node dependency relationships are used to describe the operational dependencies between different data operation behaviors, to determine whether one data operation behavior depends on another, and to determine the corresponding dependency direction when a dependency relationship exists.
[0060] Specifically, based on the operation context information corresponding to at least one data operation behavior, the operation dependency relationship between different data operation behaviors can be determined, and based on the operation dependency relationship, the node dependency relationship between corresponding calling nodes can be determined.
[0061] For example, if the target object first obtains first data through a first data access behavior and then calls the target tool to perform a data analysis operation based on the first data, it can be determined that the second data operation behavior corresponding to the data analysis operation depends on the first data access behavior, and a node dependency relationship is established from the first calling node corresponding to the first data access behavior to the second calling node corresponding to the second data operation behavior.
[0062] For example, if the target object obtains the target analysis result through the tool invocation behavior and directly uses the target analysis result in subsequent operations, it can be determined that the result referencing behavior corresponding to the subsequent operations depends on the tool invocation behavior, and a node dependency relationship is established from the calling node corresponding to the tool invocation behavior to the calling node corresponding to the result referencing behavior.
[0063] In one implementation, node dependencies may include one or more of the following: data flow relationships, tool invocation relationships, result referencing relationships, and operation dependencies. Specifically, a data flow relationship represents data flowing from one data operation to another; a tool invocation relationship represents data or operation results generated by one data operation being used by a tool invoked by another data operation; a result referencing relationship represents operation results generated by one data operation being directly referenced by another data operation; and an operation dependency relationship represents the execution of a subsequent data operation depending on the execution result of a preceding data operation.
[0064] It should be noted that node dependencies can be directional. For example, when a second data operation uses data or operation results generated by a first data operation, a directed dependency can be established from the calling node corresponding to the first data operation to the calling node corresponding to the second data operation, to represent the direction of data or operation results from the previous operation to the next operation.
[0065] Furthermore, based on the operation context information corresponding to at least one data operation behavior, the data dependency between different data operation behaviors can be determined, and the node dependency strength of the corresponding calling node can be determined based on the data dependency.
[0066] The data dependency quantity is used to characterize the amount of data or operational result used in one data operation that is related to the data or operational result produced by another data operation. The data dependency quantity can be determined based on one or more of the following: data volume, number of data records, number of data fields, number of data items, data proportion, data lineage mapping ratio, or other information that can characterize the degree of correlation between data sources. This specification does not limit this.
[0067] For example, a first data operation produces a first operation result, and a second data operation uses the first operation result and other data as input. If, based on data lineage, a predetermined proportion of the input data for the second data operation originates from the first operation result, then the node dependency strength between the first and second data operations can be determined based on this predetermined proportion.
[0068] For example, if 30% of the input data for the second data operation comes from the output result of the first data operation, then the 30% data source ratio can be used as the basis for determining the node dependency strength, or the corresponding node dependency strength can be determined based on the data source ratio according to a preset mapping rule.
[0069] In another implementation, the data dependency can be determined based on the correspondence between data fields between different data operation behaviors. For example, when a second data operation behavior uses some of the data fields generated by a first data operation behavior, the corresponding data dependency can be determined based on the ratio between the number of data fields used and the number of data fields contained in the result of the first operation, and the node dependency strength can be further determined.
[0070] In one implementation, node dependency strength can be represented numerically. The higher the value of the node dependency strength, the greater the dependence of the current data operation on the data or operation result generated by the corresponding preceding data operation. Node dependency strength can also be represented in the form of levels, intervals, or other forms, which are not limited in this specification.
[0071] After determining the node dependencies and node dependency strengths, the node dependencies and node dependency strengths can be used as the node dependency attributes of the corresponding calling nodes, and the target call chain can be constructed based on the node attributes and node dependency attributes of each calling node.
[0072] For example, during the execution of the target task, the target object sequentially executes a first data access behavior, a second data access behavior, a tool invocation behavior, and a result referencing behavior. Specifically, the first and second data access behaviors acquire first and second data, respectively; the tool invocation behavior uses both the first and second data as input and generates the target analysis result; and the result referencing behavior references the target analysis result in subsequent operations.
[0073] For the aforementioned data manipulation process, corresponding first, second, third, and fourth calling nodes can be generated respectively, and node dependencies can be established such that the first calling node points to the third calling node, the second calling node points to the third calling node, and the third calling node points to the fourth calling node. The node dependency strength between the first and second calling nodes and the third calling node can be determined based on the amount of data dependency in the input data corresponding to the third calling node that originates from the first and second calling nodes, respectively. The node dependency strength between the third and fourth calling nodes can be determined based on the amount of data dependency in the operation results generated by the third calling node referenced by the fourth calling node.
[0074] Therefore, the actual data operation behavior that occurs during the execution of the target task can be represented by the call node, the operation characteristics of each data operation behavior can be described by the node attributes, and the data source relationship and dependence degree between different data operation behaviors can be described by the node dependency relationship and the node dependency strength, thus forming a target call chain that can reflect the continuous data operation behavior and its correlation degree during the execution of the target task.
[0075] By applying the embodiments of this specification, the operational dependencies between data operation behaviors are determined based on the operational context information corresponding to the data operation behaviors. Furthermore, the node dependencies between calling nodes are determined based on these operational dependencies. Additionally, the data dependency quantity of the data operation behaviors is determined based on the operational context information, and the node dependency strength between calling nodes is determined based on the data dependency quantity. This allows for a simultaneous description of the correlation between data operation behaviors occurring consecutively within the lifecycle of the target task from two dimensions: the dependency type and the degree of data dependency. This enables a more accurate characterization of the correlation between different data operation behaviors and their degree of correlation, thereby reducing information gaps caused by judging solely based on a single data operation behavior or solely based on the existence of a dependency relationship. This provides a more accurate basis for subsequently determining data usage risks based on the correlation between consecutive data operation behaviors and improves the accuracy of data access control for the target task.
[0076] In one optional embodiment of this specification, the operation context information includes the operation time of the corresponding data operation behavior, and the node attributes include node activity. Based on the operation context information corresponding to at least one data operation behavior, determine the node attributes corresponding to at least one calling node, including: Based on the time difference between the operation occurrence time corresponding to at least one data operation behavior and the target access time corresponding to the target data access request, the node activity of at least one calling node is determined, wherein the node activity decreases as the time difference increases.
[0077] The operation occurrence time characterizes the actual time when the corresponding data operation behavior occurs. Node activity characterizes the degree to which the corresponding data operation behavior affects the timeliness of the data usage process of the current target data access request. For the same target task, the closer the data operation behavior is to the occurrence time of the current target data access request, the higher its corresponding node activity can be; the earlier the data operation behavior occurs, the lower its corresponding node activity can be.
[0078] Specifically, the occurrence time of at least one data operation behavior can be obtained, and the time difference between each operation occurrence time and the target access time corresponding to the target data access request can be determined. For any data operation behavior, the node activity of the calling node corresponding to the data operation behavior can be determined based on the time difference between the operation occurrence time and the target access time, wherein the node activity decreases as the time difference increases.
[0079] For example, the target task performs a first data access action at a first moment, a second data access action at a second moment, and the current target data access request occurs at a third moment, with the first moment preceding the second moment. Since the time difference between the second data access action and the current target data access request is smaller than the time difference between the first data access action and the current target data access request, it can be determined that the calling node corresponding to the second data access action has a higher node activity than the calling node corresponding to the first data access action.
[0080] In one implementation, node activity can be determined according to a preset time decay rule. For example, a time decay period can be set, and as the time difference between the data operation and the target access time gradually increases, the node activity of the corresponding calling node is reduced according to the time decay period. The time decay rule can be a preset decay function, a segmented decay rule, or other rules that can reduce node activity as the time difference increases; this specification does not limit this to any particular rule.
[0081] In one implementation, node activity can be calculated using the following formula (1).
[0082] Formula (1) in, For nodes activity level The current time (the target access time corresponding to the target data access request). For nodes The time when the corresponding data manipulation action occurred. The time decay coefficient ( ), It is a natural constant.
[0083] Furthermore, when determining the target call chain, the node activity, node attributes, and node dependency attributes corresponding to each call node can be stored together in the target call chain. This allows the target call chain to simultaneously record the occurrence of each data operation behavior during the execution of the target task, the dependencies between behaviors, and the timeliness of different historical behaviors relative to the current target data access request.
[0084] In one implementation, the target task can be executed continuously for a considerable period, generating multiple data manipulation actions during execution. For earlier data manipulation actions, the activity of the corresponding node can be reduced as the time difference between the earlier data manipulation action and the current target data access request increases, without needing to remove the calling node corresponding to the data manipulation action from the target call chain. Therefore, while preserving the complete data manipulation history of the target task, the impact of earlier data manipulation actions on the current target data access request can be gradually reduced.
[0085] By applying the embodiments of this specification, the node activity of the calling node is determined by the time difference between the occurrence time of the data operation behavior and the target access time corresponding to the target data access request. The node activity decreases as the time difference increases, so that data operation behaviors occurring at different times within the life cycle of the target task have different degrees of influence on the current data access request. This allows for the retention of historical data operation behaviors while giving higher influence weight to data operation behaviors that are closer to the current data access request time. The influence of earlier data operation behaviors on the current data usage risk assessment gradually decreases, which is more in line with the characteristic of continuous data usage risk changing over time. This improves the accuracy of data usage risk assessment for the current data access request and the accuracy of data access control based on the risk assessment.
[0086] In one optional implementation of this specification, the target call chain is determined based on at least one calling node, at least one node attribute corresponding to the calling node, and at least one node dependency attribute of the calling node, including: The initial call chain is determined based on at least one call node, the node attributes corresponding to at least one call node, and the node dependency attributes of at least one call node; If at least one of the calling nodes has a first calling node whose node activity is lower than the activity threshold, then the first calling node in the initial calling chain is pruned to obtain the target calling chain. The first calling node can be any one of the at least one calling nodes.
[0087] In the embodiments of this specification, when determining the target call chain, the initial call chain can be determined first based on at least one call node, the node attributes corresponding to at least one call node, and the node dependency attributes of at least one call node.
[0088] The initial call chain can include multiple call nodes determined during the execution of the target task, as well as the relationships established between different call nodes based on node dependency attributes. The initial call chain can be used to fully or relatively fully represent the data operation behaviors that have occurred during the execution of the target task and the relationships between different data operation behaviors.
[0089] Furthermore, the node activity level corresponding to at least one calling node can be obtained, and the node activity level corresponding to each calling node can be compared with a preset activity threshold. The activity threshold can be determined based on the data usage characteristics of the target task, risk assessment requirements, or a preset time range; this specification does not limit this.
[0090] If at least one of the calling nodes has a first calling node whose activity level is below the activity threshold, then the first calling node in the initial calling chain can be pruned to obtain the target calling chain. The first calling node can be any node among the at least one calling node.
[0091] Specifically, pruning may include removing the first call node from the call chain used for current data risk analysis, and severing the association between the first call node and other call nodes. In one implementation, the connection relationships of other call nodes may also be adjusted based on the pruning of the first call node to maintain the analyzability of data and operational dependencies among the remaining call nodes.
[0092] For example, during the execution of a target task, a first call node, a second call node, and a third call node are generated sequentially. The data operation corresponding to the first call node occurred relatively recently, far removed from the current target data access request, resulting in the first call node's activity level being below a preset activity threshold. The second and third call nodes, however, have activity levels above the threshold. In this case, the first call node in the initial call chain can be pruned, while the second and third call nodes and their corresponding relationships are retained in the target call chain. This ensures that the target call chain primarily reflects recent data operation behaviors with a high time-sensitivity to the current target data access request.
[0093] In one implementation, the node activity of the initial call chain can be checked at a preset period, and pruning can be performed when a call node with an activity level below an activity threshold is detected. Alternatively, when a new target data access request is received, the node activity of each call node can be re-determined based on the target access time corresponding to the target data access request, and pruning can be performed according to the re-determined node activity. This specification does not limit this approach.
[0094] It should be noted that pruning the first calling node refers to removing it from the target calling chain used for current data risk assessment, and is not limited to deleting the historical data operation records corresponding to the first calling node. Optionally, the data operation behavior corresponding to the first calling node, as well as its node attributes and node dependency attributes, can still be saved for subsequent auditing, historical querying, or rebuilding of the calling chain.
[0095] By applying the embodiments of this specification, and comparing the node activity of the calling node with an activity threshold, calling nodes in the initial calling chain whose node activity is lower than the activity threshold are pruned. This reduces the interference of earlier data operation behaviors that have a low impact on the timeliness of the current target data access request on the current calling chain. It allows the target calling chain to focus more on continuous data operation behaviors in the lifecycle of the target task that have a high timeliness correlation with the current data access request. As a result, when judging data usage risk based on the target calling chain, it can better highlight recent data operation behaviors and their correlations, reduce the impact of low-timeliness historical behaviors on the current risk judgment, improve the accuracy of the current data usage risk judgment, and thus improve the accuracy of data access control based on the risk judgment.
[0096] Step 104: Based on the target call chain, determine at least one risk assessment parameter corresponding to the target data access request.
[0097] In one optional implementation of this specification, at least one risk assessment parameter includes at least one of comprehensive data sensitivity, data aggregation breadth, data resource association depth, and task path deviation. Based on the target call chain, determine at least one risk assessment parameter for the target data access request, including at least one of the following: Based on the data sensitivity information of the data resources corresponding to each calling node in the target call chain, determine the comprehensive data corresponding to the target data access request; Based on the number of data sources and the cumulative amount of data involved in the target call chain, the data aggregation breadth of the target data access request is determined. The data aggregation breadth is positively correlated with the number of data sources and the amount of data. The data association depth corresponding to the target data access request is determined based on the node association depth of at least one node pair in the target call chain. Based on the node dependencies in the target call chain, the actual task access path of the target task is determined. Based on the path difference between the actual task access path and the baseline task access path, the task path deviation corresponding to the target data access request is determined.
[0098] The overall data sensitivity is used to characterize the overall data sensitivity of the data resources involved in the execution of the target task. Data sensitivity can be used to characterize the sensitivity of the information contained in data resources to the data subject, business activities, or data security. For example, data resources can be divided into different data sensitivity levels according to preset data classification and grading rules, and the overall data sensitivity of the target task can be determined based on the data sensitivity levels corresponding to multiple data resources involved in the target call chain.
[0099] Data aggregation breadth characterizes the range and scale of data sources involved in the execution of a target task, reflecting the extent to which a target object acquires or uses data from multiple data sources within the same task. Data aggregation breadth is positively correlated with the number of data sources involved in the target call chain and the cumulative amount of data. The more data sources involved in the target call chain, or the larger the amount of data acquired, processed, or used from data sources, the greater the corresponding data aggregation breadth.
