Permission credible mutual recognition method, device and system of cross-domain agent

Abstract

The invention provides a credible permission mutual recognition method, device and system for cross-domain agents, and relates to the technical field of artificial intelligence security, and the method comprises the steps: receiving a call request from a source domain agent for executing a corresponding operation on a target domain agent; obtaining an affine transformation parameter set through a permission conversion model according to the space relation characteristics of a source domain permission space and a target domain permission space involved in the calling request, and finally screening out affine transformation parameters suitable for permission voucher mapping by utilizing cone volume measurement evaluation and security verification, and mapping an original permission voucher of the source domain agent from a source domain permission space to a target domain permission space by using the affine transformation parameter, and generating a mapped permission voucher which can be identified and verified by the target domain. According to the method, the permission space is geometrized, and affine transformation, space division and geometric calculation are utilized, so that the permission credible mutual recognition problem in cross-domain agent collaboration is effectively solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence security, in particular to a cross-domain intelligent agent permission trusted mutual recognition method, device and system. BACKGROUND

[0002] In the artificial intelligence driven cross-domain collaboration scenario, the heterogeneity of the permission system has become the main challenge to realize the trusted mutual recognition between intelligent agents. In medical research, the intelligent data analysis agent (Chinese translation: intelligent agent, is the core concept in the field of artificial intelligence) of the medical research center, namely Med-Research-AI (full name: Medical Research AI, Chinese translation: medical research intelligent agent), needs to access the intelligent medical record library management Hosp-Record-DB (full name: Hospital Record Database AI, Chinese translation: hospital medical record library management intelligent agent) of the cooperative hospital to perform the operation of reading the anonymized medical record data of the diabetes patient. The permission of the research center is defined in its two-dimensional geometric space with data sensitivity level and research project level as coordinate axes, and the certificate held by Med-Research-AI is point (4, 8); while the coordinate axes of the hospital permission space are also two-dimensional structure, but its coordinate axes are defined as operation risk level and accessible data range; the permission dimension semantics between the two domains is different, and the numerical range is also inconsistent, resulting in that the permission certificate (4, 8) of the source domain cannot be directly understood and used in the target domain.

[0003] Therefore, cross-domain permission verification depends on pre-established static mapping table or complex dynamic trust negotiation protocol; the static mapping table lacks flexibility and is difficult to adapt to the dynamic adjustment of the permission policy of any domain; and the complex negotiation protocol introduces significant performance overhead and implementation complexity, which cannot meet the needs of efficient collaboration. SUMMARY

[0004] The technical problem to be solved by the present application is to provide a cross-domain intelligent agent permission trusted mutual recognition method, device and system, which effectively solves the permission trusted mutual recognition problem in cross-domain intelligent agent collaboration by geometricizing the permission space and using affine transformation, space division and geometric calculation, and brings significant improvement in accuracy, security, flexibility, fineness and efficiency.

[0005] To solve the above technical problems, the technical solutions of the present application are as follows: In a first aspect, a cross-domain intelligent agent permission trusted mutual recognition method is provided, which comprises: receiving a calling request from a source domain intelligent agent to a target domain intelligent agent to perform a corresponding operation; According to the spatial relationship characteristics of the source domain authority space and the target domain authority space involved in the calling request, affine transformation parameter sets are obtained through an authority conversion model, and a cone volume metric is evaluated and security is verified, so that affine transformation parameters suitable for authority credential mapping are finally screened out, and the original authority credentials of the source domain agent are mapped from the source domain authority space to the target domain authority space using the affine transformation parameters, to generate mapped authority credentials that can be recognized and verified in the target domain; Based on the mapped authority credentials, the target domain authority space is divided into multiple authority regions through a clustering algorithm in combination with a set of authority center points of the agent in the target domain authority space, a target authority region to which the mapped authority credentials belong is determined, and a permission range polygon of the agent in the target domain is generated according to a set of boundary points of the target authority region; According to the corresponding operation in the calling request, the corresponding operation point in the target domain authority space is obtained, and a ray method geometric inclusion detection algorithm is used to judge whether the operation point is located inside the permission range polygon based on the permission range polygon, if it is determined that the operation point is located inside, the corresponding operation is allowed to be executed, and the calling request is forwarded to the target domain agent for execution.

[0006] Further, the calling request contains the original authority credentials of the source domain agent, which are represented as geometric coordinate points in the source domain authority space, for affine transformation and cross-domain mapping.

[0007] Further, according to the spatial relationship characteristics of the source domain authority space and the target domain authority space involved in the calling request, affine transformation parameter sets are obtained through an authority conversion model, and a cone volume metric is evaluated and security is verified, so that affine transformation parameters suitable for authority credential mapping are finally screened out, including: Based on the source domain authority space and the target domain authority space involved in the calling request, spatial relationship characteristics between the two are extracted, including authority dimension mapping rules, coordinate scaling ratios, and spatial rotation angles; The extracted spatial relationship characteristics are input into a pre-trained authority conversion model for processing to obtain output affine transformation parameter sets; A cone is generated covering the target domain authority space with the center of the target domain authority space as the apex and the mapping direction defined by the affine transformation parameters as the cone axis direction; the actual intersection volume of the cone and the target domain authority space is calculated, and the ratio of the volume value to the total volume of the target domain authority space is taken as the cone volume metric value, which is used to quantitatively evaluate the coverage degree of the affine transformation parameters on the target domain authority space; The parameters in the affine transformation parameter set are sorted in descending order based on the cone volume metric, and a subset of parameters with metric values ​​higher than a preset threshold are selected. This subset of parameters is then subjected to compatibility and risk verification according to the security policy of the target domain permission space to obtain the parameters that pass the verification. The parameters that pass the verification are used as the final affine transformation parameters for permission credential mapping.

[0008] Furthermore, the original authorization credentials of the source domain agent are mapped from the source domain authorization space to the target domain authorization space using the affine transformation parameters, generating a mapped authorization credential that can be recognized and verified in the target domain, including: Read the final affine transformation parameters that have passed verification and parse them into translation vectors, scaling matrices, and rotation matrices for coordinate transformation; Based on translation vectors, scaling matrices, and rotation matrices, the geometric coordinates of the original permission credential are sequentially scaled, rotated, and translated to obtain new coordinates in the target domain permission space. The obtained new coordinate points are encapsulated according to the format of the target domain permission credential, and a spatial location verification label is attached to generate a mapped permission credential that can be identified and verified by the target domain.

[0009] Furthermore, based on the mapped permission credentials and combined with the set of permission center points of agents in the target domain permission space, a clustering algorithm is used to divide the target domain permission space into multiple permission regions, including: Obtain the coordinates of the mapped permission credential in the target domain permission space; Based on the coordinate positions, load a predefined set of permission center points for the target domain, where each permission center point corresponds to a permission level; Using the loaded set of permission center points as cluster centers, a distance metric algorithm is used to dynamically divide the target domain permission space, generating multiple convex polygon permission regions defined by boundary point sets; Based on the divided permission areas, calculate the spatial distance between the coordinates of the mapped permission credential and the center point of each permission; Based on the calculated spatial distance, the mapped permission credentials are assigned to the target permission area corresponding to the permission center point with the smallest distance.

[0010] Further, the target permission region to which the mapped permission credential belongs is determined, and a permission range polygon mapped to the source domain agent in the target domain is generated based on the boundary point set of the target permission region, including: Obtain the target permission region identifier corresponding to the mapped permission credential; Based on the target permission region identifier, retrieve the set of boundary points corresponding to the target permission region; A closed convex polygon geometry structure is constructed in spatial order connection by using the retrieved boundary point set; The convex polygon geometry structure is topologically checked, and the polygon passing the check is bound to the mapped permission credential as the permission range polygon of the source domain agent in the target domain.

[0011] Further, according to the corresponding operation in the call request, the corresponding operation point in the target domain permission space is obtained, and based on the permission range polygon, a ray method geometric inclusion detection algorithm is used to judge whether the operation point is located inside the permission range polygon. If it is judged that the operation point is located inside, the corresponding operation is allowed to be executed, and the call request is forwarded to the target domain agent for execution, including: The operation description information in the call request is parsed, and the corresponding operation point coordinates in the target domain permission space are extracted; Based on the bound permission range polygon structure, the boundary vertex sequence is read; Using the operation point coordinates, an infinite extension detection ray is emitted along the preset direction of the target domain permission space; According to the boundary vertex sequence of the detection ray and the permission range polygon, all intersection point coordinates are calculated; Based on the obtained intersection point coordinates, the number of valid intersection points is counted. If it is odd, it is judged that the operation point is located inside the permission range polygon, and if it is even, it is judged that it is located outside; If the operation point is located inside the permission range polygon, an evaluation result of allowing the corresponding operation to be executed is generated; The evaluation result is attached to the call request, and is forwarded to the target domain agent to trigger the execution of the corresponding operation.

