A heterogeneous network node multi-level trust management evaluation method and system

Abstract

The application discloses a kind of heterogeneous network node multilevel trust management evaluation method and system.The method obtains node interaction record and context information, constructs the scene six-tuple containing interaction source subject, target subject, interaction type, interaction target, governance rule and privacy level;Based on the scene six-tuple, interaction record and trust weight, the context-aware trust graph that supports node multi-scene multi-edge representation is constructed;After receiving target trust evaluation request, combine scene label, maximum hop count and privacy permission level to filter effective trust path;Comprehensive trust result is calculated using trust propagation, trust synthesis and cross-scene attenuation operator;And according to the evaluation result of security threshold and privacy policy, execute release, empowerment, degradation verification or access interception.The application can reduce the semantic drift, overreach and privacy leakage risk caused by cross-scene trust direct reuse, improve the accuracy, interpretability and security control ability of general evaluation management of blockchain trust mechanism.

Description

Technical Field

[0001] This invention belongs to the field of network security and distributed identity trust management technology, specifically involving a multi-level trust management and evaluation method and system for heterogeneous network nodes. Background Technology

[0002] With the development of technologies such as distributed systems, Web 3.0, blockchain networks, cloud-edge collaboration, and cross-domain service calls, the types of entities, interaction relationships, and business scenarios in networks are becoming increasingly complex, covering various scenarios including e-commerce, government services, enterprise cooperation, financial services, and daily social interactions. In heterogeneous networks where centralized trusted third-party verification capabilities are limited or cross-domain verification costs are high, how to dynamically assess node trust based on historical interaction records, relationship links, and business context is a crucial technical issue for ensuring distributed identity authentication, access control, and reliable interaction. Therefore, trust assessment technologies for heterogeneous networks have significant application value in areas such as distributed identity authentication, cross-domain authorization, on-chain interaction risk control, privacy protection, and permission management.

[0003] Among existing trust assessment methods, a common approach is to calculate trust based on network topology, global reputation, or probability propagation. For example, the PageRank graph ranking method typically calculates node importance based on node connectivity and global topology, and this can be used for reputation or trust evaluation. Methods such as EigenTrust aggregate local trust evaluations to obtain a global trust value. Other methods use trust propagation, probability multiplication, or matrix iteration to calculate indirect trust on multi-hop paths. In recent years, trust prediction methods based on deep learning or graph neural networks have also been used to model trust relationships in complex networks. While these methods can reflect the topological relationships, historical interactions, or potential relationship characteristics between nodes to some extent, there is still room for improvement when dealing with heterogeneous interactive networks with multiple service scenarios, multiple privacy levels, and various governance rules.

[0004] In real-world heterogeneous networks, trust relationships are not solely determined by node connectivity; they are also influenced by business scenarios and interaction contexts. For example, a node may gain a high trust rating in a business-to-consumer (B2C) e-commerce scenario due to its high transaction fulfillment rate, but this does not necessarily mean it should have the same level of trust in a government service (G2C) scenario or a business-to-business (B2B) scenario involving confidential data. Without a unified modeling of interaction types, objectives, governance rules, and privacy levels, high trust values ​​formed in scenarios with low security requirements or low privacy levels may be unreasonably transferred and reused in scenarios with high security requirements or high privacy levels, thereby increasing the risks of privilege escalation, trust abuse, and privacy breaches.

[0005] In distributed trust management, trust assessment between nodes often requires indirect transmission along multi-hop paths. Existing methods for handling trust links mostly employ path weight accumulation, probability multiplication, global reputation iteration, or matrix calculation. These methods can reflect the impact of link length or topology on trust transmission, but they typically struggle to clearly quantify the semantic differences in trust across different business scenarios. For example, when trust evidence migrates from a public community interaction scenario (C2C) to a financial services scenario (B2C) or a government services scenario (G2C), due to differences in interaction goals, governance rules, and privacy levels, the trust assessment results usually need to be attenuated or constrained accordingly. Without a cross-scenario trust attenuation mechanism, cross-domain trust scores may be inflated, failing to meet the security expectations of the target scenario.