[0100] Data resource association depth characterizes the degree of association that can be established between different data resources in the target call chain through data operation behaviors. It reflects the degree to which multiple data resources are used in conjunction to form data associations or combined analyses. Data resource association depth can be determined based on the node association depth between calling nodes in the target call chain. When different data resources form multi-layered associations through multiple calling nodes with data dependencies or operation dependencies, the corresponding data resource association depth can be determined to be high.
[0101] Task path deviation is used to characterize the degree of deviation between the data access path formed during the actual execution of the target task and the predetermined task execution path.
[0102] Specifically, it is possible to obtain the data sensitivity information of the data resources involved in each calling node in the target call chain, and determine the comprehensive data sensitivity of the target data access request based on the data sensitivity information.
[0103] Data sensitivity information is used to indicate the degree of data sensitivity of the corresponding data resource. Data sensitivity information can be determined by the data classification and grading results corresponding to the data resource, or by one or more of the following: resource attributes, data content characteristics, data usage restrictions, or preset sensitivity identifiers.
[0104] For example, data resources can be pre-divided into multiple sensitivity levels, and corresponding sensitivity values can be set for different sensitivity levels. For multiple data resources involved in the target call chain, the sensitivity value corresponding to each data resource can be obtained separately, and the overall data sensitivity can be determined based on each sensitivity value.
[0105] In one alternative implementation, the overall data sensitivity can be determined based on the sensitivity value of each data resource and the node activity of the corresponding calling node, so that the data resources involved in recent data operation behaviors have a greater impact on the current overall data sensitivity than the data resources involved in earlier data operation behaviors.
[0106] In another alternative implementation, the corresponding data sensitivity can be weighted according to the degree of data dependency of different data resources in the target call chain to obtain a comprehensive data sensitivity. This specification does not limit this.
[0107] Specifically, the breadth of data aggregation corresponding to the target data access request can be determined based on the number of data sources and the cumulative amount of data involved in the target call chain.
[0108] The number of data sources indicates the number of data sources accessed, invoked, referenced, or processed in the target call chain. Data sources can correspond to databases, data tables, data files, data interfaces, data services, or other data resources that can provide data. For multiple data operations belonging to the same data source, they can be merged and counted based on a preset data source identifier to avoid duplicate counting of the same data source.
[0109] The cumulative data volume indicates the total amount of data involved in the target call chain during the execution of the target task. The cumulative data volume can be determined based on the number of data records, data bytes, data items, data return volume, or other information that can characterize the data scale for each data operation.
[0110] When determining the breadth of data aggregation, both the number of data sources and the cumulative data volume can be considered. For example, if at least one of the number of data sources and the cumulative data volume increases, the breadth of data aggregation can be increased accordingly; when the target task involves only a few data sources and the cumulative data volume is small, a relatively low breadth of data aggregation can be determined.
[0111] In one optional implementation, the number of data sources and the cumulative data volume can be normalized or segmented, and then the data aggregation breadth can be determined based on the processed results to reduce the impact of different units on the determination of the data aggregation breadth. In another optional implementation, the number of data sources or the cumulative data volume can be weighted according to the data sensitivity corresponding to the data sources to obtain the data aggregation breadth; this specification does not limit this approach.
[0112] Specifically, the data resource association depth corresponding to the target data access request can be determined based on the node association depth between at least one pair of nodes in the target call chain.
[0113] In this context, a node pair represents two call nodes in the target call chain that have a dependency or association relationship. The node association depth characterizes the degree of association between the data resources involved in the two call nodes corresponding to the node pair. The degree of association can be determined based on the similarity, relevance, or shared characteristics between the node attributes corresponding to the two call nodes. For example, the degree of node association between the two call nodes can be determined based on the data attributes, data fields, identification information, data tags, or other attribute information that can characterize the association relationship between the data resources corresponding to the two call nodes, and the corresponding node association depth can be determined based on the degree of node association.
[0114] For example, a first calling node corresponds to a first data access behavior, and a second calling node corresponds to a second data access behavior. The first data access behavior accesses a first data resource, and the second data access behavior accesses a second data resource. If the first data resource and the second data resource have the same user identifier field, or have multiple identical or related data attributes, then it can be determined that the first calling node and the second calling node have a high degree of node association, and a high node association depth can be determined accordingly. If the first data resource and the second data resource do not have obvious common attributes or related attributes, then it can be determined that the corresponding node association degree is low, and a low node association depth can be determined accordingly.
[0115] Specifically, at least one node pair can be identified from the target call chain, and the data resource association depth corresponding to the target data access request can be determined based on the node association depth corresponding to each node pair.
[0116] In this context, a node pair can correspond to two calling nodes in the target call chain, and the data resources involved in the two calling nodes have certain data attribute associations. For the data resources involved in multiple calling nodes in the target call chain, the node association depths corresponding to different data resources can be determined separately, and the data resource association depths can be obtained by comprehensively processing the corresponding node association depths based on multiple nodes.
[0117] For example, the target call chain includes a first data access node, a second data access node, and a third data access node, corresponding to the first data resource, the second data resource, and the third data resource, respectively. If the first data resource and the second data resource have a high degree of attribute association, and the second data resource and the third data resource also have a high degree of attribute association, then the high association depth of the data resources involved in the target task can be determined based on the node association depth corresponding to the first data resource and the second data resource, and the node association depth corresponding to the second data resource and the third data resource.
[0118] In one optional implementation, the node association depth can be determined based on one or more of the following: the number of common attributes between data resources corresponding to node pairs, the degree of attribute similarity, the number of associated fields, the degree of data tag overlap, or a preset association level. Furthermore, the data resource association depth can be determined based on the node association depth of multiple node pairs.
[0119] In another alternative implementation, the data resource association depth can also be determined by combining the node dependency strength. For example, when there is a deep node association relationship and a high node dependency strength between two data resources, the corresponding data resource association depth can be determined to be high; when there is only a weak or indirect dependency relationship between two data resources, the corresponding data resource association depth can be determined to be low.
[0120] Specifically, the actual task access path of the target task can be determined based on the node dependencies in the target call chain, and the task path deviation corresponding to the target data access request can be determined based on the path difference between the actual task access path and the baseline task access path.
[0121] The actual task access path characterizes the actual data operation trajectory formed by the node dependencies between calling nodes during the actual execution of the target task. The actual task access path may include one or more of the following: data resources actually involved in the target task, data operation behaviors, tool calls, and the execution or dependency order between different data operation behaviors.
[0122] The baseline task access path is a predefined task execution trajectory used to indicate the target task. The baseline task access path can be determined based on one or more of the following: the task type corresponding to the target task, preset task flow, historical normal execution records, data usage strategy, task template, or administrator configuration. The baseline task access path may include one or more of the following: allowed data resources, allowed data operation behaviors, expected data operation sequence, and expected dependencies between data resources.
[0123] In one alternative implementation, the actual task access path and the baseline task access path can be represented as path structures consisting of data resource nodes, operation nodes, and dependencies between nodes, respectively. The path differences can be determined by comparing one or more of the following: differences in nodes, differences in node order, differences in node dependencies, or differences in path length between the two path structures.
[0124] In another alternative implementation, the actual task access path and the baseline task access path can be encoded as path features, and the path difference can be determined based on the similarity between the two path features. For example, the task path deviation can be determined based on path similarity, wherein the more similar the actual task access path is to the baseline task access path, the lower the task path deviation; the greater the difference between the actual task access path and the baseline task access path, the higher the task path deviation.
[0125] The above methods allow for the determination of risk assessment parameters corresponding to target data access requests from different dimensions. Specifically, comprehensive data sensitivity reflects the sensitivity of the data involved in the target task; data aggregation breadth reflects the range and scale of data sources involved in the target task; data resource association depth reflects the degree of association formed between different data resources through continuous data operation behaviors; and task path deviation reflects the degree of deviation between the actual data operation path of the target task and the expected task execution trajectory. These risk assessment parameters can be determined individually or in combination; this specification does not limit this. For example, risk assessment information corresponding to the target data access request can be obtained by comprehensively processing based on the preset weights of at least one risk assessment parameter.
[0126] By applying the embodiments of this specification, and determining at least one risk assessment parameter among comprehensive data sensitivity, data aggregation breadth, data resource association depth, and task path deviation based on the target call chain, the data usage status formed by continuous data operation behaviors within the target task lifecycle can be characterized from different dimensions such as data sensitivity, data aggregation scope, the degree of association between data resources, and the degree of deviation between the actual task execution path and the expected task execution trajectory. This allows data risk assessment to no longer rely solely on a single data access behavior, but to comprehensively consider the data sensitivity, data aggregation, data association, and changes in the task execution path involved in continuous data operation behaviors. This improves the accuracy of identifying and judging data usage risks during the execution of the target task and provides a more accurate basis for implementing corresponding data access control on target data access requests based on data usage risks.
[0127] In one optional implementation of this specification, the comprehensive data sensitivity of the target data access request is determined based on the data sensitivity information of the data resources corresponding to each calling node in the target call chain, including: Based on the data sensitivity information of the data resources corresponding to the target call node in the target call chain, and the node activity of the target call node, the node data sensitivity of the target call node is determined. Based on the node data sensitivity of at least one call node, the comprehensive data sensitivity corresponding to the target data access request is determined. The target call node is any one of the call nodes.
[0128] In the embodiments of this specification, when determining the comprehensive data sensitivity corresponding to the target data access request, the node activity of each calling node in the target call chain can be further considered so that the comprehensive data sensitivity can reflect the actual impact of the data sensitivity corresponding to different data operation behaviors on the current target data access request.
[0129] Specifically, for any calling node in the target calling chain, data sensitivity information of the data resources corresponding to the calling node can be obtained, as well as node activity. The data sensitivity information characterizes the data sensitivity of the data resources corresponding to the calling node, while node activity characterizes the timeliness or impact of the data operation behavior of the calling node relative to the current target data access request.
[0130] Based on this, the node data sensitivity can be determined by considering the data sensitivity information and node activity of the calling node. Node data sensitivity characterizes the effective sensitivity of the data resources corresponding to the calling node during the execution of the task corresponding to the current target data access request.
[0131] For example, for two calling nodes with the same data sensitivity, if the data operation of the first calling node occurred closer to the time of the current target data access request, then the node corresponding to the first calling node has higher activity, and it can be determined that the node corresponding to the first calling node has higher data sensitivity. Conversely, if the data operation of the second calling node occurred earlier, resulting in lower activity, then it can be determined that the node corresponding to the second calling node has relatively lower data sensitivity. Therefore, newer data operations can have a greater impact on the current overall data sensitivity, while earlier data operations have a relatively weaker impact on the current overall data sensitivity.
[0132] Furthermore, after determining the node data sensitivity of each calling node in the target call chain, the overall data sensitivity corresponding to the target data access request can be determined based on the node data sensitivity of at least one calling node. For example, the node data sensitivity of each calling node can be weighted, averaged, statistically analyzed in stages, or processed according to preset sensitivity rules to obtain the overall data sensitivity.
[0133] In a specific example, the target call chain includes a first call node, a second call node, and a third call node. The first call node corresponds to ordinary business data and has low data sensitivity; the second call node corresponds to user identity data and has high data sensitivity; and the third call node corresponds to transaction data and also has high data sensitivity. If the data operations corresponding to the second and third call nodes occur within a timeframe close to the current target data access request, then the second and third call nodes have high node activity, and therefore their corresponding node data sensitivity is high. Although the first call node corresponds to data resources with lower sensitivity, its high node activity also allows it to participate in determining the overall data sensitivity. By comprehensively processing the node data sensitivities corresponding to each call node, the overall data sensitivity corresponding to the current target data access request can be obtained.
[0134] In this way, the overall data sensitivity is not determined solely by the data sensitivity information of the data resources involved in the target call chain, but is further combined with the node activity of different data operation behaviors to dynamically adjust the data sensitivity of different call nodes, so that the overall data sensitivity can reflect the effective data sensitivity involved in the target task at the current execution stage.
[0135] Optionally, data sensitivity information can be represented using at least one of sensitivity levels, sensitivity scores, or sensitivity ranges. For example, data resources can be pre-divided into multiple sensitivity levels according to their sensitivity, and corresponding sensitivity values can be configured for different sensitivity levels. When determining the sensitivity of node data, the sensitivity value corresponding to the data sensitivity information can be correlated with the node activity level to obtain the node data sensitivity.
[0136] Optionally, data sensitivity information can be weighted based on node activity. For example, when data sensitivity information is represented by a sensitivity score, a corresponding activity weight can be determined based on node activity, and the node's data sensitivity can be determined based on the sensitivity score and the activity weight. The activity weight can increase as node activity increases, thus allowing more active calling nodes to have a greater impact on the overall data sensitivity.
[0137] Optionally, the data sensitivity level can be adjusted based on node activity. For example, for calling nodes with high activity and corresponding high-sensitivity data resources, their node data sensitivity can be set to high sensitivity; for calling nodes with low activity, the impact of their data sensitivity in the current overall data sensitivity can be reduced. It should be noted that reducing the current impact of calling nodes with low activity does not mean deleting the historical data operation records corresponding to the calling nodes.
[0138] Optionally, when determining the overall data sensitivity based on the node data sensitivity corresponding to at least one calling node, the data sensitivity of multiple nodes can be processed according to preset integration rules. For example, the maximum value among multiple node data sensitivities can be selected as the overall data sensitivity; a weighted sum or weighted average can be performed based on the data sensitivity of each node; or statistics can be compiled based on the sensitivity levels of multiple node data sensitivities, and the overall data sensitivity can be determined based on the statistical results. The specific integration rules adopted can be configured according to the data space security policy, data resource management rules, or data usage rules corresponding to the target task.
[0139] Optionally, when determining the overall data sensitivity, only the node data sensitivity of the calling node whose node activity reaches a preset activity level can be comprehensively processed. For example, an activity threshold can be preset. When the node activity of a calling node reaches the activity threshold, the node data sensitivity of that calling node is included in the process of determining the overall data sensitivity. When the node activity of a calling node is lower than the activity threshold, the impact of the node data sensitivity of that calling node can be reduced, or the node data sensitivity can be excluded from the current calculation process of the overall data sensitivity.
[0140] In one alternative implementation, the overall data sensitivity can be calculated based on the following formula (2).
[0141] Formula (2) in, Indicates overall data sensitivity. For a set of data access nodes, For nodes The corresponding node data sensitivity, This represents the node's activity level.
[0142] It should be noted that for different types of calling nodes, their sensitivity attributes can be determined by matching them with their data operation behavior. Specifically, for data access nodes, the preset sensitivity level quantification value of the data resource corresponding to the data access node can be determined as the node data sensitivity of the data access node; for tool calling nodes, the node data sensitivity of the tool calling node can be determined based on the sensitivity attributes corresponding to the input data nodes of the tool calling node, according to preset inheritance rules or aggregation rules; for result referencing nodes, the node data sensitivity of the result referencing node can be determined based on the sensitivity attributes corresponding to the results referenced by the result referencing node.