[0012] The second aspect is a cross-domain intelligent agent permission trusted mutual recognition system, including: A receiving module is used to receive a call request from a source domain intelligent agent to a target domain intelligent agent to execute a corresponding operation; A generating module is used to obtain an affine transformation parameter set through a permission conversion model according to the spatial relationship characteristics of the source domain permission space and the target domain permission space involved in the call request, and finally filter out the affine transformation parameters suitable for permission credential mapping, and use the affine transformation parameters to map the original permission credential of the source domain intelligent agent from the source domain permission space to the target domain permission space, and generate a target domain recognizable and verifiable mapped permission credential; The dividing module is configured to divide the target domain authority space into a plurality of authority areas by a clustering algorithm based on the mapped authority credential and a set of authority center points of agents in the target domain authority space, determine a target authority area to which the mapped authority credential belongs, and generate a permission range polygon of the mapped authority credential of the source domain agent in the target domain according to a boundary point set of the target authority area. The processing module is configured to acquire a corresponding operation point of the corresponding operation in the target domain authority space according to the operation in the calling request, judge whether the operation point is located inside the permission range polygon by using a ray method geometric inclusion detection algorithm based on the permission range polygon, and if the operation point is located inside, allow the corresponding operation to be executed and forward the calling request to the target domain agent for execution.

[0013] In a third aspect, a computing device includes: one or more processors; a memory device storing one or more programs, which when executed by the one or more processors, cause the one or more processors to implement the method.

[0014] In a fourth aspect, a computer-readable storage medium stores a program, which when executed by a processor, implements the method.

[0015] The above scheme of the present application at least has the following beneficial effects: By introducing the analysis of the topological relationship of the authority space, the spatial relationship characteristics including the authority dimension mapping rule, the coordinate scaling ratio and the space rotation angle are dynamically extracted, and the pre-trained authority conversion model is used to generate the affine transformation parameters, so that the semantics and value range of the source domain authority credential can be accurately mapped to the target domain authority space to generate the authority credential that can be recognized and verified in the target domain, which overcomes the limitations of the traditional static mapping table in dealing with semantic differences and value range mismatches, and realizes the lossless conversion and accurate expression of the authority meaning between different domains.

[0016] The affine transformation parameters are dynamically determined and need to be verified by the target domain security policy, which means that when the authority policy, space definition or security requirement of any domain changes, the system can quickly adapt by recalculating and verifying the transformation parameters without the need to redesign or frequently update the complex static mapping protocol, which significantly improves the adaptability to dynamic environment and solves the problem of poor flexibility and difficulty in coping with dynamic adjustment of the traditional method.

[0017] By determining the attribution of mapped permission credentials to convex polygon permission regions divided based on predefined permission center points of the target domain, and binding the credentials to their corresponding permission range polygons, the permission verification problem is transformed into a rigorous geometric spatial inclusion relationship judgment problem. This effectively prevents unauthorized access and ensures the reliability and security of authorization decisions. Furthermore, the security policy verification step added during the mapping process further guarantees that the conversion process and results meet the security requirements of the target domain.

[0018] This invention not only realizes cross-domain mapping of agent permissions, but also further refines the granularity of permission control to specific operations. By mapping the corresponding operation to the operation point in the target domain permission space, and determining whether the point is located inside the permission range polygon of the requesting agent, it determines whether to allow the execution of the operation. This mechanism realizes fine-grained judgment on whether an agent is allowed to execute a corresponding operation in the target domain, improving the precision and flexibility of permission management. Attached Figure Description

[0019] Figure 1 This is a flowchart illustrating a cross-domain intelligent agent authorization trust mutual recognition method provided by an embodiment of the present invention.

[0020] Figure 2 This is a schematic diagram of a cross-domain intelligent agent authorization trust mutual recognition system provided by an embodiment of the present invention. Detailed Implementation

[0021] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0022] like Figure 1 As shown, an embodiment of the present invention proposes a method for trusted mutual recognition of permissions between cross-domain intelligent agents, the method comprising the following steps: Step 1: Receive a call request from the source domain agent to perform a corresponding operation on the target domain agent; Step 2: Based on the spatial relationship characteristics between the source domain permission space and the target domain permission space involved in the call request, obtain the affine transformation parameter set through the permission transformation model, and use the cone volume metric for evaluation and security verification to finally select the affine transformation parameters suitable for permission credential mapping. Then, use the affine transformation parameters to map the original permission credential of the source domain agent from the source domain permission space to the target domain permission space to generate a mapped permission credential that can be identified and verified in the target domain. Step 3, based on the mapped permission credential, in combination with the permission center point set of the agent in the target domain permission space, the target domain permission space is divided into multiple permission areas through a clustering algorithm, the target permission area to which the mapped permission credential belongs is determined, and the boundary point set of the target permission area is used to generate a permission range polygon of the agent in the source domain mapped in the target domain; Step 4, according to the corresponding operation in the calling request, the corresponding operation point in the target domain permission space is obtained, based on the permission range polygon, a ray method geometric inclusion detection algorithm is used to judge whether the operation point is located inside the permission range polygon, if it is judged that the operation point is located inside, the corresponding operation is allowed to be executed, and the calling request is forwarded to the target domain agent for execution.

[0023] In the embodiment of the application, by normalizing the reception and processing of cross-domain calling requests, the secure access and initialization of permission requests between heterogeneous systems are realized, by dynamically analyzing the permission space topological relationship and determining the affine transformation parameters, the source domain permission credential is accurately mapped into a form recognizable by the target domain, effectively solving the heterogeneity problem of different permission systems in semantics and numerical range, and improving the accuracy and cross-domain adaptability of the mapping. Then, based on the pre-defined permission center point of the target domain, the spatial division algorithm is used to generate the permission area and determine the permission range polygon, realizing the fine and geometric area management of the mapped permission, enhancing the granularity and flexibility of the permission control, and by mapping the operation to a spatial point and using the geometric inclusion detection algorithm for permission verification, it is ensured that only the operations within the permission range will be released, thereby realizing the trusted mutual recognition of cross-domain permissions and the automatic execution of operations while ensuring security.

[0024] In a preferred embodiment of the application, the calling request contains the original permission credential of the source domain agent, and the original permission credential is represented as a geometric coordinate point in the source domain permission space for affine transformation and cross-domain mapping.

[0025] In the embodiment of the application, by including the original permission credential of the source domain agent in the calling request, the original permission credential is represented as a geometric coordinate point in the source domain permission space for affine transformation and cross-domain mapping, which converts abstract permission information into concrete spatial geometric points, so that the permission credential has precise, calculable and measurable mathematical representation. At the same time, this method enhances the standardization and consistency of permission expression, avoids semantic ambiguity caused by permission description language or structural differences, and significantly improves the explainability and processing efficiency in the cross-domain permission mutual recognition process.

[0026] In a preferred embodiment of the application, step 1, receiving a calling request from a source domain agent to a target domain agent for performing a corresponding operation, the specific process is as follows: A standardized secure communication interface module is deployed as a unified access point for cross-domain permission mutual recognition process, which is responsible for continuous listening and receiving cross-domain operation call requests initiated by source domain agents. When the request arrives, the interface module first performs strict inbound request protocol compliance check to verify whether the message format, encryption method and transmission protocol meet the predefined interaction specification; at the same time, identity preliminary authentication is performed through digital signature verification and other technical means, the key operation semantic elements are accurately extracted from the encrypted message, including the unique identifier of the target domain agent, the specific operation type to be executed, and the most critical original permission credentials of the source domain agent, which are represented as one or more geometric coordinate points with explicit semantics in the source domain permission space.

[0027] In the specific embodiment of cross-medical collaboration, the intelligent data analysis Agent of the medical research center, i.e. Med-Research-AI, as a source domain agent, needs to access the intelligent medical record database management Agent of the cooperative hospital, i.e. Hosp-Record-DB, to execute the operation of reading the anonymized medical record data of diabetic patients. Med-Research-AI applies for and obtains a geometric coordinate point (4, 8) representing its permission in its source domain permission space, where the horizontal coordinate value 4 represents that it is authorized to process more sensitive health data in the data sensitivity level dimension, and the vertical coordinate value 8 represents that it belongs to a research project with high priority and trust in the research project level dimension. The agent initiates a call request to the system through the encrypted and digitally signed HTTPS protocol, which explicitly contains the target domain agent identifier Hosp-Record-DB-01, the operation type of reading the anonymized medical record data of diabetic patients, and the above permission credential point (4, 8). After verifying the digital signature to ensure that the request comes from a legal source, the receiving module successfully parses these key elements and completes the initialization of the cross-domain mutual recognition process.