[0006] Meanwhile, while some context-aware trust management or access control methods introduce contextual attributes such as time, location, service type, and device status, they are mostly designed for specific network environments or specific business scenarios, lacking a unified graph structure representation and composable trust computation mechanisms. When multiple types of interaction relationships exist simultaneously in the system, such as consumer-to-consumer (C2C), business-to-consumer (B2C), business-to-business (B2B), government-to-business (G2B), and government-to-consumer (G2C), existing methods struggle to isolate and store trust evidence across multiple scenarios, perform path propagation, evidence synthesis, and privacy constraint checks within the same model.

[0007] Furthermore, while some trust prediction methods based on deep learning or graph neural networks can learn node relationship features from large amounts of interaction data, they are typically highly dependent on the quality of training data, sample size, and model parameters. In high-security scenarios such as government services, financial services, and distributed identity authentication, trust assessment results not only need to be accurate but also require a traceable, auditable, and verifiable computational process. If the model cannot clearly define the source path of the trust value, the scenario basis, and the process of privacy constraint enforcement, its applicability in security control and compliance governance scenarios will be limited.

[0008] From an engineering perspective, trust relationships in heterogeneous networks are dynamically changing. New business scenarios, new node interaction records, and new trust edges may continuously be added to the system. Existing solutions that rely on global iteration or a single global trust graph may need to recalculate a large number of node trust values ​​when scenarios expand or trust edges are updated, affecting real-time evaluation efficiency and potentially causing new scenarios to interfere with existing trust relationships. Furthermore, if privacy and permission levels are not considered during the path selection phase, path evidence that does not meet privacy requirements may enter the subsequent trust calculation process, reducing the reliability of trust evaluation results in security control. Summary of the Invention

[0009] To address the shortcomings of existing technologies, the present invention aims to provide a multi-level trust management and evaluation method and system for heterogeneous network nodes, thereby overcoming the aforementioned deficiencies in existing technologies.

[0010] To achieve the above objectives, the present invention provides the following technical solution:

[0011] A multi-level trust management evaluation method for heterogeneous network nodes includes:

[0012] Scene modeling steps: Obtain the historical interaction records of nodes in the heterogeneous network and their business context information, and construct a scene six-tuple containing the interaction subject and multi-dimensional scene semantic attributes accordingly;

[0013] The graph construction steps are as follows: Based on the scenario six-tuple, the node historical interaction records and the corresponding trust weights, a context-aware trust graph reflecting the multi-scenario association between nodes is constructed. For the interaction of the same source node and target node in different business scenarios, it is mapped as independent connection edges with different scenario labels.

[0014] Path filtering step: Receive a trust evaluation request containing the source node to be evaluated, the target node, and the target evaluation scenario. Within the context-aware trust graph, combine the scenario labels of the connecting edges, the path hop limit, and the privacy permission conditions to search and filter out the valid trust paths from the source node to the target node.

[0015] Trust calculation steps: The trust evidence corresponding to the effective trust path is quantitatively calculated using a preset set of trust evaluation operators. The calculation process includes accumulated trust transfer calculation along a single path, trust evidence fusion calculation for multiple paths or multiple source scenarios, and cross-scenario trust decay calculation based on the semantic attribute differences between the source scenario and the target scenario, thereby obtaining the comprehensive trust result under the target evaluation scenario.

[0016] Security control steps: The calculated comprehensive trust result is strictly matched and verified with the security access threshold and privacy constraint policy corresponding to the target evaluation scenario to generate a node evaluation result for the source node, and the corresponding access authorization, downgrade verification or access interception control actions are executed based on the evaluation result.

[0017] Preferably, in the trust calculation step, the specific process of cross-scenario trust decay calculation is as follows: compare the interaction type, interaction target, governance rules and privacy level attributes in the six-tuples corresponding to the source scenario and the target scenario, and calculate the semantic scenario distance between the two scenarios;

[0018] For categorical attributes, the degree of local attribute difference is determined based on whether the attribute classification is consistent or a preset semantic difference matrix; for privacy level attributes, the degree of local attribute difference is determined based on a preset normalized sequence level difference.