[0143] Optionally, comprehensive data sensitivity can be used as one of at least one risk assessment parameter to further determine the risk assessment information corresponding to the target data access request, and to determine the data exposure control parameter corresponding to the target data access request based on the risk assessment information. This data exposure control parameter is used to control the degree of data exposure. For example, when the comprehensive data sensitivity is high, a higher data usage risk can be identified, and the degree of data exposure corresponding to the target data access request can be reduced accordingly; when the comprehensive data sensitivity is low, a higher degree of data exposure can be maintained. The data exposure control parameter can be used to control the data granularity, data precision, or data integrity of the returned data.
[0144] By applying the embodiments of this specification, the data sensitivity of each calling node is determined based on the data sensitivity information of the data resources corresponding to each calling node in the target calling chain and the node activity of the calling nodes. Furthermore, the comprehensive data sensitivity of the target data access request is determined based on the data sensitivity of at least one calling node. This ensures that the determination of data sensitivity not only considers the sensitivity of the data resources involved in the target task itself, but also combines the activity of different data operation behaviors relative to the current data access request. This allows for a dynamic characterization of the data sensitivity corresponding to different calling nodes, enabling the comprehensive data sensitivity to more accurately reflect the actual data sensitivity impact during the execution of the current target task. It also reduces the inappropriate influence of historical data operations on the current data risk assessment and provides a more accurate basis for implementing corresponding data exposure control and data access control for the target data access request based on data risk.
[0145] In one optional implementation of this specification, the breadth of data aggregation can be calculated based on the following formula (3).
[0146] Formula (3) in, To aggregate the breadth of data, The number of data sources for different business domains. To accumulate the amount of data acquired, The base-10 logarithm is used to non-linearly amplify the risks arising from a surge in data volume.
[0147] In one optional implementation of this specification, determining the data association depth corresponding to the target data access request based on the node association depth of at least one pair of nodes in the target call chain includes: Based on the node attributes of the target node pair, determine the node similarity of the target node pair; The node association depth of the target node pair is determined based on the node similarity and node activity of the target node pair, wherein the target node pair is any one of at least one node pair.
[0148] After determining the node association depth of at least one node pair in the target call chain, the data association depth corresponding to the target data access request can be determined based on the node association depth of at least one node pair. For example, the node association depths of each node pair can be weighted, averaged, statistically analyzed in stages, or the maximum value can be selected to obtain the data association depth corresponding to the target data access request.
[0149] In this embodiment, data association depth is used to characterize the degree to which data resources involved in different calling nodes are used in association during the execution of the target task. A higher data association depth indicates that different data resources are more likely to be used in combination during the current task execution, thus creating a higher degree of data association risk. For example, when multiple different data resources have high attribute similarity, and the corresponding data operation behaviors have high activity and strong data dependencies, it can be determined that the data association depth corresponding to the target data access request is high.
[0150] In one alternative implementation, the data association depth corresponding to the target data access request is calculated based on the following formula (4).
[0151] Formula (4) in, Indicates the depth of data association. Represents a node and The strength of node dependency. and Representing nodes respectively and Corresponding node attribute set; Indicates the number of elements in the set; and These represent the intersection and union operations of sets, respectively. This represents the maximum similarity obtained when traversing any two distinct node pairs in the call chain. If A value close to 1 indicates an extremely high risk of cross-table association re-identification.
[0152] By applying the embodiments of this specification, node similarity is determined based on the node attributes corresponding to different calling nodes in the target call chain. Furthermore, node association depth is determined by combining node activity and node dependency strength. Then, the data association depth corresponding to the target data access request is determined based on the association depth of at least one node with its corresponding node. This allows the determination of data association degree to simultaneously consider the attribute association degree between different data resources, the current impact of related data operation behaviors, and the data dependency degree between calling nodes. This enables more accurate identification of data risks arising from the associated use of different data resources during the execution of the target task, avoiding judgment of data usage risk based solely on a single data resource or a single data access behavior. It also provides a more accurate basis for implementing corresponding data exposure control and data access control for target data access requests based on data association risks.
[0153] In one optional implementation of this specification, the task path deviation can be calculated based on the following formula (5).
[0154] Compare the current actual task access path with the preset baseline task access path. Convert the actual accessed node sequence into an access vector. The baseline task access path is converted into a vector. The task path conformity is calculated using cosine similarity, and then the task path deviation is calculated based on the task path conformity. Formula (5) in, Indicates the deviation of the task path. Represents the dot product (inner product) of vectors. It represents the magnitude of the vector. The closer to 0, the more aligned with the original mission objective; the closer to 1, the more serious the deviation.
[0155] Step 106: Based on at least one risk assessment parameter, determine the risk assessment information corresponding to the target data access request, and based on the risk assessment information, determine the target access result corresponding to the target data access request.
[0156] In one optional implementation of this specification, the risk assessment information is a risk assessment score; determining the risk assessment information corresponding to the target data access request based on at least one risk assessment parameter includes: Based on at least one risk assessment parameter and the influence weight corresponding to at least one risk assessment parameter, determine the risk assessment score corresponding to the target data access request.
[0157] The risk assessment information is a risk assessment score, which is used to quantitatively characterize the degree of data risk corresponding to the target data access request.
[0158] Specifically, at least one risk assessment parameter corresponding to the target data access request can be obtained. For example, the risk assessment parameter may include at least one of the risk characteristic values corresponding to comprehensive data sensitivity, data aggregation breadth, data resource association depth, and task path deviation.
[0159] For each risk assessment parameter, its corresponding influence weight can be determined, and a weighted calculation can be performed based on each risk assessment parameter and its corresponding influence weight to obtain the risk assessment score corresponding to the target data access request.
[0160] In one alternative implementation, the risk assessment score is calculated using the following formula (6).
[0161] A weighted linear combination method is used to conduct a comprehensive risk assessment of the target data access request, and a comprehensive risk score for the target data access request is calculated. (Value range normalized to) ): Formula (6) in, , , , The dynamic weight parameters for each dimension of features satisfy... , Indicates overall data sensitivity. Indicates the breadth of data aggregation. Indicates the depth of data association. Indicates the deviation of the task path. The sigmoid activation function is used to smoothly map the overall score to the (0,1) interval.
[0162] By applying the embodiments of this specification, a risk assessment score corresponding to a target data access request is determined based on at least one risk assessment parameter and its influence weight. This allows different risk characteristic values to be comprehensively quantified according to their respective impact on data risk, resulting in a risk assessment result that can characterize the overall data risk level of the target data access request. This provides a quantitative basis for subsequently determining the corresponding data access control strategy and data exposure level based on the risk assessment score, enabling data risk to act more directly on data access control.
[0163] In one optional implementation of this specification, the risk assessment information is a risk assessment score; based on the risk assessment information, determining the target access result corresponding to the target data access request includes: Based on the risk assessment score and at least one preset risk level threshold, determine the target risk level corresponding to the target data access request. Based on the target risk level, a target access control policy is determined from at least one preset access control policy, and based on the target access control policy, the target access result corresponding to the target data access request is determined. Different access control policies correspond to different data exposure granularities, and the data exposure granularity is used to represent the precision of the data returned to the target object.
[0164] After determining the risk assessment score, the target risk level corresponding to the target data access request can be determined based on the relationship between the risk assessment score and at least one preset risk level threshold. For example, a first risk level threshold, a second risk level threshold, and a third risk level threshold can be set, and the target risk level corresponding to the target data access request can be determined as safe, low risk, medium risk, or high risk according to the range in which the risk assessment score falls.
[0165] For example, when the risk assessment score is below the first risk level threshold, the target risk level can be determined as safe; when the risk assessment score reaches the first risk level threshold but is below the second risk level threshold, the target risk level can be determined as low risk; when the risk assessment score reaches the second risk level threshold but is below the third risk level threshold, the target risk level can be determined as medium risk; and when the risk assessment score reaches the third risk level threshold, the target risk level can be determined as high risk.
[0166] Furthermore, a target access control policy can be determined from at least one preset access control policy based on the target risk level. Different access control policies correspond to different data exposure granularities. The data exposure granularity is used to represent the precision of the data returned to the target object.
[0167] For example, for a security level, access control policies corresponding to higher data exposure granularity can be determined to enable the target object to obtain relatively complete or accurate data; for a low-risk level, access control policies corresponding to reduced data precision can be determined, such as appropriately generalizing, intervalizing, or reducing the precision of the data; for a medium-risk level, access control policies corresponding to lower data exposure granularity can be determined, such as anonymizing, masking, aggregating, or only returning statistical results for some data; for a high-risk level, access control policies that restrict data exposure can be determined, such as refusing to return target data or only returning preset non-sensitive data.
[0168] Furthermore, the target data access request can be processed based on the target access control policy to obtain the target access result. For example, when the target access control policy allows data to be returned at a first data exposure granularity, the target data can be processed according to the first data exposure granularity before being returned; when the target access control policy requires a reduction in data exposure granularity, the target data can be generalized, desensitized, masked, aggregated, or have its precision reduced before being returned; when the target access control policy prohibits data exposure, the target data can be left unreturned, and a corresponding access denial result can be generated.
[0169] Therefore, a correspondence can be established between risk assessment scores, risk levels, access control policies, and data exposure granularity, so that the data usage risk corresponding to the target data access request can be further transformed into specific data access control methods.
[0170] By applying the embodiments of this specification, a risk assessment score corresponding to a target data access request is determined based on at least one risk assessment parameter. The target risk level is then determined based on the relationship between the risk assessment score and a preset risk level threshold. Corresponding access control strategies are matched according to different target risk levels. This allows data usage risk to be further transformed from continuous risk assessment results into data access control methods with different data exposure granularities. This enables differentiated control of the accuracy of returned data based on the risk level corresponding to the target data access request, avoiding the use of a uniform data access method to handle data access requests with different risk levels. The data risk assessment results can directly affect data exposure levels and data access control decisions, improving the targeting and flexibility of data access control based on data risk.
[0171] In one optional embodiment of this specification, at least one access control policy includes a security state control policy with data exposure granularity decreasing sequentially, a low-risk state control policy, a medium-risk state control policy, and a high-risk state control policy. Based on the target risk level, a target access control policy is determined from at least one preset access control policy, and based on the target access control policy, the target access result corresponding to the target data access request is determined, including: If the target risk level is a security level, then the security state control policy is determined as the target access control policy, and based on the security state control policy, the target access result corresponding to the target data access request is determined. If the target risk level is low risk, then the low risk state control policy is determined as the target access control policy, and based on the low risk state control policy, the target access result corresponding to the target data access request is determined. If the target risk level is medium risk, then the medium risk status control policy is determined as the target access control policy, and based on the medium risk status control policy, the target access result corresponding to the target data access request is determined. If the target risk level is high risk, then the high risk state control policy is determined as the target access control policy, and based on the high risk state control policy, the target access result corresponding to the target data access request is determined.
[0172] In the embodiments described in this specification, multiple access control policies are pre-configured, each corresponding to a different data exposure granularity. These multiple access control policies include a security state control policy, a low-risk state control policy, a medium-risk state control policy, and a high-risk state control policy, with the data exposure granularity decreasing sequentially. The data exposure granularity represents the precision of the data returned to the target object; a higher granularity results in higher precision of the returned data.
[0173] After determining the target risk level corresponding to the target data access request, the target access control policy can be selected from multiple access control policies based on the target risk level, and the target data can be processed accordingly based on the target access control policy to obtain the target access result.
[0174] When the target risk level is set to a security level, the security state control policy is determined as the target access control policy, and the target data is processed according to the data exposure granularity corresponding to the security state control policy to obtain the target access result. For example, target data can be returned to the target object with a high degree of data accuracy.
[0175] When the target risk level is low, the low-risk state control strategy is determined as the target access control strategy, and the target data is processed according to the data exposure granularity corresponding to the low-risk state control strategy to obtain the target access result. For example, the target data can be appropriately generalized, have its precision reduced, or be processed into intervals before being returned.
[0176] When the target risk level is medium risk, the medium risk status control strategy is determined as the target access control strategy, and the target data is processed according to the data exposure granularity corresponding to the medium risk status control strategy to obtain the target access results. For example, the target data can be further anonymized, masked, aggregated, or only statistical results can be returned.
[0177] When the target risk level is high, the high-risk status control policy is determined as the target access control policy, and the target data is processed according to the data exposure granularity corresponding to the high-risk status control policy to obtain the target access result. For example, the return of target data can be restricted or prohibited, or only preset low-precision or non-sensitive data can be returned.
[0178] Therefore, based on the risk level corresponding to the target data access request, the data exposure granularity that matches the risk level can be selected, so that the higher the risk level, the lower the corresponding data exposure level.
[0179] Optionally, the data exposure granularity corresponding to the security state control policy, low-risk state control policy, medium-risk state control policy, and high-risk state control policy can be pre-configured and adjusted according to data security rules, data usage rules, or access requirements corresponding to the target task.
[0180] Optionally, different access control policies may correspond to one or more data processing methods. For example, data processing methods may include at least one of the following: returning raw data, data generalization, data range-based processing, data precision reduction, data masking, data anonymization, data aggregation, returning statistical results, and denying data access. Different access control policies may select different combinations of data processing methods to form different data exposure granularities.
[0181] Optionally, the high-risk status control strategy may not directly return the target data to the target object, but instead generate an access-restricted result; or, when preset security conditions are met, it may only return non-sensitive data that has undergone preset processing.
[0182] By applying the embodiments of this specification, a correspondence is established between security state control strategies, low-risk state control strategies, medium-risk state control strategies, and high-risk state control strategies and different data exposure granularities. Furthermore, a matching access control strategy is selected based on the risk level corresponding to the target data access request. This ensures that the data risk level formed by the comprehensive risk characteristic values corresponds to different levels of data exposure control. As the risk level increases, the data exposure granularity is progressively reduced, thereby avoiding the use of the same data exposure method for access requests with different data risk levels. This allows data risk to directly drive the adjustment of the data exposure level, providing data access control that matches the data risk for the target data access request.
[0183] In one optional embodiment of this specification, the security state control policy is used to instruct that the data accessed by the data access request be returned in its original state. Based on the security state control policy, the target access result corresponding to the target data access request is determined, including: Determine the target data requested in the target data access request, and use the target data as the target access result.
[0184] Security state control policies are used to instruct that data access requests should return the data in its original state. Security state control policies correspond to a high level of data exposure granularity, and are used to perform complete data access control on the target data when the target risk level corresponding to the target data access request is at a security level.
[0185] Specifically, when the target risk level corresponding to the target data access request is a security level, the security state control policy is determined as the target access control policy. Based on the security state control policy, the target data corresponding to the target data access request can be determined, and the target data can be directly returned to the target object as the target access result.