[0028] In a preferred embodiment of the present application, step 2, according to the spatial relationship characteristics of the source domain permission space and the target domain permission space involved in the call request, affine transformation parameter set is obtained through permission conversion model, and conic volume measurement evaluation and security verification are used to finally screen out affine transformation parameters suitable for permission credential mapping, including: Step 21, based on the source domain permission space and the target domain permission space involved in the call request, the spatial relationship characteristics between them are extracted, including permission dimension mapping rule, coordinate scaling ratio and space rotation angle, specifically including: To map the source domain permission coordinate points to the target domain permission space, the geometric transformation relationship between the two spaces needs to be established first. By querying the pre-set cross-domain space mapping strategy library, the permission space metadata of the source domain and the target domain are compared and analyzed for features and semantic relationships, thereby extracting three key feature parameters for defining the space affine transformation. Specifically: Permission dimension mapping rule: used to establish the semantic correspondence between the coordinate axis dimensions of the source domain permission space and the target domain permission space; by analyzing the metadata of the two domain permission spaces, based on the ontology or pre-defined semantic mapping table, a semantic bridge between the dimensions is established, for example, the data sensitivity level dimension of the source domain is mapped to the operation risk level dimension of the target domain, and the research project level dimension of the source domain is mapped to the accessible data range dimension of the target domain.

[0029] Coordinate scaling ratio: determined by calculating the value range ratio of the corresponding dimensions of the two spaces; for example, the data sensitivity level of the source domain takes a value range of [1, 5], and the operation risk level of the target domain takes a value range of [1, 10], so the scaling ratio in this dimension is 2.0.

[0030] Space rotation angle: determined by analyzing the angle between the basis vectors of the two space coordinate systems, used to correct the coordinate system direction deviation caused by differences in permission model design philosophy; for example, it is identified that there is a 30-degree orientation offset of the target domain coordinate system relative to the source domain coordinate system.

[0031] Step 21 in specific implementation, the system takes the permission coordinate points (4, 8) of Med-Research-AI in step 1, queries the mapping strategy library, determines the dimension mapping relationship, including the correspondence between the data sensitivity level and the target operation risk level, and the correspondence between the research project level and the target accessible data range; after analysis and calculation, the scaling ratio in the operation risk level dimension is 2.0, from the value range [1, 5] to [1, 10], and the space rotation angle is determined to be 30°.

[0032] Step 22, input the extracted space relationship features into the pre-trained permission conversion model for processing, and obtain the output affine transformation parameter set, specifically including: The space relationship features extracted in step 21, i.e. the dimension mapping rule, the scaling ratio 2.0 and the rotation angle 30°, are input as input vectors into the pre-trained permission conversion model; the model is trained based on a large number of historical mapping examples, and through the integrated mapping rule library and geometric calculation logic, a complete affine transformation parameter set is output, including the scaling matrix, the rotation matrix and the translation vector.

[0033] The above features are input into the permission conversion model, and the construction process of the permission conversion model is as follows: This process covers data collection, cleaning, standardization, and division. The training samples collected are successful cases of cross-domain permission mapping in various industries over the past five years, verified by manual review, with a cumulative number of at least 150,000 to ensure coverage of multiple business fields such as healthcare and different permission space configuration scenarios. Each sample includes complete permission space metadata of the source and target domains, including the definition of numerical ranges and semantic descriptions of each dimension, as well as manually confirmed spatial relationship characteristics such as dimension mapping rules, coordinate scaling ratios, spatial rotation angles, and actual verified affine transformation parameters such as scaling matrices, rotation matrices, and translation vectors.

[0034] In the data cleaning phase, the collected samples are verified multiple times. First, samples with missing metadata or incomplete parameters are removed to ensure that each sample has complete information. Second, contradictory samples are detected using pre-set logical rules, such as cases where the dimension ranges of the source and target domains do not match but the unreasonable scaling ratio is labeled. Such samples need to be manually reviewed or directly removed. Finally, statistical methods are used to identify outliers, such as samples with rotation angles exceeding the reasonable range of 0 to 60 degrees. The proportion of abnormal samples should be controlled within 3%.

[0035] In the data standardization phase, different types of features are processed accordingly. For continuous numerical features such as scaling ratios and rotation angles, the min-max method is used to map them to the interval of 0 to 1, eliminating the influence of dimension differences on model training. For semantic features such as dimension mapping rules, one-hot encoding is used to convert them into fixed-length binary vectors, enabling the model to effectively identify the corresponding relationships between different dimensions. For affine transformation parameters as labels, they need to be converted into a unified vector form, such as converting a two-dimensional scaling matrix into a vector containing two elements, directly representing the rotation angle as a numerical value, and converting the translation vector into a vector containing two elements. The dataset division needs to divide the processed samples into training, validation, and test sets in a 7:2:1 ratio. The training set is used for model parameter learning, the validation set is used for adjusting hyperparameters during training, and the test set is used for final model performance evaluation. When dividing, it is necessary to ensure that the proportion of samples in each business field in the three sets is consistent with the original sample set, avoiding the impact of uneven data distribution on the generalization ability of the model.

[0036] The model architecture uses a deep neural network with an overall structure consisting of an input layer feature fusion layer and multiple output layers. Each layer is designed to optimize the characteristics of permission space transformation. The number of neurons in the input layer is consistent with the length of the standardized feature vector. For example, when the feature vector contains scaling ratios, rotation angles, and dimension mapping rule encodings, a total of 6 elements, the input layer needs to have 6 neurons. Its main function is to perform a preliminary linear transformation on the input feature vector, mapping the features to a high-dimensional space to prepare for subsequent feature fusion.

[0037] The feature fusion layer includes three fully connected sub-layers for deep fusion and abstraction of input features to extract high-level feature representations; the first sub-layer is provided with 64 neurons and uses a ReLU activation function to enhance the non-linear expression capability of the model, the second sub-layer is provided with 32 neurons and also uses a ReLU activation function to further compress the feature dimension and extract more abstract features, and the third sub-layer is provided with 16 neurons and uses a LeakyReLU activation function to avoid the problem of neuron death, while a residual connection is added between the first sub-layer and the third sub-layer to directly superimpose the output of the first sub-layer on the output of the third sub-layer, alleviate the gradient decay problem of deep network, and ensure the integrity of feature transmission.

[0038] The multi-head output layer designs three independent output branches according to different types of affine transformation parameters, respectively generating a scaling matrix, a rotation matrix and a translation vector; the scaling matrix branch is provided with 2 neurons and uses a linear activation function, and the output result is processed by a ReLU function to ensure that the scaling ratio is non-negative, the rotation angle branch is provided with 1 neuron and uses a linear activation function, and the output result is range-constrained to ensure that it is between 0 and 60 degrees, and the translation vector branch is provided with 2 neurons and uses a linear activation function, directly outputting two components of the translation vector, and the outputs of the branches are combined to form a complete set of affine transformation parameters.

[0039] The model training adopts a supervised learning method, and the network parameters are optimized through multiple iterations. Before training, the network weights need to be initialized, the weights of the input layer and the feature fusion layer are initialized by the He method, the weights of the output layer are initialized by the Xavier method, and the bias terms are all initialized to 0. During the training process, the Adam optimizer is used to adjust the parameters, the initial learning rate is set to 0.001, when the validation set loss does not decrease for 5 consecutive rounds, the learning rate is automatically halved, and the minimum is reduced to 0.00001.

[0040] The loss function adopts a weighted mean square error, and the weights are set according to the importance of different parameters, wherein the weights of the scaling matrix and the rotation matrix are set to 1.0, and the weight of the translation vector is set to 0.8, in order to highlight the requirement for the accuracy of spatial scaling and rotation. During each training, the training set samples are first input into the network for forward calculation to obtain the predicted affine transformation parameters, and then the predicted value and the true value are substituted into the loss function to calculate the error, and then the error is transmitted from the output layer to the input layer through the back propagation algorithm, and the weights and biases are adjusted according to the error contribution of each layer of neurons to update the network parameters iteratively.