[0019] After obtaining the semantic scene distance by weighted summation and normalization of the local differences, the target attenuation coefficient is calculated using a monotonically non-increasing attenuation function or a first-order approximation function of the attenuation function, and then the attenuation coefficient is mapped to the trust evidence to be transmitted.

[0020] Preferably, in the graph construction step: the context-aware trust graph includes a node set, a context-aware connection edge set, a trust weight mapping function, and a scene label mapping function. The context-aware connection edge is composed of a source node, a target node, a scene six-tuple, and a trust weight. When an interaction record with the same source node and target node but different scene six-tuples occurs in the network, the combination of the source node, target node, and scene six-tuple is used as the retrieval condition to establish multiple context-aware connection edges with different scene labels for the node pair in the context-aware trust graph.

[0021] Preferably, in the path filtering step: the trust assessment request further includes a set of source scenarios for extracting initial trust evidence, and the target assessment scenario is used to determine the final business scenario to which the node assessment result applies and its corresponding privacy constraint strategy; the system searches for candidate paths from the source node to the target node within a preset maximum hop count range, and performs preliminary filtering on the candidate paths based on the scenario labels of the connecting edges; the privacy level attribute corresponding to the context-aware connecting edge in the candidate path is compared with the permission level corresponding to the trust assessment request for privacy constraints, and when it is determined that there is a connecting edge in the candidate path with a privacy level higher than the permission level, the candidate path containing the connecting edge is blocked and eliminated to complete the filtering of effective trust paths.

[0022] Preferably, in the trust calculation step: the trust transfer calculation is used to quantify the overall trust evidence on a single valid trust path. The specific calculation process includes: forming a trust evidence group by combining the original trust score and evidence confidence level of each context-aware connection edge on the valid trust path; and performing recursive cumulative product calculation on the trust score and evidence confidence level of each connection edge along the valid trust path that is connected end to end, to obtain a comprehensive path trust evidence that includes the overall path trust score and the overall path evidence confidence level.

[0023] Preferably, in the trust calculation step: the trust evidence fusion calculation is used to perform overall fusion of comprehensive path trust evidence corresponding to multiple valid trust paths or multiple source scenarios; the fusion calculation method includes confidence-weighted fusion, maximum value fusion, minimum value fusion or harmonic average fusion; when performing the confidence-weighted fusion, the comprehensive path trust evidence of each valid path is merged, and the final trust score after fusion is obtained by weighted summation of the confidence of each comprehensive path trust evidence, and the final evidence confidence after fusion is calculated by accumulating the confidence of each path and adding a preset upper bound constraint, and this calculation process satisfies the commutative law.

[0024] Preferably, in the security control step: the calculated comprehensive trust result is compared and verified with the security allowance threshold and the downgrade processing threshold corresponding to the target evaluation scenario; if the comprehensive trust result reaches the security allowance threshold and satisfies the privacy constraint policy, the corresponding authentication allowance or permission granting instruction is triggered; if the comprehensive trust result does not reach the security allowance threshold but is higher than the downgrade processing threshold and satisfies the privacy constraint policy, a downgrade verification instruction including restricting access permissions, secondary authentication, supplementary verification elements, or manual review is triggered; if the comprehensive trust result is lower than the downgrade processing threshold or does not satisfy the privacy constraint policy, an access interception instruction is directly triggered.

[0025] Preferably, in the scenario modeling step: the scenario six-tuple includes an interaction source subject, an interaction target subject, an interaction type, an interaction target, a governance rule, and a privacy level attribute; the interaction source subject and the interaction target subject are used to determine the nodes and interaction directions in the context-aware trust graph; the interaction type, interaction target, governance rule, and privacy level are used to characterize the scenario semantic differences of node interactions; the interaction type includes interaction forms formed by any combination of consumers, enterprises, and governments; and the privacy level is divided into a strictly increasing sequence of levels according to the degree of data disclosure and confidentiality.