[0186] In one specific implementation, the target data can be the query results, data records, data fields, or other data content corresponding to the target data access request. After determining the target data, no data generalization, precision reduction, masking, desensitization, aggregation, or other processing that reduces the granularity of data exposure is performed on the target data. Instead, the data content and precision of the target data are maintained, and the target data is directly returned as the target access result.
[0187] For example, when a target object initiates a data query request for a target task, and the target risk level is determined to be a security level based on risk assessment parameters, the security status control policy can be determined as the target access control policy. The data access module retrieves the target data according to the target data access request and directly returns the complete query results to the target object.
[0188] By using the above methods, when the data risk corresponding to the target data access request is within the security level, the original data granularity and data precision of the target data can be maintained, enabling the target object to obtain complete data access results.
[0189] By applying the embodiments of this specification, a security status control strategy is adopted when the target risk level corresponding to the target data access request is a security level, and the target data is returned directly as the target access result in its original state. This enables the data risk assessment result to establish a correspondence with the data exposure granularity. When the data risk corresponding to the risk characteristic value is at a security level, data access control can be completed without additionally reducing the data precision or data integrity. While meeting the data risk control requirements, the original data granularity of the target data is maintained, thereby achieving data access control that matches the level of data risk.
[0190] In one optional implementation of this specification, the low-risk state control strategy is used to instruct the range mapping of the values corresponding to sensitive fields in the data requested by the data access request. Based on a low-risk state control strategy, the target access result corresponding to the target data access request is determined, including: Determine the target data requested in the target data access request, and determine the value corresponding to the target sensitive field from the target data; Based on the data exposure granularity control factor, the target interval width corresponding to the target sensitive field is determined. The data exposure granularity control factor is determined based on the risk assessment score and is used to control the data exposure granularity of the data returned to the target object based on the target data access request. Based on the target interval width, determine the range of values that the corresponding values of the target sensitive fields fall into; Replace the values corresponding to the target sensitive fields in the target data with the range representative value or range of the value interval to obtain the target access results.
[0191] Among these, the target sensitive fields can be determined based on at least one of the following: the field's preset sensitivity level, whether the field has a direct or indirect identifying function, whether the current target task requires the use of precise values, the risk level corresponding to the target data access request, and the output constraints in the data usage contract. For example, a numeric field that may improve data association, inference, or re-identification capabilities through precise values, but for which the current target task does not require the use of precise values, can be identified as a target sensitive field; for fields for which the task must use precise values to complete calculations, interval processing can be omitted.
[0192] Once the target sensitive fields are identified, their corresponding values can be obtained, and the target interval width can be determined based on the data exposure granularity control factor. The data exposure granularity control factor can be determined based on a risk assessment score and is used to control the granularity of data exposure returned to the target object based on the target data access request. The higher the risk level corresponding to the risk assessment score, the wider the target interval width can be determined, thereby reducing the accuracy of the returned values.
[0193] In one alternative implementation, the data exposure granularity control factor is calculated based on the following formula (7).
[0194] To adhere to the conservative principle of security defense (i.e., the higher the risk, the faster the restriction measures increase), the system employs a decay model based on risk scores. Calculate the data exposure granularity control factor : Formula (7) in, The calculated risk assessment score, ; The sensitivity adjustment coefficient pre-configured for the system ( (Recommended value: 2-4) is used to control the convergence speed of data exposure granularity as risk increases.
[0195] Furthermore, the numerical range into which the value corresponding to the target sensitive field falls can be determined based on the target range width, and the original value corresponding to the target sensitive field can be replaced with the range representative value or range range of that numerical range to obtain the target access result.
[0196] For example, if the target sensitive field in the target data is the revenue amount, and the original value corresponding to the target sensitive field is 12680, and the target interval width is determined to be 500 based on the current data exposure granularity control factor, then it can be determined that 12680 falls into the value interval of 12500 to 13000. The original value is then replaced with the interval range of "12500 to 13000" or with the interval representative value corresponding to the value interval, thereby reducing the accuracy of the original value while retaining certain numerical information of the target data.
[0197] Using the above method, only sensitive fields that meet preset conditions can be processed for range, without having to process all fields in the target data uniformly.
[0198] Optionally, a data exposure granularity control factor can be determined based on the risk assessment score, and a target interval width can be determined based on the data exposure granularity control factor. For example, a base interval width can be set, and the base interval width can be adjusted according to the data exposure granularity control factor so that the larger the data exposure granularity control factor, the larger the target interval width, and the lower the numerical precision of the returned data.
[0199] In one alternative implementation, the target interval width is calculated based on the following formula (8).
[0200] Interval width Data exposure granularity control factor Related, defined as: Formula (8) in, The basic compartment width. The smaller the size, the wider the binning interval, and the lower the data accuracy.
[0201] Optionally, the numerical range corresponding to the target sensitive field can be determined by rounding down, rounding up, or by using preset binning rules. For example, for the original value 12680 and the range width 500, the corresponding range can be determined to be 12500 to 13000 according to preset binning rules. The representative value of the range can be the range median, the lower bound of the range, the upper bound of the range, or other preset representative values.
[0202] By applying the embodiments of this specification, the target sensitive fields that need to be range-processed are determined based on the preset sensitivity level of the fields, the function of field identification, the task's requirement for precise numerical values, and the output constraints in the data usage contract. The data exposure granularity control factor is determined based on the risk assessment score. Then, the target range width is adjusted according to the data exposure granularity control factor, and the original value of the target sensitive field is converted into a range representative value or range range. This allows data risk to affect the data exposure precision of sensitive values. While meeting the target task's requirement for using inaccurate data, the precision of sensitive values is reduced, thereby reducing the data association and inference risks brought about by precise values, and achieving data access control that matches the level of risk.
[0203] In one optional embodiment of this specification, the medium-risk status control strategy is used to indicate the data type of the data requested by the data access request and the amount of data accumulated by the task corresponding to the data access request during task execution, and to determine the target access result corresponding to the data access request. Based on the medium-risk status control strategy, the target access result corresponding to the target data access request is determined, including: Determine the target data requested in the target data access request; The target access result is determined based on the data type of the target data and the amount of target data accumulated by the target task during task execution.
[0204] Specifically, after determining that the medium-risk status control strategy is the target access control strategy, the target data corresponding to the target data access request can be determined, and the data type of the target data and the amount of target data accumulated by the target task during the current task execution can be obtained.
[0205] The data type is used to characterize the data category to which the target data belongs, and can be determined based on the data content, data fields, data resource attributes, or preset data classification rules. For example, data types can include identity information data, contact information data, location data, transaction data, behavioral data, statistical data, etc.
[0206] The cumulative data volume is used to characterize the scale of data acquired by the target task during task execution. It can be determined cumulatively based on the amount of data acquired by the target task through data access, tool calls, or result referencing. Data volume can be represented by the number of data records, the number of data fields, the number of data bytes, or other quantitative methods that can represent the scale of data.
[0207] Furthermore, the target access result can be determined based on the data type of the target data and the cumulative amount of target data acquired by the target task during task execution. For example, data volume control rules corresponding to different data types can be established in advance. When the cumulative amount of data acquired by the target task is small, the target data can be processed according to the first data exposure method corresponding to the corresponding data type; when the cumulative amount of data acquired reaches a preset data volume threshold, the target data can be processed according to the second data exposure method corresponding to the corresponding data type, wherein the data exposure level of the second data exposure method is lower than that of the first data exposure method.
[0208] For example, for identity information data, when the target task has accumulated a data volume threshold, it can return only a preset data summary, statistical results, or data after further processing, instead of directly returning the complete data record; for statistical data, when the accumulated data volume is low, it can return the corresponding statistical results, and when the accumulated data volume reaches a higher range, it can further limit the range or granularity of the returned data.
[0209] Furthermore, the data processing method for the target data can be determined based on the data type and the cumulative amount of data acquired, and the target data can be processed according to the determined data processing method to obtain the target access result. In this way, data of the same data type can correspond to different data access results when the cumulative amount of data acquired in the target task is different.
[0210] Optionally, one or more data volume thresholds can be set for different data types, and the target access result can be determined based on the data volume range in which the cumulative data volume acquired by the target task falls. For example, different data volume thresholds can be set for data types with different levels of sensitivity, so that more sensitive data types can trigger stricter data exposure control with a smaller cumulative data volume.
[0211] Optionally, the cumulative amount of data acquired can be counted only as the amount of data that has been successfully returned to the target object within the current target task's lifecycle, or it can be counted as the amount of data that the target task actually acquires or uses through data operation behaviors such as data access, tool calls, or result references.
[0212] Optionally, when the target data belongs to a preset data type and the cumulative amount of data acquired by the target task reaches the corresponding data volume threshold, the data exposure level of the target access results can be reduced. For example, the number of returned data records can be reduced, the returned fields can be limited, statistical results can be returned, or the target data can be aggregated.
[0213] Optionally, multiple data exposure levels can be determined based on the data type and the cumulative data volume, and corresponding data return methods can be configured for different data exposure levels. This allows for a higher cumulative data volume to be accessed at a lower data exposure granularity for the same data type.
[0214] By applying the embodiments of this specification, the target access result is determined based on the data type of the target data and the amount of data accumulated by the target task during task execution. This allows data access control to simultaneously consider the data category involved in the current request and the scale of data already accumulated by the target task. When the accumulated amount of data reaches different levels, corresponding data access results are adopted for different data types. This avoids determining the degree of data exposure based solely on a single data access request. It allows the data risk corresponding to the risk characteristic value to further affect the accumulated data acquisition status during task execution, and accordingly, differentiated control is applied to subsequent data access, reducing the data usage risk caused by the continuous acquisition of similar or related data.
[0215] In one optional implementation of this specification, the target access result is determined based on the data type of the target data and the amount of target data accumulated by the target task during task execution, including: If the data type indicates that the target data is aggregated statistical data, then the target noise corresponding to the target data is determined, and the target data is noise-added based on the target noise to obtain the target access result; If the data type indicates that the target data is detailed data and the target data volume is less than or equal to the set data volume threshold, then the target sensitive data in the target data is identified, and the target sensitive data is masked to obtain the target access result. If the data type indicates that the target data is detailed data and the target data volume is higher than the set data volume threshold, then the data distribution information of the target data is determined and the data distribution information is used as the target access result.
[0216] Specifically, the first step is to determine the data type of the target data. Data type can be used to characterize the data content and organization of the target data. For example, data type can include aggregated statistical data and detailed data. Aggregated statistical data can be data formed after statistical analysis, calculation, or summarization based on multiple data records, such as statistical results like quantity, proportion, average, sum, and quantiles. Detailed data can be data corresponding to specific data records, specific fields, or specific data objects.
[0217] In one implementation, when the data type indicates that the target data is aggregated statistical data, the target noise corresponding to the target data can be determined, and the target data can be noise-added based on the target noise to obtain the target access result.
[0218] Target noise can be determined based on the data exposure granularity control factor, the sensitivity parameter of the query function corresponding to the target data, and a preset noise distribution. For example, Laplace noise can be used as the target noise. For the statistical results corresponding to the target data, target noise can be generated according to preset noise parameters and then superimposed on the statistical results to obtain the noisy statistical results. The noise amplitude of the target noise can be adjusted according to the data exposure granularity control factor, so that the returned statistical results maintain the usability of basic statistical information while reducing the accuracy of the statistical results.
[0219] For example, if the target data is a statistical result of the number of users of a certain type, and the original statistical result is 10,000, after determining that the target noise is a random number following a preset Laplace distribution, the target noise can be superimposed on 10,000 to obtain a noisy statistical result. Thus, the target object can obtain data for statistical analysis, but it is difficult to directly perform data correlation based on the precise statistical results.
[0220] In another implementation, when the data type indicates that the target data is detailed data, the target data volume accumulated during the task execution can be further obtained, and the target data volume can be compared with the set data volume threshold.
[0221] The target data volume is used to characterize the scale of data that the target task has accumulated during the current task execution. The target data volume can be determined based on the number of data records returned to the target object, the number of data fields, the number of data bytes, the number of data items, or other indicators that can characterize the data scale. For example, the target data volume can be obtained by accumulating the number of data records acquired by the target task during the current and historical data access processes.
[0222] When the target data volume is lower than or equal to the set data volume threshold, the target sensitive data can be identified, and the target sensitive data can be masked to obtain the target access result.
[0223] Target sensitive data can be determined based on pre-defined sensitive field information. For example, fields such as name, contact information, exact address, and account identifier in detailed data can be identified as target sensitive data based on their corresponding sensitive attributes. After identifying the target sensitive data, irreversible masking processing can be performed on it, such as replacing some characters with preset mask characters, thereby reducing the recognizability of sensitive fields while retaining other non-sensitive detailed data.
[0224] Irreversible masks include one or more of the following: field masking, local preservation, generalization, tokenization, and irreversible mapping.
[0225] When the target data volume exceeds the set data volume threshold, the data distribution information of the target data can be determined, and the data distribution information can be used as the target access result.
[0226] In this context, data distribution information describes the overall data distribution characteristics of the target data, rather than returning specific details within the target data. For example, data distribution information may include data quantity, numerical range, frequency distribution, proportion distribution, mean, quantiles, or other statistical characteristics. Instead of directly returning the corresponding detailed data to the target object, data distribution information that characterizes the overall distribution of the target data is determined based on the target data, and this data distribution information is used as the target access result.
[0227] For example, when the amount of detailed data already acquired by the target task exceeds a set data volume threshold, instead of returning newly added detailed records, statistical characteristics such as the quantity, value range, and proportional distribution of the newly added detailed data can be returned. This preserves the target task's ability to analyze the overall data distribution while reducing the likelihood of acquiring a large amount of further detailed data.
[0228] Therefore, the medium-risk status control strategy can select different data processing methods according to the data type of the target data: for aggregated statistical data, noise is added to reduce the accuracy of the statistical results; for detailed data whose cumulative data acquisition amount does not exceed the threshold, masking is used to reduce the identifiability of sensitive fields; for detailed data whose cumulative data acquisition amount exceeds the threshold, detailed content is no longer returned directly, but data distribution information is returned instead.
[0229] In the embodiments described in this specification, the data volume threshold can be set according to the data type of the target data. Different data types may have different data volume thresholds. For example, a lower data volume threshold can be set for detailed data containing many sensitive fields, while a relatively higher data volume threshold can be set for detailed data with lower sensitivity.
[0230] In the embodiments described in this specification, the target data volume can be determined using a single data size indicator or by combining multiple data size indicators. For example, the target data volume can be determined based on the cumulative number of data records acquired, or it can be calculated comprehensively based on at least two of the cumulative number of data records acquired, the number of fields, and the number of data bytes.