[0041] To prevent overfitting during training, in addition to using a validation set to monitor model performance, a dropout layer with a dropout rate of 0.2 is added after each sub-layer of the feature fusion layer to randomly discard some neurons and reduce the network's dependence on local features. L2 regularization is also used with a regularization coefficient of 0.0001 to limit the size of the weight parameters. When the validation set loss does not decrease for 10 consecutive rounds, the early stopping mechanism is triggered, and the training is stopped and the current optimal model parameters are saved.

[0042] After the model is trained, the overall performance is verified on an independent test set. The evaluation indicators include the mean absolute error and the root mean square error, which measure the average deviation and overall dispersion of the predicted parameters from the true parameters. For fields such as medical finance that require high precision, the mean absolute error should be controlled within 0.05, and the root mean square error should be controlled within 0.08. At the same time, the adaptability of the model in different business fields needs to be verified to ensure that the prediction error in each field is within the preset threshold. If the error in a certain field exceeds the threshold, additional samples need to be supplemented for incremental training in that field.

[0043] After the model is deployed, a dynamic updating mechanism needs to be established to regularly collect new cross-domain permission mapping success cases. Incremental training is performed on the model once a quarter to update the network parameters to adapt to the dynamic changes in the permission space, such as the expansion of the target domain permission dimension or the adjustment of the value range. After each update, the performance needs to be verified on the test set to ensure that the model always maintains high mapping accuracy.

[0044] Finally, in the model deployment and application stage, the trained model is packaged as a callable service module. This module receives the spatial relationship features extracted in step 21 as input and outputs the complete affine transformation parameter set after internal forward calculation. At the same time, this service also integrates version management, performance monitoring, and regular updating mechanisms to adapt to changes in security policies in different target domains and ensure the continuous provision of accurate and reliable spatial transformation parameter prediction services. After internal calculation, the affine transformation parameter set is output: scaling matrix [2.0, 0; 0, 1.2], rotation matrix [cos 30°, -sin 30°; sin 30°, cos 30°], and translation vector [1, 0.5].

[0045] In the specific implementation of step 22, the features input into the permission conversion model in step 21 are used to train the model using cross-domain access history data in the medical field. When the input contains data sensitivity to operation risk level mapping rules, scaling factor 2.0, and rotation angle 30°, etc. features, the affine transformation parameter set is output after internal calculation: scaling matrix [2.0, 0; 0, 1.2], rotation matrix [cos 30°, -sin 30°; sin 30°, cos 30°], and translation vector [1, 0.5].

[0046] Step 23, with the center of the target domain authority space as the vertex, the mapping direction defined by the affine transformation parameters as the cone axis direction, generate a cone covering the target domain authority space; calculate the actual intersection volume of the cone and the target domain authority space, and take the ratio of the volume value to the total volume of the target domain authority space as the cone volume metric value, which is used to quantitatively evaluate the coverage of the affine transformation parameters to the target domain authority space; based on the cone volume metric value, sort the parameters in the affine transformation parameter set in descending order, and select the parameter subset whose metric value is higher than the preset threshold; for the parameter subset, perform compatibility and risk verification according to the security policy of the target domain authority space to obtain the parameters that pass the verification; take the parameters that pass the verification as the final affine transformation parameters used for authority credential mapping, which specifically includes: Take the geometric center of the target domain authority space as the vertex of the cone (the center is calculated by averaging all the authority coordinate points predefined by the target domain, for example, the center coordinates of the target domain (cooperative hospital) authority space in the medical scenario are (5, 6)); take the mapping direction of the source domain authority point to the target domain authority point defined by a certain set of affine transformation parameters output in step 22 as the axis direction of the cone (i.e. the coordinate transformation vector direction corresponding to the parameter, for example, the mapping direction of a certain parameter set is the vector direction from the source domain (4, 8) to the target domain (10.24, 9.36)).

[0047] Based on the overall boundary of the target domain authority space (determined by the maximum and minimum values of the target domain authority dimensions, for example, in the medical scenario, the target domain operation risk level dimension range [1, 10] and the accessible data range dimension range [0, 15] form a rectangular boundary), adjust the radius and height of the base of the cone to ensure that the generated cone can completely cover the entire target domain authority space, avoiding incomplete coverage that may cause subsequent volume calculation to be missed.

[0048] Calculate the volume of the actual intersection part of the generated cone and the target domain authority space (usually a rectangular or polyhedral structure) (i.e. the part of the volume of the cone falling inside the target domain authority space); then calculate the total volume of the target domain authority space (obtained by multiplying the numerical range of each authority dimension, for example, in the medical scenario, the total volume of the target domain is the length of the operation risk level range multiplied by the length of the accessible data range, i.e. (10-1) x (15-0) = 135).

[0049] Divide the intersection volume of the cone and the target domain authority space by the total volume of the target domain authority space to obtain the ratio, which is the cone volume metric value corresponding to the set of affine transformation parameters. This value is used to quantitatively evaluate the coverage of the parameters to the target domain authority space. The higher the value, the better the integrity of the parameter in mapping the authority to cover the legal authority range of the target domain.

[0050] A volume metric threshold is preset (the threshold is set by the target domain according to business requirements, for example, in a medical scenario, to ensure that the permission covers the core data area, the threshold is set to 0.7); all affine transformation parameters output in step 22 are sorted in descending order according to their respective conical volume metrics; parameters with a metric value greater than or equal to the preset threshold are selected to form a high-coverage parameter subset, and parameters with insufficient coverage (such as a parameter with a metric value of 0.6, which is excluded because it is lower than the 0.7 threshold) are excluded.

[0051] According to all predefined security policy rules of the target domain (the rules are configured by the target domain administrator and are strongly associated with the permission dimension); taking a medical scenario (a cooperative hospital) as an example, the security policy rules include: the mapping result of the operation risk level dimension ≤ 8 (to avoid high-risk permission mapping), the mapping result of the accessible data range dimension > 0 (to exclude invalid mappings without data access permission), and specific constraints such as the anonymous medical record data access permission cannot be mapped to the complete medical record access permission.

[0052] To comprehensively verify the security of the parameters, representative feature points (covering different permission levels to ensure comprehensiveness) are selected from the source domain permission space, including: extremely high permission points of the source domain, such as the extremely high permission points (5, 9) of the source domain (research center) in a medical scenario, representing the highest data sensitivity and the highest research priority; medium permission points, such as (3, 6); the lowest permission points, such as (1, 2); for each group of parameters in the high-coverage parameter subset, these source domain feature points are respectively substituted into the affine transformation logic corresponding to the parameter, and the mapping results of each feature point in the target domain permission space are calculated, i.e., the permission mapping effect when the parameter is actually used is simulated.

[0053] The affine transformation logic specifically includes: based on preset affine transformation parameters (including scaling matrix, rotation matrix and translation vector), converting the coordinates of the feature points in the source domain authority space into coordinates recognizable by the target domain authority space, the specific implementation process is as follows: first, perform scaling transformation, express the source domain feature points as (x1, y1) (x1 corresponds to the first authority dimension of the source domain, y1 corresponds to the second authority dimension of the source domain), extract the diagonal elements a (scaling ratio of the first dimension of the source domain and the first dimension of the target domain) and d (scaling ratio of the second dimension of the source domain and the second dimension of the target domain) of the scaling matrix from the affine transformation parameter set, and obtain the scaled coordinates (x2, y2) through the calculation of x2=x1×a, y2=y1×d, to eliminate the difference in numerical range of the corresponding authority dimensions between the source domain and the target domain; then perform rotation transformation, extract the rotation matrix (elements are cosθ, -sinθ, sinθ, cosθ, wherein cosθ is the cosine value of θ, and sinθ is the sine value of θ) constructed based on the included angle θ between the coordinate system of the source domain and the target domain from the affine transformation parameter set, and substitute the scaled coordinates (x2, y2) into the calculation of x3=x2×cosθ-y2×sinθ, y3=x2×sinθ+y2×cosθ, to obtain the rotated coordinates (x3, y3), to correct the directional deviation of the coordinate system of the authority space between the source domain and the target domain.

[0054] Finally, perform translation transformation, extract the translation vector (e, f) from the affine transformation parameter set, e is the first dimension translation offset, f is the second dimension translation offset, substitute the rotated coordinates (x3, y3) into the calculation of x4=x3+e, y4=y3+f, to obtain the final coordinates (x4, y4) of the target domain, to realize the alignment of the coordinates and the reference range of the target domain authority space.