[0026] Preferably, the method further includes a result caching step following the security control step: using the combined features of the source node, target node, source scenario, and target evaluation scenario as a cache index, the calculated comprehensive trust result is stored in the cache unit; during the graph state update process, when a new operation, a deletion operation, or a corresponding trust weight value update is detected in the context-aware connection edge in the context-aware trust graph, the system retrieves and clears all historical cache results associated with this connection edge based on its feature information.

[0027] A multi-level trust management and evaluation system for heterogeneous network nodes, the system comprising system modules for executing the method, specifically including:

[0028] The scene modeling module is used to obtain the historical interaction records of nodes in the heterogeneous network and their business context information, and to construct a scene six-tuple containing the interaction subject and multi-dimensional scene semantic attributes.

[0029] The graph construction module is used to construct a context-aware trust graph that reflects the multi-scenario association between nodes based on the scenario six-tuple, the node's historical interaction records, and the corresponding trust weights.

[0030] The path filtering module is used to receive trust evaluation requests containing the source node to be evaluated, the target node, and the target evaluation scenario. It combines the scenario labels of the connecting edges, the path hop limit, and the privacy permission conditions to search and filter out valid trust paths.

[0031] The trust calculation module is used to quantify the trust evidence corresponding to the effective trust path using a preset set of trust evaluation operators. The calculation includes performing trust transfer calculation, trust evidence fusion calculation, and cross-scenario trust decay calculation to obtain a comprehensive trust result.

[0032] The evaluation and decision-making module is used to strictly match and verify the comprehensive trust result with the security access threshold and privacy constraint policy of the target evaluation scenario, generate the node evaluation result for the source node, and execute the corresponding permission control action.

[0033] The beneficial effects of this invention are as follows: The invention incorporates interaction type, interaction goal, governance rules, and privacy level into the trust modeling process through a scenario six-tuple, enabling the same node pair to form different trust evaluation criteria in different business scenarios, reducing the problem of directly reusing a single global trust value in heterogeneous scenarios. This invention represents nodes, context-aware edges, trust weights, and scenario labels through a context-aware trust graph, allowing node interaction relationships in heterogeneous networks to simultaneously express topological connectivity and business context relationships, improving the completeness of trust evidence expression. This invention calculates and fuses trust evidence under single-path, multi-path, and multi-source scenarios through trust propagation and trust synthesis operators, enabling trust assessment to utilize both direct trust evidence and multi-hop indirect trust evidence simultaneously. Since the source path, scenario label, privacy level, and calculation process of trust evidence can all be recorded and traced, the applicability, interpretability, and auditability of trust assessment in complex network environments are improved. This invention employs a cross-scenario attenuation operator to attenuate trust evidence based on the scenario distance between the source and target scenarios. This constrains cross-scenario trust transfer by differences in interaction type, interaction target, governance rules, and privacy level, reducing excessive trust migration and semantic shifts across scenarios. After generating the trust assessment results, this invention combines target scenario security constraint policies to implement security controls such as authentication granting, permission assignment, downgrade verification, or access blocking. This allows the trust assessment results to directly serve access control, identity authentication, and permission management in heterogeneous networks, improving the engineering feasibility, assessment efficiency, and security control effectiveness of the trust assessment method. Attached Figure Description

[0034] Figure 1 A schematic diagram of the overall process of a multi-level trust evaluation method for heterogeneous networks based on context-aware trust graph algebra provided by the present invention.

[0035] Figure 2 This is a context-aware trust graph algebraic framework diagram in this invention;

[0036] Figure 3 This is a schematic diagram of the overall framework of the context-aware trust graph algebra in this invention;

[0037] Figure 4 This is an ablation analysis diagram of the CTAG operator in the example. Detailed Implementation

[0038] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0039] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0040] The embodiments of the present invention will be further described in detail below with reference to the accompanying drawings:

[0041] This invention provides a multi-level trust assessment method for heterogeneous networks, comprising the following steps: a scenario modeling step, acquiring historical interaction records of nodes in the heterogeneous network and their business context information, and constructing a scenario six-tuple containing the interaction subject and multi-dimensional scenario semantic attributes; a graph construction step, constructing a context-aware trust graph reflecting the multi-scenario association relationship between nodes based on the scenario six-tuple, historical interaction records of nodes, and corresponding trust weights, mapping the interactions of the same source node and target node under different business scenarios to independent connection edges with different scenario labels; and a path filtering step, receiving a trust assessment request containing the source node to be evaluated, the target node, and the target evaluation scenario, and within the context-aware trust graph, combining the scenario labels of the connection edges, path hop count limits, and privacy permissions. The process involves several steps: First, searching and filtering for valid trust paths from the source node to the target node. Second, a trust calculation step utilizes a pre-defined set of trust evaluation operators to quantify the trust evidence corresponding to the valid trust paths. This calculation includes cumulative trust transfer calculation along a single path, trust evidence fusion calculation for multiple paths or multiple source scenarios, and cross-scenario trust attenuation calculation based on semantic attribute differences between the source and target scenarios, ultimately yielding a comprehensive trust result for the target evaluation scenario. Third, a security control step rigorously matches and verifies the calculated comprehensive trust result against the security access threshold and privacy constraint policies corresponding to the target evaluation scenario, generating a node evaluation result for the source node. Based on this evaluation result, corresponding access authorization, downgrade verification, or access interception control actions are executed.

[0042] In the scenario modeling step, the system collects historical interaction records between nodes from sources such as on-chain interaction logs, transaction records, service call records, identity authentication records, and inter-node evaluation records of the heterogeneous network, and extracts the business context information carried by each interaction. A heterogeneous network refers to a network with various participating entities of different natures, such as consumer nodes, enterprise nodes, and government nodes, with differences in interaction types, objectives, and compliance constraints among these entities. Since a single global reputation value is insufficient to simultaneously characterize the trust differences under these multiple scenarios, this invention provides a structured description of each interaction record according to its business context, forming a scenario six-tuple.

[0043] Scene Six-Tuple Used to describe the business environment in which node interactions take place, it can be represented as:

[0044]

[0045] in, Indicates the main body of the interaction source. Indicates the target subject of the interaction. Indicates the interaction type. Indicate the interaction target, Indicates governance rules, Indicates the privacy level. Main body of the interaction source. and the target subject of interaction Used to determine nodes and interaction directions in the context-aware trust graph, it serves as the topological anchor point for trust path routing and does not directly participate in the measurement of scene semantic differences; interaction type Interaction goals Governance rules and privacy level These characteristics collectively represent the semantic differences in the interaction between nodes, which are then used for the subsequent quantitative calculation of cross-scene semantic distance.

[0046] In the six-tuple of a scene, the interaction type This includes interaction forms formed by any combination of consumers, businesses, and governments, specifically covering at least one of consumer-to-consumer, business-to-consumer, business-to-business, government-to-business, and government-to-consumer interactions, corresponding to different business backgrounds such as personal social networking, e-commerce, cross-organizational cooperation, government supervision, and public services. Interaction Objectives This describes the specific business objective to be achieved in this interaction, such as product transaction, data sharing, identity verification, or compliance audit. Governance Rules This describes the applicable regulatory policies, compliance frameworks, or contractual constraints for this interaction, such as general data protection regulations, industry licensing agreements, or on-chain smart contract terms. Privacy Level The data is classified into a strictly increasing hierarchy based on its level of openness and confidentiality, including at least one of the following: open, internal, confidential, and secret. The hierarchy shows a monotonically increasing relationship from low to high confidentiality level. The higher the level, the higher the sensitivity of the data and the stricter the access control constraints.

[0047] Through the above-mentioned six-tuple scenario, the historical interaction records generated by the same node in different business scenarios can be modeled as trust evidence with different scenario semantics, thereby avoiding the indiscriminate mixing of trust evaluation results formed in different business scenarios and reducing the risk of semantic deviation and privacy leakage caused by direct reuse of cross-scenario trust.

[0048] In the graph construction step, the system constructs a context-aware trust graph based on scenario six-tuples, node historical interaction records, and corresponding trust weights to reflect the multi-scenario relationships between nodes. Context-aware Trust Graph It can be represented as:

[0049]

[0050] in, For a set of nodes, For a context-aware set of connection edges, For the trust weight mapping function, This is the scene label mapping function. The context-aware trust graph is formally a directed multigraph, where any two nodes can have multiple connections with different scene labels, rather than retaining only a single connection.