[0231] In the embodiments described in this specification, data distribution information can be determined based on the data characteristics of the target data. For example, for numerical data, the numerical range, mean, quantiles, and the number of data points corresponding to different ranges can be determined; for categorical data, the number and proportion of data points corresponding to different categories can be determined; and for time-based data, the number and proportion of data points corresponding to different time intervals can be determined. By using different methods of expressing data distribution information, the overall statistical characteristics of the target data can be preserved without returning specific detailed data.
[0232] In one optional implementation of this specification, the noise amplitude of the target noise can be adjusted according to the data exposure granularity control factor. When the data exposure degree corresponding to the data exposure granularity control factor is low, the amplitude of the target noise can be increased; when the data exposure degree corresponding to the data exposure granularity control factor is high, the amplitude of the target noise can be decreased, thereby adjusting the return accuracy of the aggregated statistical data according to different data exposure requirements.
[0233] By applying the embodiments in this specification, noise addition, masking, and data distribution information return methods are adopted according to the data type of the target data. For detailed data, the control method is further determined by combining the cumulative data volume acquired by the target task during task execution with a set data volume threshold. This allows data access control to simultaneously consider the data content type and the cumulative data scale of the task: For aggregated statistical data, noise is added to reduce the accuracy of statistical results; for detailed data with a low cumulative acquisition volume, masking is used to reduce the identifiability of sensitive data; for detailed data with a cumulative acquisition volume exceeding the threshold, the return of detailed data is cut off and only data distribution information is provided, reducing the data association and inference risks caused by the continuous acquisition of a large amount of detailed data. Thus, differentiated data access control is implemented according to different data types and the cumulative data exposure scale, reducing the data risks generated during the continuous acquisition of data by the target task.
[0234] In one optional implementation of this specification, determining the target noise corresponding to the target data includes: Based on the global sensitivity of the query function corresponding to the target data and the data exposure granularity control factor, the noise parameters of the target noise are determined, and the target noise is determined based on the noise parameters. Among them, the data exposure granularity control factor is determined based on the risk assessment score and is used to control the data exposure granularity returned to the target object based on the target data access request. The data exposure granularity query function is a function that performs aggregate statistical processing on the original data corresponding to the target data access request to obtain the target data. The global sensitivity is used to indicate the degree of impact of a change in a piece of data in the original data on the query result. The noise parameter is positively correlated with the global sensitivity and negatively correlated with the data exposure granularity control factor.
[0235] In one optional implementation of this specification, determining the target noise corresponding to the target data includes: determining the noise parameters of the target noise based on the global sensitivity of the query function corresponding to the target data and the data exposure granularity control factor, and determining the target noise based on the noise parameters.
[0236] The query function is a function that performs aggregated statistical processing on the raw data corresponding to the target data access request to obtain the target data. For example, the query function can be a function used to calculate the total amount of data, average value, count, proportion, or other statistical results. The raw data may include the data records corresponding to the target data access request, and the target data may be the statistical results obtained by aggregating and statistically analyzing multiple raw data records.
[0237] Global sensitivity indicates the maximum impact a change in a single piece of data in the original dataset has on the query results of the corresponding query function. A higher global sensitivity means that an addition, deletion, or numerical change in a single piece of original data may cause a greater change in the statistical results corresponding to the target data.
[0238] For example, for a query function used to count the number of records, adding or deleting a single original data entry might cause a fixed change in the statistical result; for a query function used to calculate the sum of data, a change in the value of a single original data entry might cause a significant change in the statistical result. Therefore, the appropriate global sensitivity can be determined based on the calculation method corresponding to different query functions.
[0239] The data exposure granularity control factor is used to control the granularity of data exposure returned to the target object based on the target data access request. The data exposure granularity control factor can be determined based on the risk assessment score. For example, the corresponding data exposure granularity control factor can be determined based on the risk range in which the risk assessment score falls. A higher risk assessment score corresponds to a lower data exposure granularity control factor, thus reducing the likelihood of the target object obtaining highly accurate statistical results.
[0240] After determining the global sensitivity and the data exposure granularity control factor, the noise parameters of the target noise can be determined based on these two factors. The noise parameters are positively correlated with the global sensitivity and negatively correlated with the data exposure granularity control factor.
[0241] Specifically, with other conditions remaining constant, increasing the global sensitivity allows for a larger noise parameter, resulting in a greater amplitude of the generated target noise. This reduces the deterministic impact of changes in a single piece of original data on the final returned result. Conversely, decreasing the data exposure granularity control factor allows for a larger noise parameter, further reducing the accuracy of the returned target data.
[0242] In one specific implementation, a Laplace distribution can be used to generate target noise. First, the noise parameters of the Laplace noise are determined based on global sensitivity and data exposure granularity control factors. Then, random noise conforming to the corresponding Laplace distribution is generated based on the noise parameters, and the random noise is added to the target data to obtain the noisy target access result.
[0243] For example, the target data is the total transaction amount calculated based on user transaction details. The global sensitivity corresponding to the query function is used to characterize the maximum impact of changes in a single transaction on the total transaction amount. When the risk assessment score corresponding to the target task is high, the data exposure granularity control factor decreases accordingly, while the noise parameter determined based on the global sensitivity and the data exposure granularity control factor increases. Target noise is generated based on the increased noise parameter and added to the total transaction amount before being returned. Thus, the returned result can still be used to obtain the overall statistical characteristics of the transaction data, but the accuracy of the returned result is reduced.
[0244] In one implementation, the noise parameters can be calculated using the method indicated by the following formula (9). .
[0245] Formula (9) in, For global sensitivity of the query function, Expose granularity control factors for data.
[0246] Furthermore, when the target data corresponds to different query functions, the global sensitivity for each query function can be determined separately, and the target noise can be determined in conjunction with the corresponding data exposure granularity control factor. For example, for count queries, summation queries, and average queries, the global sensitivity can be determined based on the data change characteristics of the corresponding query function, and then the corresponding noise parameters can be determined based on their respective global sensitivity, so that different types of aggregated statistical results have noise processing effects that match their own data change sensitivity.
[0247] In one optional implementation of this specification, the noise parameter can be determined based on the ratio of global sensitivity to the data exposure granularity control factor. That is, when the data exposure granularity control factor remains constant, the higher the global sensitivity, the higher the noise parameter; conversely, when the global sensitivity remains constant, the lower the data exposure granularity control factor, the higher the noise parameter. The specific calculation method can be set according to actual data access control requirements.
[0248] In one optional implementation of this specification, the data exposure granularity control factor can be continuously adjusted based on the risk assessment score, or it can be divided into multiple preset value ranges according to the risk level. For example, when the risk assessment score is in a higher risk range, a smaller data exposure granularity control factor can be determined for the target data to increase the corresponding noise parameter; when the risk assessment score is in a lower risk range, a larger data exposure granularity control factor can be determined for the target data to reduce the corresponding noise parameter.
[0249] In one optional embodiment of this specification, the target noise is not limited to Laplace noise; other preset noise distributions can also be used to generate target noise based on the data type of the target data, the query function type, and the preset data access control strategy. Accordingly, the corresponding noise parameters can be determined based on the selected noise distribution.
[0250] By applying the embodiments of this specification, the noise parameters of the target noise are determined based on the global sensitivity of the query function and the data exposure granularity control factor. The noise parameters increase with the increase of global sensitivity and increase with the decrease of data exposure granularity control factor. This allows the data return accuracy of aggregated statistical data to simultaneously adapt to the degree of influence of a single piece of original data on the statistical results and the current data exposure control requirements. When the influence of a single piece of original data on the query results is large, the noise amplitude is increased. When the data exposure granularity control factor decreases, the noise amplitude is further increased. This reduces the possibility that the target object obtains or associates original data features based on the accurate aggregation results, reduces the data risk caused by the aggregated statistical results being used for further data association or inference, and realizes data access control that differentiates the control of the return accuracy of aggregated statistical data based on risk assessment results.
[0251] In one optional implementation of this specification, a high-risk state control strategy is used to indicate that the access result of a data access request is empty; Based on the high-risk state control strategy, the target access result corresponding to the target data access request is determined, including: The target access result corresponding to the target data access request is determined to be empty.
[0252] Based on the high-risk state control strategy, the data access operation corresponding to the target data access request can be directly rejected, and the target data will not be returned to the target object. In this case, it can be determined that the target access result corresponding to the target data access request is empty.
[0253] In one specific implementation, upon receiving a target data access request, a risk assessment score corresponding to the request can be determined first. If the risk assessment score does not reach a high-risk threshold, the target access result can be determined according to a data access control strategy matching the corresponding risk status; if the risk assessment score reaches a high-risk threshold, the data acquisition operation corresponding to the current data access request is terminated, and the target access result is determined to be empty.
[0254] For example, when a target object requests access to a data resource, a risk assessment score is determined based on the call chain corresponding to the target task. When the risk assessment score reaches a high-risk threshold, the system can stop executing the current data query, data read, or data acquisition operation, and will no longer return query results, data records, or data fields to the target object, returning an empty result as the target access result.
[0255] Optionally, if the target access result is determined to be empty, status information indicating that access was denied can also be returned to the target object, such as an access denial message, a preset error code, or other access control results. An access denial message indicates that the target data access request did not obtain a data access result, but does not contain the target data itself.
[0256] By applying the embodiments of this specification, when the risk assessment result corresponding to the target data access request reaches a high-risk condition, a high-risk state control strategy is adopted to directly terminate the data acquisition operation corresponding to the current data access request and determine the target access result as empty. This reduces the granularity of data exposure in high-risk situations to the point that no target data content is provided to the target object. Compared with simply masking, reducing precision, or digesting the data, this can further prevent the target object from acquiring the target data, thereby reducing the data risk caused by the continued exposure of data content during high-risk data access and realizing data access control that directly blocks data access requests based on risk assessment results.
[0257] In one optional embodiment of this specification, after determining that the target access result corresponding to the target data access request is empty, the method further includes: Revoke the target data access token corresponding to the target task to freeze subsequent tool calls for the target task.
[0258] Specifically, the target data access token is used to characterize the authorization status of a target task accessing target data resources or invoking data processing tools. When the risk assessment score corresponding to a target data access request reaches a high-risk condition, and the target access result is determined to be empty based on the high-risk status control strategy, the target data access token currently held by the target task can be further determined, and the target data access token can be marked as invalid or directly removed from the set of valid tokens.
[0259] After the target data access token is revoked, subsequent data access requests or tool call requests initiated by the target task can be verified based on the token's validity. When it is detected that the target data access token corresponding to the target task is invalid, the target task can be denied further calls to tools that require data access permissions, or the tool can be denied providing data access results to the target task, thereby freezing subsequent tool calls of the target task.
[0260] For example, if the risk assessment score corresponding to the target task reaches a high-risk threshold during execution, the system first determines the target access result of the current target data access request to be empty; then, it revokes the data access token corresponding to the target task. Afterwards, even if the target task continues to initiate new tool call requests, it will be unable to continue calling tools requiring data access permissions using the revoked data access token without obtaining a valid new data access authorization.
[0261] By applying the embodiments of this specification, after a high-risk data access request is blocked, the data access token corresponding to the target task is further revoked, and subsequent tool calls of the target task are frozen based on the validity of the token. This extends the scope of data access control from the current data access request to the subsequent data acquisition behavior of the target task, preventing the target task from continuing to acquire data through subsequent tool calls after the current access is denied. This reduces the data risk caused by the continuous acquisition of data by high-risk tasks and achieves continuous control over the subsequent data access behavior of the target task.
[0262] In one optional embodiment of this specification, after determining that the target access result corresponding to the target data access request is empty, the method further includes: The target task is suspended, and risk alarm information is generated based on the target call chain and risk assessment score. The risk alarm information is then pushed to the administrator terminal. In response to an access recovery command initiated by the administrator terminal, upon receiving a new target data access request from the target object, the target access result of the target data access request is determined based on the medium-risk status control policy.
[0263] Specifically, after a target data access request is blocked by a high-risk state control policy, the target task's state can be set to suspended. A target task in the suspended state will suspend further operations requiring data access permissions.
[0264] After the target task enters a suspended state, the target call chain and the risk assessment score corresponding to the target data access request can be obtained, and risk alarm information can be generated based on the target call chain and risk assessment score. The target call chain can be used to characterize the data operation behaviors that have occurred during the current task execution and their corresponding relationships, while the risk assessment score can be used to characterize the data risk level corresponding to the current target data access request. The risk alarm information can include the target task's task identifier, the target call chain, the risk assessment score, and relevant information that triggered the high-risk state.
[0265] Furthermore, risk alert information can be generated based on risk alert information and pushed to the administrator terminal so that the administrator can manually review the current data access behavior of the target task.
[0266] During manual review, administrators can determine whether data access to a target task can be restored based on the target call chain and risk assessment score in the risk alert information. If the administrator believes that the target task meets the conditions for resumption of execution, they can initiate an access restoration command through the administrator terminal.
[0267] In response to the access recovery command, the system can resume receiving target data access requests initiated by the target object and adjust the data access control status corresponding to the target task from suspended to medium-risk. At this point, the system does not directly restore the data access method of the target task before the high-risk state; instead, it determines the target access result corresponding to the target data access request based on the medium-risk state control strategy.
[0268] For example, upon receiving a renewed request for access to the target data, the access result can be determined based on the data type and the amount of data accumulated during task execution. When the target data is aggregated statistical data, noise can be added; when the target data is detailed data, the method of returning sensitive data masking or data distribution information can be determined based on the accumulated amount of data, thereby restoring data access to the target task with limited data exposure granularity.
[0269] Therefore, even if the administrator allows the target task to resume execution, the target task will not be directly restored to the data access state before the high-risk blocking, but will continue to execute in the restricted data access mode corresponding to the medium-risk state.
[0270] By applying the embodiments in this specification, after a high-risk data access request is blocked, the target task is suspended, and a risk alarm message is generated based on the target call chain and risk assessment score and pushed to the administrator terminal for manual review. This adds a manual review step based on actual data operation behavior and risk assessment results before the high-risk task resumes data access. After the administrator initiates an access recovery command, the target task is restored to a medium-risk state, and the target access result is determined based on the medium-risk state control strategy, rather than directly restoring the original data access method. This ensures that data access after manual approval is still subject to restricted data exposure control, thereby reducing the data risk caused by high-risk tasks continuously acquiring data without review and directly acquiring high-precision data after recovery. This achieves a closed loop of data access control, including blocking, manual review, and restricted recovery of high-risk data access.
[0271] In one optional embodiment of this specification, after determining the target access result corresponding to the target data access request based on the target access control policy, the method further includes: Based on at least one of the target data access request, the target call chain, at least one risk assessment parameter, and the target access result, determine the target access control information corresponding to the target task. Based on the target access control information corresponding to the target task and the historical access control information corresponding to the historical tasks processed by the target object, the object access control information of the target object is determined and stored.