[0055] The simulation mapping results of all feature points of each set of parameters are compared and verified with the loaded target domain security policy rules one by one, if all simulation results of a set of parameters completely comply with all security policy rules (no rule is violated), it is determined that the parameter passes the security verification; if any one of the simulation results of a set of parameters violates any one of the security policy rules (for example, in the medical scene, a parameter maps the source domain high permission point (5, 9) to the target domain (11.0, 11.3), wherein the operation risk level 11.0 violates the rule of less than or equal to 8), it is determined that the parameter does not pass the security verification, and it is removed from the parameter subset. Taking the medical scene as an example, the initial parameter set is removed because the simulation result (11.0, 11.3) violates the operation risk level ≤8; the new parameter set (scaling matrix [1.6, 0; 0, 1.0], rotation matrix [cos30°, -sin30°; sin30°, cos30°], translation vector [1, 1]) output by the permission conversion model after re-computation, maps the source domain high permission point (5, 9) to (7.8, 10.5), wherein the operation risk level 7.8 ≤8, the accessible data range 10.5 >0, and other medical record access permission constraints are met, so it is determined that the parameter passes the security verification.

[0056] From the high-coverage parameter subset, all parameters that pass the security policy verification are selected; if there is only one set of parameters that pass the verification (such as the new parameter set in the medical scene), the set of parameters is directly determined as the final affine transformation parameter; if there are multiple sets of parameters that pass the verification, the conical volume metric values of the parameters are compared again, and the set with the highest metric value is selected as the final parameter (to ensure that the permission coverage is optimal under the premise of security compliance); the finally determined parameter will be used for formal mapping of the source domain original permission credential (such as (4, 8) in the medical scene) to the target domain permission space.

[0057] In a preferred embodiment of the present application, step 2, using the affine transformation parameter to map the original permission credential of the source domain agent from the source domain permission space to the target domain permission space, generates a target domain recognizable and verifiable mapped permission credential, comprising: Step 24, read the final affine transformation parameter that passes the verification, and parse it into a translation vector, a scaling matrix and a rotation matrix for coordinate transformation, specifically comprising: The final affine transformation parameter set outputted by step 23, which has passed the target domain security policy verification; the implementation process of this step is to call a special parameter parser, read the data structure of the parameter set, and according to the predefined parameter organization format, deconstruct and instantiate it into three independently operable mathematical transformation components: a scaling matrix that defines the scaling factor of each dimension, a rotation matrix that defines the rotation direction of the coordinate system, and a translation vector that defines the coordinate origin offset. Through this parsing process, the comprehensive parameter package outputted by step 23 is converted into a serialized basic geometric transformation operation unit that can be directly called by step 25.

[0058] Step 24, in specific implementation, takes the aforementioned medical collaboration case, the system reads the final affine transformation parameter set verified by step 23, and parses its specific composition: the scaling matrix is [1.6, 0; 0, 1.0], which means scaling 1.6 times in the operation risk level dimension and keeping the original value unchanged in the accessible data range dimension; the rotation matrix is [cos30°, -sin30°; sin30°, cos30°], corresponding to a 30-degree spatial rotation transformation; the translation vector is [1, 1], used for final offset adjustment of the coordinate position after transformation. At this point, all basic operation units required for mapping calculation are ready.

[0059] Step 25, based on the translation vector, scaling matrix and rotation matrix, the original permission credential geometric coordinate points are sequentially scaled, rotated and translated to obtain new coordinate points in the target domain permission space, which specifically includes: According to the strict geometric transformation sequence, the source domain original permission credential geometric coordinate points received in step 1 are sequentially subjected to the three basic geometric transformations parsed by step 24; first, apply the scaling matrix to the coordinate points for dimension scaling processing, multiply the coordinate point values of each dimension by the corresponding scaling factor in the scaling matrix, to realize the linear conversion of permission values from the source domain value range to the target domain value range; then, apply the rotation matrix to the scaled coordinate points for spatial rotation alignment operation, adjust the direction of the coordinate points in the multi-dimensional space through matrix operation, so that it is consistent with the coordinate system direction of the target domain permission space; finally, apply the translation vector to the rotation-aligned coordinate points for coordinate position offset, add the component values of the translation vector to the corresponding dimensions of the coordinate points, to finally complete the precise positioning of the coordinate points in the target domain permission space.

[0060] Step 25, in a specific implementation, according to the embodiment, the original permission coordinate point (4, 8) of Med-Research-AI is mapped to the target domain space. First, the scaling matrix [1.6, 0; 0, 1.0] is applied to process, and the intermediate result (6.4, 8.0) is obtained; then, the rotation matrix [cos 30°, -sin 30°; sin 30°, cos 30°] is applied to the intermediate result for spatial rotation alignment; finally, the translation vector [1, 1] is applied to the rotation result for offset, and the new coordinate position of the point in the target domain permission space is finally calculated as (7.8, 10.5), and thus the permission certificate of the source domain has been accurately converted to the target domain space in mathematics.

[0061] Step 26, the new coordinate point obtained is encapsulated according to the format of the target domain permission certificate, and a space position verification label is attached, to generate a mapped permission certificate that can be identified and verified by the target domain, specifically including: The new coordinate point calculated in step 25 is standardized and encapsulated according to the data structure and certificate specification required by the target domain permission system; specifically, the dimension values of the new coordinate point are filled into the corresponding fields of the target domain permission certificate template, and in addition, to ensure the verifiability and tamper resistance of the mapped certificate, the generation process also includes attaching a space position verification label to it; the label is calculated by a cryptographic hash function based on the values of the new coordinate point, a timestamp, and a key that can only be verified by the target domain permission center, and finally, a mapped permission certificate that is compliant in format, complete in content, and has verification information is output.

[0062] Step 26, in a specific implementation, the system obtains the mapped coordinate point (7.8, 10.5). Then, a permission certificate data structure conforming to the hospital domain specification is created, 7.8 is filled into the operation risk level field, and 10.5 is filled into the accessible data range field. At the same time, the system uses the coordinate point data, the current timestamp, and a secure key to generate a digital label through cryptographic hash operation, and attaches it to the certificate to form the final mapped permission certificate that can be identified by the target domain, which represents the equivalent permissions granted to Med-Research-AI in the hospital permission space.

[0063] In a preferred embodiment of the present application, step 3, based on the mapped permission certificate, in combination with the permission center point set of the intelligent agent in the target domain permission space, the target domain permission space is divided into multiple permission regions through a clustering algorithm, including: Step 31, obtaining the coordinate position of the mapped permission certificate in the target domain permission space, specifically including: Step 31, obtain the coordinate position of the mapped permission credential in the target domain permission space, which is the data preparation link of the permission area division; its implementation process is to parse and extract information from the standardized permission credential generated in step 26, the system calls the special credential parsing submodule, reads the complete data structure of the credential, which follows the permission credential format specification defined by the target domain, and the parsing submodule accurately extracts the numerical values representing the permission position by positioning the pre-defined coordinate field. These numerical values are the accurate geometric representation of the source domain agent permission in the target domain permission space after affine transformation; after extraction, the coordinate values are converted into geometric point data objects that can be processed internally by the system.

[0064] In the specific implementation of step 31, the aforementioned medical collaboration case is continued, and the target domain permission credential containing the mapped coordinate point (7.8, 10.5) is generated in step 26. The system parses the credential and extracts its coordinate position in the target domain permission space: the operation risk level dimension value is 7.8, and the accessible data range dimension value is 10.5, i.e. the coordinate point (7.8, 10.5).

[0065] Step 32, based on the coordinate position, load the target domain pre-defined permission center point set, wherein each permission center point corresponds to a permission level, specifically including: Step 32, based on the coordinate position, load the target domain pre-defined permission center point set, wherein each permission center point corresponds to a permission level. Its implementation process is that the system retrieves and loads the permission center point set pre-defined and configured by the domain administrator from the persistent storage system of the target domain. Each center point in the set has a fixed coordinate position in the target domain permission space and a unique correspondence with a specific permission level. These permission levels are defined according to the security policy and business requirements of the target domain, forming a complete classification system for all permission states in the target domain. During the loading process, the system verifies the integrity and validity of the center point data, and after completion, loads the entire set into the memory data structure as a stable reference system.

[0066] In the specific implementation of step 32, the system continues the aforementioned example and loads the pre-defined permission center point set from the configuration library of the cooperating hospital. The set contains three core points: permission center point A (2, 3), corresponding to the permission to access only public health statistical data; permission center point B (6, 7), corresponding to the permission to access de-identified medical record data; and permission center point C (8, 10), corresponding to the permission to access detailed medical record data, but requiring additional approval permission.