[0272] The target access control information is used to record the data access control process and results corresponding to the target task. For example, the target access control information may include at least one of the following: the target task's task identifier, the target data access request, the target call chain, risk assessment parameters, the target access control policy, the target access result, and the corresponding data access control time.
[0273] Risk assessment parameters may include at least one of the following: comprehensive data sensitivity, breadth of data aggregation, depth of data resource association, and deviation from the task path. By recording these risk assessment parameters, the risk characteristics information corresponding to the target task during execution can be retained for subsequent analysis of the target task's data access control.
[0274] The target call chain can be used to record data manipulation behaviors that occur during the execution of a target task and the relationships between different data manipulation behaviors. The target access result can be used to record the data content or data exposure form actually provided to the target object after the target data access request is processed by the target access control policy. Therefore, the target access control information can simultaneously reflect the target task's data access requests, the data operation relationships during task execution, risk characteristics, and the final data access control result.
[0275] Furthermore, historical access control information corresponding to historical tasks processed by the target object can be obtained, and based on the target access control information corresponding to the target task and the historical access control information, the object access control information of the target object can be determined.
[0276] Among them, object access control information is used to record the data access control status of multiple tasks from the perspective of the target object. For example, access control information of multiple historical tasks can be summarized based on the risk assessment parameters, risk assessment scores, target access control policies, and target access results of multiple tasks corresponding to the target object to form object access control information for the target object.
[0277] In one specific implementation, access control information corresponding to multiple tasks can be associated according to the task execution time corresponding to the target object, and the risk characteristic values corresponding to different risk assessment parameters, the number of tasks corresponding to different risk levels, and the usage of different access control policies can be counted to form access control records corresponding to the target object.
[0278] Once the object access control information is determined, it can be stored. The stored object access control information can be used to query data access control status by target object dimension and serves as the historical data basis for subsequent control effectiveness evaluation.
[0279] By applying the embodiments of this specification, access control information corresponding to the target task is formed based on the target data access request, target call chain, risk assessment parameters, and target access result. Furthermore, access control information corresponding to historical tasks is combined to form access control information at the target object level. This allows data access behavior, risk characteristics, and control results in a single data access control process to be associated and recorded, forming a continuous object-level data access control record. This provides a basis for subsequently identifying the data access control status of the target object, reduces the data risk caused by controlling based solely on a single data access behavior, which fails to reflect historical data usage, and provides an information foundation for continuous data access control.
[0280] In one optional embodiment of this specification, after storing the object access control information, the method further includes: Retrieve stored object access control information at set intervals; Based on object access control information, control effectiveness evaluation indicators are determined. These indicators characterize the effectiveness of data access control implemented on data access requests initiated by the target object. Based on the control effectiveness evaluation indicators, adjust the impact weight of at least one risk assessment parameter and / or at least one risk level threshold.
[0281] The set period can be pre-configured according to data access control requirements; for example, stored object access control information can be retrieved daily, weekly, or monthly. Control effectiveness evaluation metrics are used to characterize the effectiveness of data access control execution for data access requests initiated against target objects.
[0282] In one specific implementation, data access control results corresponding to different risk assessment parameters, risk levels, and target access control policies can be statistically analyzed based on object access control information, and control effectiveness evaluation indicators can be determined based on the statistical results.
[0283] For example, the number of data access requests identified as high-risk, the number of blocked data access requests, the number of data access requests using different data exposure control strategies, and the number of data access requests that need to have their control status adjusted after subsequent manual review can be counted within a preset time period. Based on the above statistical results, control effectiveness evaluation indicators can be determined.
[0284] Furthermore, the impact of different risk assessment parameters on data access control results can be determined by establishing a correspondence between the risk assessment parameters recorded in the object access control information and the actual data access control results. For example, the risk level distribution and access control policy distribution corresponding to different risk assessment parameter value ranges can be analyzed to determine whether the current risk assessment parameter configuration can accurately distinguish different data access risks.
[0285] Once the control effectiveness evaluation indicators are determined, the impact weight of at least one risk assessment parameter and / or at least one risk level threshold can be adjusted based on the control effectiveness evaluation indicators.
[0286] The influence weights corresponding to the risk assessment parameters are used to determine the degree of influence of different risk characteristic values in the risk assessment score. When the control effectiveness evaluation index indicates that the control effect corresponding to a certain risk characteristic value is insufficient, the influence weight corresponding to that risk characteristic value can be adjusted; when the control effectiveness evaluation index indicates that the current risk level classification cannot effectively distinguish different data access risks, the threshold corresponding to the risk level can be adjusted.
[0287] For example, historical access control information may indicate that, in multiple tasks, data access requests with higher risk characteristic values corresponding to data resource association depth tend to lead to subsequent high-risk behaviors. However, if the current data resource association depth has a lower impact weight, the impact weight corresponding to that risk characteristic value can be increased, allowing the data resource association depth to have a greater impact on the subsequent risk assessment score.
[0288] For example, if historical access control information shows that a large number of data access requests are concentrated around a certain risk level threshold, and the access control results corresponding to that threshold differ significantly from the actual risk situation later, then the corresponding risk level threshold can be adjusted according to the control effectiveness evaluation indicators to make the data access control range corresponding to different risk levels more consistent with the historical data access situation.
[0289] In this way, the stored object access control information can be used periodically to evaluate the current risk assessment parameters, and the impact weights and / or risk level thresholds corresponding to the risk assessment parameters can be adjusted according to the evaluation results, so that the risk assessment and data access control of subsequent data access requests can be adjusted based on historical control results.
[0290] By applying the embodiments of this specification, object access control information is acquired according to a set period and control effectiveness evaluation indicators are determined. Then, the influence weights and / or risk level thresholds corresponding to risk assessment parameters are adjusted according to the control effectiveness evaluation indicators. This allows historical data access control results to have a reverse effect on subsequent risk assessments and data access control parameters. This avoids the influence weights of risk characteristic values and risk level classifications remaining fixed for a long time, which would not match the actual changes in data access risks. As a result, subsequent data access control can be adaptively adjusted according to historical control effects, improving the matching degree of access control for different data risks and reducing data risks caused by unreasonable configuration of risk assessment parameters.
[0291] In one optional implementation of this specification, after receiving a target data access request initiated by the target object for the target task, the task triggering context and the authorization decision context can be determined respectively, and the basic access qualification verification of the target data access request can be performed based on the authorization decision context.
[0292] The task triggering context describes the basis and business background for generating the target data access request. The task triggering context may include user input information, historical interaction content, upstream tool return results, model inference content, and other information used to trigger or form the target data access request. Because some information in the task triggering context may originate from external input, tool return content, or model-generated content, the task triggering context is not used as the basis for granting basic access permissions.
[0293] The authorization decision context is used to support the basic access qualification determination for target data access requests. The authorization decision context may include subject identity attributes, target resource attributes, preset access control policies, data usage contracts, authorization credentials, and other trusted authorization information maintained by the trusted data space. The information in the authorization decision context can be determined by the trusted data space based on the identity of the target object, the target resource corresponding to the target data access request, and the current authorization status, and an authorization context object corresponding to the target data access request can be generated.
[0294] In one specific implementation, it can be determined whether the target object has the basic access qualifications to access the target resource based on the authorization decision context. For example, the identity of the target object can be determined based on the subject identity attribute, and whether the target object has the permission to access the target resource can be determined based on at least one of the resource attributes, preset access control policies, data usage contracts, and authorization credentials.
[0295] If the authorization decision context determines that the target object does not have basic access qualifications, the target data access request will be rejected, and the target object will not be granted access to the target resource based on user input information, historical interaction content, tool return results, or model inference content in the task triggering context.
[0296] If the authorization decision context determines that the target object has basic access qualifications, the target data access request is allowed to enter the subsequent data access control process, and the target call chain is determined based on the data operation behavior and operation context information corresponding to the target data access request. Furthermore, risk assessment parameters can be determined based on the target call chain, risk assessment information can be determined based on the risk assessment parameters, and the target access control policy and target access result can be determined based on the risk assessment information.
[0297] In the above process, the task triggering context does not participate in the granting of basic access permissions, but can serve as auxiliary input for target call chain construction, behavior correlation analysis, task path deviation analysis, anomaly detection, and security auditing. For example, the background of the target data access request can be analyzed based on user input information, historical interaction content, tool return results, and model inference content in the task triggering context, and the relevant information can be correlated with the data operation behavior in the target call chain to help determine whether the actual data usage behavior meets the expectations of the target task.
[0298] The authorization decision context and the task triggering context are logically isolated. Even if the task triggering context contains instructions, tool return results, or model inference content requesting access to the target resource, it will not directly change the subject identity attributes, resource attributes, access control policies, data usage contracts, or authorization credentials in the authorization decision context. Only after the authorization decision context determines that the target object has basic access qualifications will the target data access request enter the subsequent data access control process based on the call chain and risk assessment.
[0299] The above approach can form a processing flow of "basic authorization verification - task behavior analysis - risk assessment - data access control". The authorization decision context is used to determine whether the target data access request has basic access qualifications, and the task triggering context is used to assist in analyzing the task background of the target data access request and subsequent data usage behavior. This avoids task triggering information from directly participating in the granting of basic permissions, while retaining the role of task triggering information in the supervision of data usage behavior.
[0300] See Figure 2 , Figure 2 A schematic diagram of an adaptive control system architecture for agent data access risk is shown according to one embodiment of this specification.
[0301] The request parsing and separation decision module parses data access requests, obtaining information such as the access subject, task objective, access object, and context information, and determines the information source (i.e., task triggering context) and authorization source (i.e., authorization decision context). The behavior association tracking and call chain construction module uses this information to model the associations of data access behavior, tool call behavior, information association behavior, intermediate result references, and cross-round associations, constructing a call chain to form the entire behavioral chain of the agent's task execution. The risk dynamic assessment module assesses the risk level based on the call chain, outputting a security level, low risk level, medium risk level, or high risk level. The risk-driven adaptive control module determines the access control policy based on the risk level, adjusting the data return granularity and data access permissions. After control, the data enters the trusted data space through controlled access, accesses the data source in the trusted data space in conjunction with the access control policy, and returns the access results to the agent. The audit link recording module records request records, behavior link records, risk assessment records, control decision records, and result and return records, and stores and manages the audit link. This allows for the determination of control effectiveness evaluation indicators based on the audit link, and optimization of various modules in the system based on these indicators.
[0302] In summary, this specification's embodiments address the issues of fragmented data access behavior, difficulty in identifying cross-step data association risks, and the limitations of traditional access control which can only perform static "allow / deny" judgments for single requests during continuous task execution. It constructs a dynamic data access control mechanism that addresses these problems.
[0303] The overall solution does not grant isolated authorization for a single data access request. Instead, it uses tasks as the basic unit of analysis, continuously linking multiple data accesses, tool calls, and result references that occur during the execution of a task. It also combines authorization qualifications, task execution context, risk characteristics, and data exposure control strategies to dynamically monitor the data usage process.
[0304] From the overall processing flow perspective, firstly, by isolating the task triggering context and the authorization decision context, pre-authorization judgment is performed on data access requests. Specifically, the authorization decision context determines whether the current request has basic access qualifications based on trusted information such as subject identity, resource attributes, access control policies, data usage contracts, and authorization credentials. User input, historical interaction content, tool return results, and model inference content in the task triggering context do not directly participate in basic permission granting, but serve as auxiliary information for subsequent behavior association, anomaly analysis, and security auditing. This establishes a separation of responsibilities between basic authorization and task behavior analysis, preventing external input or model-generated content from directly changing data access permissions.
[0305] After establishing basic access privileges, the solution further constructs a target call chain at the granularity of the target task. Data operations such as data access, tool calls, and result referencing are mapped to call nodes. Node attributes and dependencies between nodes are determined based on the operational context, forming a task-level call chain that reflects data flow and operational relationships. Furthermore, through node activity and time decay mechanisms, the impact of different historical operations on the current risk assessment is differentiated, and inactive nodes can be pruned, resulting in an effective call subgraph that better reflects the current task execution state. In this way, data usage behaviors originally scattered across multiple requests, tools, and execution steps can be incorporated into a single task for continuous analysis.
[0306] After obtaining the target call chain, the solution further extracts risk feature values from the call chain and conducts risk assessment based on multiple risk feature values. For example, it can combine information such as data sensitivity, data aggregation breadth, data resource association depth, and task path deviation to quantify the data risk corresponding to the current data access request. Therefore, risk judgment no longer depends solely on the data being requested, but can further consider the data already acquired and processed by the current task, the relationships between different data resources, and the actual execution path of the task. This allows for the identification of data associations, aggregations, and inferred risks that may not be obvious in a single access but gradually form after continuous data use.
[0307] Based on the risk assessment, the solution further establishes a mapping relationship between the risk assessment results and data access control policies, and determines different data exposure methods according to different risk states. For a safe state, the original state of the target data can be returned; for a low-risk state, data precision can be reduced by methods such as sensitive field range filtering; for a medium-risk state, noise can be injected into aggregated statistical data according to the data type and the amount of data acquired by the task, sensitive field masking can be performed when the cumulative amount of detailed data acquired is low, and only data distribution information can be returned when the cumulative amount acquired exceeds the threshold; for a high-risk state, the target access result can be directly determined to be empty, and the data access token can be revoked, subsequent tool calls can be frozen, and the task can be suspended and handed over to the administrator for manual review.
[0308] Therefore, the embodiments in this specification are not simply adding a de-identification algorithm, but rather forming a complete technical chain from basic authorization qualification judgment, task-level behavior association, risk feature extraction, dynamic risk assessment, to risk-driven data access control. Among them, the call chain solves "how to associate multiple data operation behaviors", risk assessment solves "how much risk is there in the continuous use of data after association", data access control solves "how much data should be exposed to the target object for different risks", and the token revocation, task suspension and manual review mechanism further solves "how to prevent subsequent data acquisition in high-risk situations".
[0309] Furthermore, this specification establishes a feedback optimization mechanism based on historical access control information. The system can record target data access requests, target call chains, risk characteristic values, access control policies, and target access results, further forming access control information at the target object level. By periodically analyzing historical access control information to determine control effectiveness evaluation indicators, the system can adjust the impact weights corresponding to risk characteristic values and risk level thresholds, enabling historical data access control results to have a reverse effect on subsequent risk assessments and access controls. Therefore, data access control is no longer a one-time execution of fixed rules, but can be continuously adjusted based on actual control results.
[0310] From the perspective of the correspondence between technical problems and technical means, the embodiments of this specification mainly solve four types of interrelated problems: First, to address the problem that a single request cannot reflect continuous data usage behavior, multiple data operation behaviors are associated through a task-level call chain; Second, to address the problem that traditional static authorization cannot identify continuous data usage risks, dynamic risk assessment is performed through multi-dimensional risk feature values in the call chain; Third, to address the problem that traditional "allow / deny" control methods cannot adapt to different risk levels, hierarchical data access control is achieved through a mapping between risk status and data exposure granularity; Fourth, to address the problem that fixed risk parameters cannot adapt to changes in actual data usage, the effectiveness of control is evaluated through historical access control information, and the influence weight of risk feature values and risk level thresholds are further adjusted.
[0311] Therefore, the core innovation of this specification is not limited to a single risk indicator, a single desensitization method, or a single call chain construction step. Rather, it lies in organizing "basic authorization—task behavior association—risk feature extraction—dynamic risk assessment—risk-driven control—historical feedback optimization" into a continuous data access control closed loop, establishing a link between data access permission judgments and actual data usage behavior during task execution. Even with basic access qualifications, the system can continuously assess task-level data usage risks and dynamically reduce the granularity of data exposure based on the risk level, up to and including blocking data access and freezing subsequent data acquisition behavior. This achieves a shift from "request-level static authorization" to "task-level continuous data access control."
[0312] The following is in conjunction with the appendix Figure 3 Taking the application of the data access control method for trusted data spaces provided in this specification in the scenario of Agent data usage in trusted data spaces as an example, the data access control method for trusted data spaces will be further explained. Figure 3 The present specification illustrates a process flowchart of a data access control method for a trusted data space according to an embodiment of this specification, which specifically includes the following steps.
[0313] Step 302: Intercept the target data access request initiated by the intelligent agent for the target task.
[0314] Step 304: Based on the data operation behaviors involved in the execution of the target task by the intelligent agent, and the operation context information corresponding to the data operation behaviors, construct the target call chain.
[0315] Step 306: Based on the data sensitivity information of the data resources corresponding to each calling node in the target call chain, determine the comprehensive data sensitivity of the target data access request; based on the number of data sources and the cumulative data volume involved in the target call chain, determine the data aggregation breadth of the target data access request; based on the node association depth of at least one node pair in the target call chain, determine the data association depth of the target data access request; based on the node dependency relationship in the target call chain, determine the actual task access path of the target task; based on the path difference between the actual task access path and the baseline task access path, determine the task path deviation of the target data access request.
[0316] Step 308: Based on the comprehensive data sensitivity, data aggregation breadth, data association depth, and task path deviation corresponding to the target data access request, and the influence weights corresponding to the comprehensive data sensitivity, data aggregation breadth, data association depth, and task path deviation respectively, determine the risk assessment score corresponding to the target data access request.
[0317] Step 310: Based on the risk assessment score and the three preset risk level thresholds, determine the target risk level corresponding to the target data access request.
[0318] Step 312: If the target risk level is a security level, then determine the target data requested by the target data access request and use the target data as the target access result.
[0319] Step 314: If the target risk level is low risk, determine the target data requested by the target data access request, and determine the value corresponding to the target sensitive field from the target data; determine the target interval width corresponding to the target sensitive field based on the data exposure granularity control factor; determine the numerical interval into which the value corresponding to the target sensitive field falls based on the target interval width; replace the value corresponding to the target sensitive field in the target data with the interval representative value or interval range of the numerical interval to obtain the target access result.
[0320] Step 316: If the target risk level is medium risk, determine the data type of the target data requested in the target data access request. If the data type indicates that the target data is aggregated statistical data, determine the target noise corresponding to the target data and add noise to the target data based on the target noise to obtain the target access result. If the data type indicates that the target data is detailed data and the target data volume is lower than or equal to the set data volume threshold, determine the target sensitive data in the target data and perform masking processing on the target sensitive data to obtain the target access result. If the data type indicates that the target data is detailed data and the target data volume is higher than the set data volume threshold, determine the data distribution information of the target data and use the data distribution information as the target access result.
[0321] Step 318: If the target risk level is high risk, determine that the target access result corresponding to the target data access request is empty, revoke the target data access token corresponding to the target task to freeze the subsequent tool calls of the target task, suspend the target task, generate risk alarm information based on the target call chain and risk assessment score, and push the risk alarm information to the administrator terminal.
[0322] Step 320: In response to the access recovery command initiated by the administrator terminal, and upon receiving a new target data access request initiated by the target object, determine the target access result of the target data access request based on the medium-risk status control policy.
[0323] Step 322: Based on at least one of the target data access request, target call chain, at least one risk assessment parameter, and target access result, determine the target access control information corresponding to the target task; based on the target access control information corresponding to the target task and the historical access control information corresponding to the historical tasks processed by the target object, determine the object access control information of the intelligent agent, and store the object access control information.
[0324] Step 324: Obtain the stored object access control information according to the set period; determine the control effectiveness evaluation index based on the object access control information; adjust the impact weight and / or risk level threshold based on the control effectiveness evaluation index.
[0325] Corresponding to the above method embodiments, this specification also provides embodiments of a data access control device for a trusted data space. Figure 4 A schematic diagram of a data access control device for a trusted data space, according to one embodiment of this specification, is shown. Figure 4 As shown, the device includes: The call chain determination module 402 is configured to determine the target call chain in response to a target data access request initiated by the target object for the target task. The target call chain is determined based on at least one data operation behavior involved in the execution of the target task and the corresponding operation context information. The operation context information is information related to the operation control of the corresponding data operation behavior. The risk parameter determination module 404 is configured to determine at least one risk assessment parameter corresponding to the target data access request based on the target call chain. The access result determination module 406 is configured to determine the risk assessment information corresponding to the target data access request based on at least one risk assessment parameter, and to determine the target access result corresponding to the target data access request based on the risk assessment information.
[0326] Optionally, the call chain determination module 402 is further configured as follows: Based on at least one data operation behavior, identify at least one calling node in the target call chain; Based on the operation context information corresponding to at least one data operation behavior, determine the node attributes corresponding to at least one calling node; Based on the operation context information corresponding to at least one data operation behavior, determine the node dependency attributes of at least one calling node; The target call chain is determined based on at least one call node, the node attributes corresponding to at least one call node, and the node dependency attributes of at least one call node.
[0327] Optionally, the node dependency attributes include node dependency relationship and node dependency strength; the call chain determination module 402 is further configured as follows: Based on the operation context information corresponding to at least one data operation behavior, determine the operation dependency relationship of at least one data operation behavior, and based on the operation dependency relationship of at least one data operation behavior, determine the node dependency relationship of at least one calling node. Based on the operation context information corresponding to at least one data operation behavior, determine the data dependency of at least one data operation behavior, and based on the data dependency of at least one data operation behavior, determine the node dependency strength of at least one calling node.
[0328] Optionally, the operation context information includes the operation time of the corresponding data operation behavior, and the node attributes include node activity; the call chain determination module 402 is further configured as follows: Based on the time difference between the operation occurrence time corresponding to at least one data operation behavior and the target access time corresponding to the target data access request, the node activity of at least one calling node is determined, wherein the node activity decreases as the time difference increases.
[0329] Optionally, the call chain determination module 402 is further configured as follows: The initial call chain is determined based on at least one call node, the node attributes corresponding to at least one call node, and the node dependency attributes of at least one call node; If at least one of the calling nodes has a first calling node whose node activity is lower than the activity threshold, then the first calling node in the initial calling chain is pruned to obtain the target calling chain. The first calling node can be any one of the at least one calling nodes.
[0330] Optionally, at least one risk assessment parameter includes at least one of comprehensive data sensitivity, data aggregation breadth, data resource association depth, and task path deviation; the risk parameter determination module 404 is further configured to include at least one of the following: Based on the data sensitivity information of the data resources corresponding to each calling node in the target call chain, determine the comprehensive data sensitivity of the target data access request; Based on the number of data sources and the cumulative amount of data involved in the target call chain, the data aggregation breadth of the target data access request is determined. The data aggregation breadth is positively correlated with the number of data sources and the amount of data. The data association depth corresponding to the target data access request is determined based on the node association depth of at least one node pair in the target call chain. Based on the node dependencies in the target call chain, the actual task access path of the target task is determined. Based on the path difference between the actual task access path and the baseline task access path, the task path deviation corresponding to the target data access request is determined.
[0331] Optionally, the risk parameter determination module 404 is further configured as follows: Based on the data sensitivity information of the data resources corresponding to the target call node in the target call chain, and the node activity of the target call node, the node data sensitivity of the target call node is determined. Based on the node data sensitivity of at least one call node, the comprehensive data sensitivity corresponding to the target data access request is determined. The target call node is any one of the call nodes.
[0332] Optionally, the risk parameter determination module 404 is further configured as follows: Based on the node attributes of the target node pair, determine the node similarity of the target node pair; The node association depth of the target node pair is determined based on the node similarity and node activity of the target node pair, wherein the target node pair is any one of at least one node pair.
[0333] Optionally, the risk assessment information is a risk assessment score; the access result determination module 406 is further configured as follows: Based on at least one risk assessment parameter and the influence weight corresponding to at least one risk assessment parameter, determine the risk assessment score corresponding to the target data access request.
[0334] Optionally, the access result determination module 406 is further configured as follows: Based on the risk assessment score and at least one preset risk level threshold, determine the target risk level corresponding to the target data access request. Based on the target risk level, a target access control policy is determined from at least one preset access control policy, and based on the target access control policy, the target access result corresponding to the target data access request is determined. Different access control policies correspond to different data exposure granularities, and the data exposure granularity is used to represent the precision of the data returned to the target object.
[0335] Optionally, at least one access control policy includes a security state control policy with data exposure granularity decreasing sequentially, a low-risk state control policy, a medium-risk state control policy, and a high-risk state control policy; the access result determination module 406 is further configured to: If the target risk level is a security level, then the security state control policy is determined as the target access control policy, and based on the security state control policy, the target access result corresponding to the target data access request is determined. If the target risk level is low risk, then the low risk state control policy is determined as the target access control policy, and based on the low risk state control policy, the target access result corresponding to the target data access request is determined. If the target risk level is medium risk, then the medium risk status control policy is determined as the target access control policy, and based on the medium risk status control policy, the target access result corresponding to the target data access request is determined. If the target risk level is high risk, then the high risk state control policy is determined as the target access control policy, and based on the high risk state control policy, the target access result corresponding to the target data access request is determined.
[0336] Optionally, the security state control policy is used to instruct that the data accessed by the data access request be returned in its original state; the access result determination module 406 is further configured to: Determine the target data requested in the target data access request, and use the target data as the target access result.
[0337] Optionally, the low-risk state control strategy is used to instruct the range mapping of values corresponding to sensitive fields in the data requested by the data access request; the access result determination module 406 is further configured to: Determine the target data requested in the target data access request, and determine the value corresponding to the target sensitive field from the target data; Based on the data exposure granularity control factor, the target interval width corresponding to the target sensitive field is determined. The data exposure granularity control factor is determined based on the risk assessment score and is used to control the data exposure granularity of the data returned to the target object based on the target data access request. Based on the target interval width, determine the range of values that the corresponding values of the target sensitive fields fall into; Replace the values corresponding to the target sensitive fields in the target data with the range representative value or range of the value interval to obtain the target access results.
[0338] Optionally, the medium-risk status control strategy is used to indicate the data type of the data requested in the data access request and the amount of data accumulated by the task corresponding to the data access request during task execution, and to determine the target access result corresponding to the data access request; the access result determination module 406 is further configured to: Determine the target data requested in the target data access request; The target access result is determined based on the data type of the target data and the amount of target data accumulated by the target task during task execution.
[0339] Optionally, the access result determination module 406 is further configured as follows: If the data type indicates that the target data is aggregated statistical data, then the target noise corresponding to the target data is determined, and the target data is noise-added based on the target noise to obtain the target access result; If the data type indicates that the target data is detailed data and the target data volume is less than or equal to the set data volume threshold, then the target sensitive data in the target data is identified, and the target sensitive data is masked to obtain the target access result. If the data type indicates that the target data is detailed data and the target data volume is higher than the set data volume threshold, then the data distribution information of the target data is determined and the data distribution information is used as the target access result.
[0340] Optionally, the access result determination module 406 is further configured as follows: Based on the global sensitivity of the query function corresponding to the target data and the data exposure granularity control factor, the noise parameters of the target noise are determined, and the target noise is determined based on the noise parameters. Among them, the data exposure granularity control factor is determined based on the risk assessment score and is used to control the data exposure granularity returned to the target object based on the target data access request. The data exposure granularity query function is a function that performs aggregate statistical processing on the original data corresponding to the target data access request to obtain the target data. The global sensitivity is used to indicate the degree of impact of a change in a piece of data in the original data on the query result. The noise parameter is positively correlated with the global sensitivity and negatively correlated with the data exposure granularity control factor.
[0341] Optionally, the high-risk status control strategy is used to indicate that the access result of the data access request is empty; the access result determination module 406 is further configured to: The target access result corresponding to the target data access request is determined to be empty.
[0342] The access result determination module 406 is also configured to: revoke the target data access token corresponding to the target task to freeze subsequent tool calls of the target task.
[0343] The access result determination module 406 is also configured to: suspend the target task, generate risk alarm information based on the target call chain and risk assessment score, and push the risk alarm information to the administrator terminal; In response to an access recovery command initiated by the administrator terminal, upon receiving a new target data access request from the target object, the target access result of the target data access request is determined based on the medium-risk status control policy.
[0344] Optionally, the above-mentioned device further includes a storage module configured to: Based on at least one of the target data access request, the target call chain, at least one risk assessment parameter, and the target access result, determine the target access control information corresponding to the target task. Based on the target access control information corresponding to the target task and the historical access control information corresponding to the historical tasks processed by the target object, the object access control information of the target object is determined and stored.
[0345] Optionally, the above-mentioned device further includes a feedback module, configured to: Retrieve stored object access control information at set intervals; Based on object access control information, control effectiveness evaluation indicators are determined. These indicators characterize the effectiveness of data access control implemented on data access requests initiated by the target object. Based on the control effectiveness evaluation indicators, adjust the impact weight of at least one risk assessment parameter and / or at least one risk level threshold.
[0346] The above is an illustrative scheme of a data access control device for a trusted data space according to this embodiment. It should be noted that the technical solution of this data access control device for a trusted data space and the technical solution of the data access control method for a trusted data space described above belong to the same concept. Details not described in detail in the technical solution of the data access control device for a trusted data space can be found in the description of the technical solution of the data access control method for a trusted data space described above.
[0347] Figure 5 A structural block diagram of a computing device 500 according to one embodiment of this specification is shown. The components of the computing device 500 include, but are not limited to, a memory 510 and a processor 520. The processor 520 is connected to the memory 510 via a bus 530, and a database 550 is used to store data.