[0067] Step 33, using the loaded permission center point set as the clustering center, using the distance measurement algorithm to dynamically divide the target domain permission space, generating multiple convex polygon permission areas defined by boundary point sets, specifically including: Step 33, with the loaded set of authority center points as clustering centers, the target domain authority space is dynamically divided using distance measurement algorithm to generate a plurality of convex polygon authority regions defined by the boundary point set. The implementation process is to use the authority center point set loaded in step 32 as the fixed clustering center, use the space division algorithm such as Voronoi diagram division, and use the Euclidean distance as the measurement basis to divide the entire target domain authority space into several non-overlapping convex polygon sub-regions; Each sub-region corresponds to a unique authority center point, containing all points in the space closest to the center point, and the output result of the division algorithm is that each authority region is precisely defined by the boundary point set, thereby discretizing the continuous space into authority regions with clear geometric boundaries.

[0068] Step 33 in specific implementation, the system takes the three loaded authority center points A(2,3), B(6,7), C(8,10) as the core, performs space division calculation, and after the calculation is completed, the target domain authority space is divided into three convex polygon regions: region A (around center point A), region B (around center point B) and region C (around center point C). The shape and range of each region are defined by the boundary vertex coordinate set, for example, region B is surrounded by its boundary point set {(4,5), (5,6), (7,8), (6,9)}.

[0069] Step 34, according to the divided authority region, the spatial distance between the coordinate position of the mapped authority credential and each authority center point is calculated, specifically including: Step 34, according to the divided authority region, the spatial distance between the coordinate position of the mapped authority credential and each authority center point is calculated; The implementation process is to obtain the mapped authority credential coordinate point obtained in step 31, and calculate the geometric spatial distance between the point and each authority center point loaded in step 32. The distance measurement is based on the principle of multidimensional Euclidean distance, and the square root of the sum of the coordinate value differences of the two points in each dimension is calculated to obtain a scalar distance value, which provides a comparable quantitative index for the proximity between the authority credential coordinate point and each authority level.

[0070] Step 34 in specific implementation, the system calculates the distance from the mapping point (7.8, 10.5) to the three authority center points. The distance d1 from point A(2,3), the distance d2 from point B(6,7), and the distance d3 from point C(8,10) are calculated. The results show that the value of distance d2 is less than d1 and d3, which means that point (7.8, 10.5) is closest to center point B representing the authority to access de-identified medical record data in geometric space.

[0071] Step 35, according to the calculated spatial distance, the mapped permission credential is attributed to the target permission area corresponding to the permission center point with the minimum distance, specifically comprising: Step 35, according to the calculated spatial distance, the mapped permission credential is attributed to the target permission area corresponding to the permission center point with the minimum distance; the implementation process is that all spatial distance results calculated in step 34 are received and compared, the smallest distance value is selected by comparison operation, and the corresponding permission center point is recorded; according to the corresponding relationship between the permission center point and the permission area established in step 33, the mapped permission credential is finally classified into the target permission area of the permission center point corresponding to the minimum distance, completing the positioning and classification of the source domain agent permission in the target domain space.

[0072] In the specific implementation of step 35, the previous step is connected, since the distance d2 from the mapping point (7.8, 10.5) to the center point B (6, 7) is the smallest, the system determines that the credential should be attributed to the permission area corresponding to the center point B, that is, area B, which can access the de-identified medical record data area; this attribution result means that the permission level of the source domain agent Med-Research-AI mapped and recognized in the target domain hospital is to access the de-identified medical record data.

[0073] In a preferred embodiment of the present application, the target permission area to which the mapped permission credential belongs is determined, and the permission range polygon of the source domain agent mapped in the target domain is generated according to the boundary point set of the target permission area, comprising: Step 36, obtaining the target permission area identifier corresponding to the mapped permission credential, specifically comprising: Obtaining the target permission area identifier corresponding to the mapped permission credential is an intermediate link connecting the permission area attribution determination and the geometric structure construction; the implementation process is that the system receives the permission area attribution determination result generated in step 35 through the internal data interface, and the result data object clearly contains the unique identifier of the target permission area to which the mapped permission credential is attributed; the identifier is a symbol code corresponding one by one to the permission center point loaded in step 32.

[0074] In the specific implementation of step 36, the previous medical collaboration case is connected, step 35 determines that the mapped credential should be attributed to area B, and the system obtains the determination result, that is, the target permission area identifier Region_B, which uniquely corresponds to the permission center point B (6, 7) and the permission level of accessing de-identified medical record data represented by it.

[0075] Step 37, based on the target permission area identifier, retrieving the boundary point set corresponding to the target permission area, specifically comprising: Based on the target permission area identifier, the corresponding boundary point set is retrieved, and the geometric data basis for the construction of the permission range polygon is provided; the implementation process is to take the target permission area identifier obtained in step 36 as a query key, and retrieve the permission area metadata database generated and stored after the space division in step 33; the system quickly locates the geometric definition data of the corresponding permission area according to the identifier, and reads the boundary point set defining the spatial range of the area, which is an ordered set containing multiple coordinate points, and these points are sequentially connected to outline the geometric contour of the target permission area.

[0076] In the specific implementation of step 37, the system takes the region identifier Region_B as an index to query the permission area geometric information library generated and stored in step 33, and after successful retrieval, reads the boundary point set of region B: {(4, 5), (5, 6), (7, 8), (6, 9)}, which accurately defines the geometric range of region B in the target domain permission space.

[0077] In step 38, the retrieved boundary point set is used to construct a closed convex polygon geometric structure according to the spatial order, which specifically includes: In step 38, the retrieved boundary point set is used to construct a closed convex polygon geometric structure according to the spatial order; the implementation process is that the system receives the boundary point set retrieved in step 37 and reconstructs it geometrically. First, according to the spatial coordinate relationship of the boundary points, a spatial sorting algorithm such as polar angle sorting is used to arrange the point set into an ordered sequence that can form a closed loop in space, and then according to this order, the adjacent boundary points are connected with straight line segments, and the first and last points are connected, finally a closed, single-connected convex polygon geometric structure is constructed, which represents the geometric range of the source domain agent allowed to exercise permission in the target domain in mathematics.

[0078] In the specific implementation of step 38, after obtaining the boundary point set {(4, 5), (5, 6), (7, 8), (6, 9)} of region B, the system determines the connection order as (4, 5), (5, 6), (7, 8), (6, 9), (4, 5), and connects the points in this order to construct a quadrilateral convex polygon, which is the specific geometric range of the permission of the de-sensitization processed medical record data in the target domain space.

[0079] In step 39, the convex polygon geometric structure is topologically verified, and the polygon that passes the verification is bound to the mapped permission credential as the permission range polygon of the source domain agent in the target domain, which specifically includes: Step 39, topological check is performed on the convex polygon geometry, and the polygon passing the check is bound to the mapped authority credential as the authority range polygon; the implementation process is that, strict topological consistency check is performed on the convex polygon constructed in step 38, including verifying whether the polygon is a closed structure, all sides are non-intersecting, for example, a simple polygon, whether it is a convex polygon, and the directionality of the vertex sequence, only when all topological checks are confirmed to be valid geometric regions, the system binds and associates it with the mapped authority credential generated in step 26, and the implementation is realized by adding a geometry attribute field to the credential data structure and storing the polygon structure.

[0080] In the specific implementation of step 39, the system performs topological check on the constructed quadrilateral convex polygon, and after the check passes, the polygon structure is attached to the mapped authority credential of Med-Research-AI as a new attribute, which is formally defined as the authority range polygon of the credential in the target domain, and thereafter, any operation authority verification of the agent in the hospital domain will be judged based on the quadrilateral area as the geometric boundary.

[0081] In a preferred embodiment of the present application, step 4, according to the corresponding operation in the call request, the corresponding operation point in the target domain authority space is obtained, and based on the authority range polygon, a ray method geometric inclusion detection algorithm is used to judge whether the operation point is located inside the authority range polygon, if it is determined that the operation point is located inside, the corresponding operation is allowed to be executed, and the call request is forwarded to the target domain agent for execution, including: Step 41, the operation description information in the call request is parsed to extract the corresponding operation point coordinates in the target domain authority space, specifically including: Parsing the operation description information in the call request to extract the corresponding operation point coordinates in the target domain authority space is a data preparation link for implementing operation authority verification; the implementation process is that, the original call request received and stored in step 1 is parsed again, and the operation description information therein is mainly processed, the system calls the operation semantic parser to identify and understand the specific operation type described in the request, and according to the operation-authority coordinate mapping rule pre-stored in the target domain system, the abstract operation type is converted into a geometric point with explicit coordinate value in the target domain authority space, which quantitatively represents the minimum authority level required to execute this operation or the specific position in the authority space.