[0348] The computing device 500 also includes an access device 540, which enables the computing device 500 to communicate via one or more networks 560. Examples of these networks include Public Switched Telephone Network (PSTN), Local Area Network (LAN), Wide Area Network (WAN), Personal Area Network (PAN), or combinations of communication networks such as the Internet. The access device 540 may include one or more of any type of wired or wireless network interface (e.g., a network interface card (NIC)), such as an IEEE 802.11 Wireless Local Area Network (WLAN) wireless interface, a Wi-MAX (Worldwide Interoperability for Microwave Access) interface, an Ethernet interface, a Universal Serial Bus (USB) interface, a cellular network interface, a Bluetooth interface, or a Near Field Communication (NFC) interface.
[0349] In one embodiment of this specification, the above-described components of the computing device 500 and Figure 5 Other components, not shown, can also be connected to each other, for example, via a bus. It should be understood that... Figure 5 The block diagram of the computing device shown is for illustrative purposes only and is not intended to limit the scope of this specification. Those skilled in the art can add or replace other components as needed.
[0350] The computing device 500 can be any type of stationary or mobile computing device, including mobile computers or mobile computing devices (e.g., tablet computers, personal digital assistants, laptop computers, notebook computers, netbooks, etc.), mobile phones (e.g., smartphones), wearable computing devices (e.g., smartwatches, smart glasses, etc.) or other types of mobile devices, or stationary computing devices such as desktop computers or personal computers (PCs). The computing device 500 can also be a mobile or stationary server.
[0351] The processor 520 is configured to execute the following computer-executable instructions, which, when executed by the processor, implement the steps of the aforementioned data access control method for trusted data space.
[0352] The above is an illustrative scheme of a computing device according to this embodiment. It should be noted that the technical solution of this computing device and the technical solution of the data access control method for trusted data space described above belong to the same concept. For details not described in detail in the technical solution of the computing device, please refer to the description of the technical solution of the data access control method for trusted data space described above.
[0353] An embodiment of this specification also provides a computer-readable storage medium storing computer-executable instructions that, when executed by a processor, implement the steps of the data access control method for a trusted data space described above.
[0354] The above is an illustrative scheme of a computer-readable storage medium according to this embodiment. It should be noted that the technical solution of this storage medium belongs to the same concept as the technical solution of the data access control method for trusted data space described above. Details not described in detail in the technical solution of the storage medium can be found in the description of the technical solution of the data access control method for trusted data space described above.
[0355] An embodiment of this specification also provides a computer program, wherein when the computer program is executed in a computer, it causes the computer to perform the steps of the data access control method for the trusted data space described above.
[0356] The above is an illustrative scheme of a computer program according to this embodiment. It should be noted that the technical solution of this computer program belongs to the same concept as the technical solution of the data access control method for trusted data space described above. For details not described in detail in the technical solution of the computer program, please refer to the description of the technical solution of the data access control method for trusted data space described above.
[0357] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.
[0358] The computer instructions include computer program code, which may be in the form of source code, object code, executable file, or certain intermediate forms. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium may be appropriately added or removed according to the requirements of patent practice. For example, in some regions, according to patent practice, computer-readable media may not include electrical carrier signals and telecommunication signals.
[0359] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that the embodiments in this specification are not limited to the described order of actions, because according to the embodiments in this specification, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in this specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to the embodiments in this specification.
[0360] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.
[0361] The preferred embodiments disclosed above are merely illustrative of this specification. The optional embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the embodiments described herein. These embodiments are selected and specifically described in this specification to better explain the principles and practical applications of the embodiments, thereby enabling those skilled in the art to better understand and utilize this specification. This specification is limited only by the claims and their full scope and equivalents.
Claims
1. A data access control method for a trusted data space, characterized in that, include: In response to a target data access request initiated by a target object for a target task, a target call chain is determined. The target call chain is determined based on at least one data operation behavior involved in the execution of the target task and the corresponding operation context information. The operation context information is information related to the operation control of the corresponding data operation behavior. Based on the target call chain, at least one risk assessment parameter corresponding to the target data access request is determined; Based on the at least one risk assessment parameter, determine the risk assessment information corresponding to the target data access request, and based on the risk assessment information, determine the target access result corresponding to the target data access request.
2. The method according to claim 1, characterized in that, The determination of the target call chain includes: Based on the at least one data operation behavior, at least one calling node in the target calling chain is determined; Based on the operation context information corresponding to the at least one data operation behavior, determine the node attributes corresponding to the at least one calling node; Based on the operation context information corresponding to the at least one data operation behavior, determine the node dependency attributes of the at least one calling node; The target call chain is determined based on the at least one calling node, the node attributes corresponding to the at least one calling node, and the node dependency attributes of the at least one calling node.
3. The method according to claim 2, characterized in that, The node dependency attributes include node dependency relationships and node dependency strength; determining the node dependency attributes of the at least one calling node based on the operation context information corresponding to the at least one data operation behavior includes: Based on the operation context information corresponding to the at least one data operation behavior, determine the operation dependency relationship of the at least one data operation behavior, and based on the operation dependency relationship of the at least one data operation behavior, determine the node dependency relationship of the at least one calling node; Based on the operation context information corresponding to the at least one data operation behavior, the data dependency of the at least one data operation behavior is determined, and based on the data dependency of the at least one data operation behavior, the node dependency strength of the at least one calling node is determined.
4. The method according to claim 2, characterized in that, The operation context information includes the operation time of the corresponding data operation behavior, and the node attributes include node activity. The step of determining the node attributes corresponding to the at least one calling node based on the operation context information corresponding to the at least one data operation behavior includes: The node activity of the at least one calling node is determined based on the time difference between the operation occurrence time corresponding to the at least one data operation behavior and the target access time corresponding to the target data access request, wherein the node activity decreases as the time difference increases.
5. The method according to claim 4, characterized in that, Determining the target call chain based on the at least one calling node, the node attributes corresponding to the at least one calling node, and the node dependency attributes of the at least one calling node includes: The initial call chain is determined based on the at least one calling node, the node attributes corresponding to the at least one calling node, and the node dependency attributes of the at least one calling node; If there is a first calling node among the at least one calling nodes whose node activity is lower than the activity threshold, then the first calling node in the initial calling chain is pruned to obtain the target calling chain, wherein the first calling node is any one of the at least one calling nodes.
6. The method according to claim 1, characterized in that, The at least one risk assessment parameter includes at least one of the following: comprehensive data sensitivity, breadth of data aggregation, depth of data resource association, and deviation of task path; The determination of at least one risk assessment parameter for the target data access request based on the target call chain includes at least one of the following: Based on the data sensitivity information of the data resources corresponding to each calling node in the target calling chain, the comprehensive data sensitivity corresponding to the target data access request is determined. Based on the number of data sources and the cumulative data volume involved in the target call chain, the data aggregation breadth of the target data access request is determined, and the data aggregation breadth is positively correlated with the number of data sources and the data volume; Based on the node association depth of at least one node pair in the target call chain, the data association depth corresponding to the target data access request is determined. Based on the node dependencies in the target call chain, the actual task access path of the target task is determined. Based on the path difference between the actual task access path and the baseline task access path, the task path deviation corresponding to the target data access request is determined.
7. The method according to claim 6, characterized in that, The determination of the comprehensive data sensitivity corresponding to the target data access request based on the data sensitivity information of the data resources corresponding to each calling node in the target call chain includes: Based on the data sensitivity information of the data resources corresponding to the target call node in the target call chain, and the node activity of the target call node, the node data sensitivity of the target call node is determined, and based on the node data sensitivity of at least one call node, the comprehensive data sensitivity corresponding to the target data access request is determined, wherein the target call node is any one of the call nodes.
8. The method according to claim 6, characterized in that, Determining the data association depth corresponding to the target data access request based on the node association depth of at least one node pair in the target call chain includes: Based on the node attributes of the target node pair, determine the node similarity of the target node pair; Based on the node similarity of the target node pair and the node activity of the target node pair, the node association depth of the target node pair is determined, wherein the target node pair is any one of the at least one node pair.
9. The method according to claim 1, characterized in that, The risk assessment information is a risk assessment score; The step of determining the risk assessment information corresponding to the target data access request based on the at least one risk assessment parameter includes: Based on the at least one risk assessment parameter and the influence weight corresponding to the at least one risk assessment parameter, the risk assessment score corresponding to the target data access request is determined.
10. The method according to claim 9, characterized in that, The step of determining the target access result corresponding to the target data access request based on the risk assessment information includes: Based on the risk assessment score and at least one preset risk level threshold, the target risk level corresponding to the target data access request is determined. Based on the target risk level, a target access control policy is determined from at least one preset access control policy, and based on the target access control policy, the target access result corresponding to the target data access request is determined. Different access control policies correspond to different data exposure granularities, and the data exposure granularity is used to represent the accuracy of the data returned to the target object.
11. The method according to claim 10, characterized in that, The at least one access control policy includes a security state control policy with data exposure granularity decreasing sequentially, a low-risk state control policy, a medium-risk state control policy, and a high-risk state control policy. The step of determining a target access control policy from at least one preset access control policy based on the target risk level, and determining the target access result corresponding to the target data access request based on the target access control policy, includes: If the target risk level is a security level, then the security status control policy is determined to be the target access control policy, and based on the security status control policy, the target access result corresponding to the target data access request is determined. If the target risk level is low risk, then the low risk state control policy is determined as the target access control policy, and based on the low risk state control policy, the target access result corresponding to the target data access request is determined. If the target risk level is medium risk, then the medium risk status control strategy is determined as the target access control strategy, and based on the medium risk status control strategy, the target access result corresponding to the target data access request is determined. If the target risk level is high risk, then the high risk state control policy is determined as the target access control policy, and based on the high risk state control policy, the target access result corresponding to the target data access request is determined.
12. The method according to claim 11, characterized in that, The security state control policy is used to instruct that the data accessed by the data access request be returned in its original state. The step of determining the target access result corresponding to the target data access request based on the security state control policy includes: Determine the target data requested by the target data access request, and use the target data as the target access result.
13. The method according to claim 11, characterized in that, The low-risk state control strategy is used to instruct the range mapping of the values corresponding to the sensitive fields in the data requested by the data access request. The step of determining the target access result corresponding to the target data access request based on the low-risk state control strategy includes: Determine the target data requested by the target data access request, and determine the value corresponding to the target sensitive field from the target data; Based on the data exposure granularity control factor, the target interval width corresponding to the target sensitive field is determined, wherein the data exposure granularity control factor is determined based on the risk assessment score and is used to control the data exposure granularity of the data returned to the target object based on the target data access request; Based on the target interval width, determine the range of values that the value corresponding to the target sensitive field falls into; Replace the value corresponding to the target sensitive field in the target data with the range representative value or range of the value range to obtain the target access result.
14. The method according to claim 11, characterized in that, The medium-risk status control strategy is used to indicate the data type of the data requested based on the data access request, as well as the amount of data accumulated by the task corresponding to the data access request during task execution, and to determine the target access result corresponding to the data access request. The determination of the target access result corresponding to the target data access request based on the medium-risk status control strategy includes: Determine the target data requested in the target data access request; The target access result is determined based on the data type of the target data and the amount of target data accumulated by the target task during task execution.
15. The method according to claim 14, characterized in that, The determination of the target access result based on the data type of the target data and the amount of target data accumulated by the target task during task execution includes: If the data type indicates that the target data is aggregated statistical data, then the target noise corresponding to the target data is determined, and the target data is noise-added based on the target noise to obtain the target access result; If the data type indicates that the target data is detailed data, and the target data volume is lower than or equal to the set data volume threshold, then the target sensitive data in the target data is determined, and the target sensitive data is masked to obtain the target access result. If the data type indicates that the target data is detailed data and the target data volume is higher than the set data volume threshold, then the data distribution information of the target data is determined, and the data distribution information is used as the target access result.
16. The method according to claim 11, characterized in that, Determining the target noise corresponding to the target data includes: Based on the global sensitivity of the query function corresponding to the target data and the data exposure granularity control factor, the noise parameters of the target noise are determined, and the target noise is determined based on the noise parameters; The data exposure granularity control factor is determined based on the risk assessment score and is used to control the data exposure granularity of the data returned to the target object based on the target data access request. The query function is a function that performs aggregate statistical processing on the original data corresponding to the target data access request to obtain the target data. The global sensitivity is used to indicate the degree of influence of a change in a piece of data in the original data on the query result. The noise parameter is positively correlated with the global sensitivity and negatively correlated with the data exposure granularity control factor.
17. The method according to claim 11, characterized in that, The high-risk status control strategy is used to indicate that the access result of the data access request is empty; The determination of the target access result corresponding to the target data access request based on the high-risk state control strategy includes: The target access result corresponding to the target data access request is determined to be empty.
18. The method according to claim 17, characterized in that, After determining that the target access result corresponding to the target data access request is empty, the method further includes: Revoke the target data access token corresponding to the target task to freeze subsequent tool calls for the target task.
19. The method according to claim 17, characterized in that, After determining that the target access result corresponding to the target data access request is empty, the method further includes: The target task is suspended, and a risk alarm is generated based on the target call chain and the risk assessment score. The risk alarm is then pushed to the administrator terminal. In response to the access recovery command initiated by the administrator terminal, upon receiving the target data access request initiated by the target object again, the target access result of the target data access request is determined based on the medium-risk status control policy.
20. The method according to claim 10, characterized in that, After determining the target access result corresponding to the target data access request based on the target access control policy, the method further includes: Based on at least one of the target data access request, the target call chain, the at least one risk assessment parameter, and the target access result, determine the target access control information corresponding to the target task; Based on the target access control information corresponding to the target task and the historical access control information corresponding to the historical tasks processed by the target object, the object access control information of the target object is determined and stored.
21. The method according to claim 20, characterized in that, After storing the object access control information, the method further includes: Retrieve the stored object access control information according to a set period; Based on the object access control information, a control effectiveness evaluation index is determined, wherein the control effectiveness evaluation index is used to characterize the effectiveness of the data access control performed on the data access request initiated by the target object; Based on the control effectiveness evaluation index, adjust the influence weight corresponding to the at least one risk assessment parameter and / or the at least one risk level threshold.
22. A data access control device for a trusted data space, characterized in that, include: The call chain determination module is configured to determine the target call chain in response to a target data access request initiated by the target object for the target task. The target call chain is determined based on at least one data operation behavior involved in the execution of the target task and the corresponding operation context information. The operation context information is information related to the operation control of the corresponding data operation behavior. The risk parameter determination module is configured to determine at least one risk assessment parameter corresponding to the target data access request based on the target call chain. The access result determination module is configured to determine the risk assessment information corresponding to the target data access request based on the at least one risk assessment parameter, and to determine the target access result corresponding to the target data access request based on the risk assessment information.
23. A computing device, characterized in that, include: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, they implement the steps of the data access control method for trusted data space according to any one of claims 1-21.
24. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, implement the steps of the data access control method for a trusted data space as described in any one of claims 1-21.
25. A computer program product, characterized in that, It includes a computer program / instruction that, when executed by a processor, implements the steps of the data access control method for a trusted data space as described in any one of claims 1-21.