[0082] Step 41, in specific implementation, takes over the medical collaboration example of steps 1-3, the system parses the calling request issued by Med-Research-AI, identifies that the operation it requests is to read the anonymized medical record data of the diabetic patient; according to the mapping rule predefined in the hospital domain, i.e. the target domain, the operation is converted into a fixed operation point in the permission space, whose coordinates are defined as (5.5, 6.5), where the operation risk level dimension value 5.5 represents that the operation has a medium-high risk level, and the accessible data range dimension value 6.5 represents that a large range of desensitized data needs to be accessed.

[0083] Step 42, based on the bound permission range polygon structure, reads its boundary vertex sequence, specifically including: Based on the bound permission range polygon structure, reading its boundary vertex sequence is the key to constructing the geometric detection environment; its implementation process is to obtain its bound geometric attribute information from the mapping permission credential processed in step 39, the system accesses the data structure of the credential, locates and reads the convex polygon geometric structure representing its permission range, and then extracts the ordered coordinate sequence of the boundary vertices from the polygon structure; this vertex sequence accurately defines a closed convex polygon region, which geometrically represents the effective permission range granted to the source domain agent in the target domain.

[0084] Step 42, in specific implementation, the system obtains the boundary of the permission range polygon bound in step 39 from the mapping permission credential of Med-Research-AI, i.e. the boundary of region B, the system reads the boundary vertex sequence of the polygon, for example, the four point coordinates stored in order: [(4, 5), (5, 6), (7, 8), (6, 9)]. This sequence defines the geometric boundary of permission verification.

[0085] Step 43, using the operation point coordinates, emits an infinitely extended detection ray along the preset direction of the target domain permission space, specifically including: Using the operation point coordinates to emit an infinitely extended detection ray along the preset direction of the target domain permission space is the starting action of the geometric containment detection algorithm; its implementation process is to simulate an infinitely long ray emanating from the operation point to judge the positional relationship between the operation point and the permission polygon. The direction of the ray is preset to be parallel to the positive direction of a coordinate axis of the permission space, such as horizontally to the right, to simplify the subsequent intersection point calculation logic.

[0086] Step 43, in specific implementation, the system takes the operation point (5.5, 6.5) as the starting point, emits an infinitely extended detection ray along the positive direction of the X-axis of the target domain permission space, horizontally to the right.

[0087] Step 44, according to the detection ray and the boundary vertex sequence of the authority range polygon, calculate all intersection point coordinates, specifically including: According to the detection ray and the boundary vertex sequence of the authority range polygon, calculate all intersection point coordinates, is the core calculation link of the ray method; its implementation process is that, based on the principle of computational geometry, traverse the polygon boundary vertex sequence obtained in step 42, process each edge of the polygon in turn, the system judges whether the detection ray intersects the current edge, if it intersects, the accurate coordinate value of the intersection point is solved according to the geometric relationship, after the traversal is completed, the coordinate set containing all effective intersection points is obtained.

[0088] Step 44 in the specific implementation, the system calculates the intersection between the horizontal right ray and the four edges defined by the vertex sequence [(4, 5), (5, 6), (7, 8), (6, 9)], and calculates that the ray intersects with two edges of the polygon, and the coordinates of the two intersection points are (6.2, 6.5) and (7.5, 6.5) respectively.

[0089] Step 45, based on the obtained intersection point coordinates, count the number of effective intersection points, if it is odd, determine that the operation point is located inside the authority range polygon, if it is even, determine that it is located outside, specifically including: Based on the obtained intersection point coordinates, count the number of effective intersection points, if it is odd, determine that the operation point is located inside the authority range polygon, if it is even, determine that it is located outside, is the decision link of making position relationship judgment; its implementation process is that, first, all intersection points calculated in step 44 are verified for effectiveness, excluding repeated intersection points or tangential cases coinciding with the polygon vertex, then the total number of intersection points after effectiveness filtering is counted, finally the core odd-even judgment rule of the ray method is applied: if the total number of intersection points is odd, it is determined that the operation point is located inside the authority range polygon; if it is even, including zero, it is determined to be located outside.

[0090] Step 45 in the specific implementation, the system verifies that the two intersection points obtained in step 44 are effective intersection points. The total number of intersection points is 2, which is even, according to the odd-even judgment rule, it is determined that the operation point (5.5, 6.5) is located outside the authority range polygon.

[0091] Step 46, if the operation point is located inside the authority range polygon, generate an evaluation result allowing the corresponding operation to be executed, specifically including: If the operation point is located inside the authority range polygon, an evaluation result allowing the corresponding operation to be performed is generated, which is a key link for generating business logic according to a geometric judgment result; the implementation process is that the system receives the position judgment conclusion output in step 45, and only when the judgment result is that the operation point is located inside the authority range polygon, a positive evaluation result allowing execution is generated; the evaluation result is a structured data object, which contains final judgment, judgment basis, information based on geometric containment detection, timestamp and the like.

[0092] Step 46, in specific implementation, because the operation point is determined by step 45 to be located outside the authority range, the system does not generate an evaluation result allowing execution.

[0093] Step 47, the evaluation result is attached to the calling request and forwarded to the target domain agent to trigger the execution of the corresponding operation, specifically including: Attaching the evaluation result to the calling request and forwarding it to the target domain agent to trigger the execution of the corresponding operation is the final output link of the entire authority mutual recognition process; the implementation process is that the system encapsulates and integrates the execution-allowing evaluation result generated in step 46 and the original calling request received in step 1 to form an enhanced, authority-attached new request message. Then the new request is forwarded to the target domain agent through a secure communication channel, and the target domain agent directly triggers the execution logic of the corresponding operation based on the authority evaluation result attached, without the need for complex authority verification again.

[0094] Step 47, in specific implementation, assuming that a certain operation point is detected to be located inside, the system generates an execution-allowing evaluation result; the system attaches the result to the original reading medical record data calling request and forwards it to the target domain agent Hosp-Record-DB. After verifying the validity of the evaluation result, Hosp-Record-DB executes the reading operation and returns the anonymous diabetes medical record data to the source domain agent Med-Research-AI.

[0095] As shown in Figure 2 , the embodiment of the application further provides a cross-domain agent authority trusted mutual recognition system, which comprises: A receiving module is configured to receive a calling request from a source domain agent to a target domain agent for performing a corresponding operation; A generating module is configured to obtain an affine transformation parameter set through an authority conversion model according to spatial relationship characteristics of a source domain authority space and a target domain authority space involved in the calling request, and finally screen out an affine transformation parameter suitable for authority credential mapping by using conic volume measurement evaluation and security verification, and map the original authority credential of the source domain agent from the source domain authority space to the target domain authority space by using the affine transformation parameter to generate a mapped authority credential recognizable and verifiable by the target domain. a partition module configured to divide the target domain authority space into a plurality of authority regions based on the mapped authority credential and a set of authority center points of agents in the target domain authority space by a clustering algorithm, determine a target authority region to which the mapped authority credential belongs, and generate a polygon of the authority range of the agent in the source domain mapped in the target domain according to a set of boundary points of the target authority region; a processing module configured to acquire a corresponding operation point in the target domain authority space according to a corresponding operation in the invocation request, determine whether the operation point is inside the polygon of the authority range by a ray method geometric inclusion detection algorithm based on the polygon of the authority range, and if the operation point is determined to be inside, allow the corresponding operation to be performed and forward the invocation request to the agent in the target domain for execution.

[0096] Embodiments of the present application also provide a computing device, comprising a processor and a memory storing a computer program, the computer program being executed by the processor to perform the method described above. All implementation manners in the above method embodiments are applicable to this embodiment and can also achieve the same technical effects.

[0097] Embodiments of the present application also provide a computer-readable storage medium storing instructions, which, when executed on a computer, cause the computer to perform the method described above. All implementation manners in the above method embodiments are applicable to this embodiment and can also achieve the same technical effects.

[0098] The present technology (including methods, devices and systems) only provides a technical solution framework for cross-domain agent authority mutual recognition, and does not involve active collection, storage or additional processing of user privacy data. Privacy data such as medical records involved in the implementation process of the technology must be strictly followed by the specific application party in accordance with laws, regulations and industry privacy protection standards to ensure that the data source is legal and the use is compliant, and the technology does not bear the legal liability caused by illegal data acquisition or use.

[0099] The authority mapping and verification effect of the present technology depends on the accuracy of the source domain / target domain authority space metadata provided by the application party, the rationality of the security policy configuration and the quality of the basic data. The present technology does not bear the relevant responsibilities for the authority mutual recognition errors, operation risks and other problems caused by the application party metadata errors, improper policy configuration or data distortion.

[0100] The present technology does not provide additional guarantees for the compliance of the business process of a specific industry (such as medical care and finance), and the application party needs to adapt and adjust the technical solution according to its own industry regulatory requirements to ensure that the overall business process complies with industry standards, and the present technology does not bear the responsibility for the problems caused by the failure to adapt to the industry compliance requirements.

[0101] The above is the preferred embodiment of the present application, it should be noted that for those skilled in the art, without departing from the principles described in the present application, can also be made several improvements and refinements, these improvements and refinements should also be considered the scope of protection of the present application.

Claims

1. A method for trusted mutual recognition of permissions among cross-domain intelligent agents, characterized in that, The method includes: Receive a request from a source domain agent to perform a corresponding operation on a target domain agent; Based on the spatial relationship characteristics between the source domain permission space and the target domain permission space involved in the call request, an affine transformation parameter set is obtained through a permission transformation model. Then, the cone volume metric is used for evaluation and security verification. Finally, the affine transformation parameters suitable for permission credential mapping are selected. The original permission credential of the source domain agent is mapped from the source domain permission space to the target domain permission space using the affine transformation parameters, generating a mapped permission credential that can be identified and verified in the target domain. Based on the mapped permission credentials, and combined with the set of permission center points of the agents in the target domain permission space, the target domain permission space is divided into multiple permission regions by a clustering algorithm. The target permission region to which the mapped permission credentials belong is determined, and the permission range polygon of the source domain agent in the target domain is generated based on the boundary point set of the target permission region. Based on the corresponding operation in the call request, obtain the corresponding operation point in the target domain permission space. Based on the permission range polygon, use the ray-matrix geometric inclusion detection algorithm to determine whether the operation point is located inside the permission range polygon. If the operation point is determined to be inside, allow the execution of the corresponding operation and forward the call request to the target domain agent for execution.

2. The method for trusted mutual recognition of permissions among cross-domain intelligent agents according to claim 1, characterized in that, The invocation request contains the original permission credentials of the source domain agent. The original permission credentials are represented as geometric coordinate points in the source domain permission space and are used for affine transformation and cross-domain mapping.

3. The method for trusted mutual recognition of permissions among cross-domain intelligent agents according to claim 2, characterized in that, Based on the spatial relationship characteristics between the source domain permission space and the target domain permission space involved in the invocation request, an affine transformation parameter set is obtained through a permission transformation model. Then, using a cone volume metric for evaluation and security verification, affine transformation parameters suitable for permission credential mapping are finally selected, including: Based on the source domain permission space and target domain permission space involved in the call request, the spatial relationship features between the two are extracted. The spatial relationship features include permission dimension mapping rules, coordinate scaling ratio and spatial rotation angle. The extracted spatial relationship features are input into a pre-trained permission conversion model for processing to obtain the output affine transformation parameter set. Using the center of the target domain permission space as the vertex and the mapping direction defined by the affine transformation parameters as the cone axis direction, a cone covering the target domain permission space is generated; the actual intersection volume of the cone and the target domain permission space is calculated, and the ratio of this volume value to the total volume of the target domain permission space is used as the cone volume metric value to quantitatively evaluate the degree of coverage of the target domain permission space by the affine transformation parameters; The parameters in the affine transformation parameter set are sorted in descending order based on the cone volume metric, and a subset of parameters with metric values ​​higher than a preset threshold are selected. This subset of parameters is then subjected to compatibility and risk verification according to the security policy of the target domain permission space to obtain the parameters that pass the verification. The parameters that pass the verification are used as the final affine transformation parameters for permission credential mapping.

4. The method for trusted mutual recognition of permissions among cross-domain intelligent agents according to claim 3, characterized in that, The original authorization credentials of the source domain agent are mapped from the source domain authorization space to the target domain authorization space using the affine transformation parameters, generating a mapped authorization credential that can be recognized and verified in the target domain, including: Read the final affine transformation parameters that have passed verification and parse them into translation vectors, scaling matrices, and rotation matrices for coordinate transformation; Based on translation vectors, scaling matrices, and rotation matrices, the geometric coordinates of the original permission credential are sequentially scaled, rotated, and translated to obtain new coordinates in the target domain permission space. The obtained new coordinate points are encapsulated according to the format of the target domain permission credential, and a spatial location verification label is attached to generate a mapped permission credential that can be identified and verified by the target domain.

5. The method for trusted mutual recognition of permissions among cross-domain intelligent agents according to claim 4, characterized in that, Based on the mapped permission credentials, and combined with the set of permission center points of agents in the target domain permission space, the target domain permission space is divided into multiple permission regions using a clustering algorithm, including: Obtain the coordinates of the mapped permission credential in the target domain permission space; Based on the coordinate positions, load a predefined set of permission center points for the target domain, where each permission center point corresponds to a permission level; Using the loaded set of permission center points as cluster centers, a distance metric algorithm is used to dynamically divide the target domain permission space, generating multiple convex polygon permission regions defined by boundary point sets; Based on the divided permission areas, calculate the spatial distance between the coordinates of the mapped permission credential and the center point of each permission; Based on the calculated spatial distance, the mapped permission credentials are assigned to the target permission area corresponding to the permission center point with the smallest distance.

6. The method for trusted mutual recognition of permissions among cross-domain intelligent agents according to claim 5, characterized in that, Determine the target permission region to which the mapped permission credential belongs, and generate a polygon representing the permission range of the source domain agent in the target domain based on the boundary point set of the target permission region, including: Obtain the target permission region identifier corresponding to the mapped permission credential; Based on the target permission region identifier, retrieve the set of boundary points corresponding to the target permission region; Using the retrieved set of boundary points, a closed convex polygon geometry is constructed by connecting them in spatial order. A topological verification is performed on the convex polygon geometry, and the polygon that passes the verification is used as the permission scope polygon of the source domain agent in the target domain and bound to the mapped permission credential.

7. The method for trusted mutual recognition of permissions among cross-domain intelligent agents according to claim 6, characterized in that, Based on the corresponding operation in the call request, obtain its corresponding operation point in the target domain permission space. Based on the permission range polygon, use a ray-mapping geometric inclusion detection algorithm to determine whether the operation point is located inside the permission range polygon. If the operation point is determined to be inside, allow the execution of the corresponding operation, and forward the call request to the target domain agent for execution, including: Parse the operation description information in the call request and extract the coordinates of the corresponding operation point in the target domain permission space; Based on the bounded permission-scoped polygon structure, read its boundary vertex sequence; Using the coordinates of the operation point, an infinitely extending detection ray is emitted along a preset direction in the target domain permission space; Calculate the coordinates of all intersection points based on the sequence of boundary vertices of the detected ray and the polygon within the permission range; Based on the obtained intersection coordinates, count the number of valid intersections. If the number is odd, the operation point is determined to be inside the permission range polygon; if the number is even, it is determined to be outside. If the operation point is located inside the permission scope polygon, an evaluation result is generated that allows the corresponding operation to be executed; The evaluation results are appended to the invocation request and forwarded to the target domain agent to trigger the corresponding operation.

8. A cross-domain intelligent agent authorization trust mutual recognition system, the system implementing the method as described in any one of claims 1 to 7, characterized in that, include: The receiving module is used to receive call requests from the source domain agent to perform corresponding operations on the target domain agent; The generation module is used to obtain a set of affine transformation parameters through a permission transformation model based on the spatial relationship characteristics between the source domain permission space and the target domain permission space involved in the call request, and to evaluate and verify the security using a cone volume metric. Finally, it selects affine transformation parameters suitable for permission credential mapping, and uses the affine transformation parameters to map the original permission credential of the source domain agent from the source domain permission space to the target domain permission space, generating a mapped permission credential that can be identified and verified in the target domain. The partitioning module is used to divide the target domain permission space into multiple permission regions based on the mapped permission credential and the set of permission center points of the agent in the target domain permission space through a clustering algorithm, determine the target permission region to which the mapped permission credential belongs, and generate a polygon of the permission range of the source domain agent in the target domain based on the boundary point set of the target permission region. The processing module is used to obtain the corresponding operation point in the target domain permission space according to the corresponding operation in the call request, and use the ray-matrix geometric inclusion detection algorithm to determine whether the operation point is located inside the permission range polygon. If the operation point is determined to be inside, the corresponding operation is allowed to be executed, and the call request is forwarded to the target domain agent for execution.

9. A computing device, characterized in that, include: One or more processors; A storage device for storing one or more programs, which, when executed by one or more processors, cause the one or more processors to implement the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a program that, when executed by a processor, implements the method as described in any one of claims 1 to 7.

Images