[0051] Context-aware connection edges consist of four parts: a source node, a target node, a scenario six-tuple, and a trust weight. The source and target nodes determine the direction of the trust relationship; the scenario six-tuple records the business context attributes of the current interaction, serving as the scenario semantic identifier for the connection edge; the trust weight is obtained by normalizing quantitative indicators such as ratings, fulfillment rates, and feedback information from historical interaction records, and its value range is constrained within a preset interval, reflecting the degree of trust the source node has in the target node within that scenario.

[0052] When interaction records occur in the network with the same source node and target node but different scenario six-tuples, the system uses the combination of source node, target node, and scenario six-tuple as the retrieval condition to establish multiple context-aware connection edges with different scenario labels for the node pair in the context-aware trust graph. For example, if node A and node B have transaction records in both an e-commerce scenario and a service call relationship in a data sharing scenario, then two independent context-aware connection edges are established in the graph, each carrying its own scenario six-tuple and trust weight. This multi-edge mechanism allows the trust relationship of the same node pair in different business scenarios to be distinguished, avoiding the direct reuse of a single global trust value in heterogeneous scenarios. Scenario label mapping function It is responsible for mapping each context-aware connection edge to its corresponding scene six-tuple to support scene filtering operations in the subsequent path selection stage.

[0053] In the path selection step, the system receives a trust assessment request containing the source node to be evaluated, the target node, and the target assessment scenario. The trust assessment request also includes a set of source scenarios for extracting initial trust evidence, and the target assessment scenario is used to determine the final business scenario to which the node assessment results apply and its corresponding privacy constraint policies. The target assessment scenario clarifies the security requirements and privacy constraints under which the trust assessment results will be used for access control decisions, and is one of the core parameters in subsequent trust computation and security control phases.

[0054] The system searches for candidate paths from the source node to the target node within a preset maximum hop count range. It then performs initial filtering based on the scenario labels of the connecting edges, retaining paths formed by connecting edges that match the source scenario set. The maximum hop count limit prevents excessive dilution of trust evidence when it travels along excessively long paths and controls the computational complexity of path search. After scenario filtering, the system compares the privacy level attributes of the context-aware connecting edges in the candidate paths with the permission level corresponding to the trust assessment request. When a candidate path contains a connecting edge with a privacy level higher than the permission level, the system blocks and removes the candidate path containing that connecting edge, thus completing the screening of valid trust paths. By performing privacy constraint checks during the path search phase, the system ensures that only trust evidence that meets privacy permission conditions is included in subsequent trust transmission and fusion calculations, preventing non-compliant trust evidence from entering the evaluation process.

[0055] In the trust calculation step, the system uses a preset set of trust evaluation operators to quantify the trust evidence corresponding to the selected effective trust paths. The set of operators includes the trust propagation operator corresponding to trust transmission calculation, the trust synthesis operator corresponding to trust evidence fusion calculation, and the cross-scenario attenuation operator corresponding to cross-scenario trust attenuation calculation.

[0056] Trust transitivity computation is used to quantify the overall trust evidence along a single valid trust path. In this embodiment, trust evidence along a single context-aware connection edge is represented as a trust evidence set containing a trust score and evidence confidence level, i.e., a trust tuple. , can be represented as:

[0057]

[0058] in, The trust score represents the degree of trust that the source node has in the target node in this scenario, and its value ranges from [0,1]. The trust score represents the confidence level of the evidence, reflecting the sufficiency and reliability of the historical evidence upon which it is based, and its value range is [0,1]. The two-component structure of the trust tuple can simultaneously characterize the degree of trust itself and the degree of certainty regarding that degree of trust, and can more accurately express trust judgments under incomplete information than a single scalar trust value.

[0059] For two consecutive segments of trust evidence connected end-to-end on an effective trust path, the trust propagation operator can be expressed as:

[0060]

[0061] in, and These represent the trust score and the confidence level of the first piece of evidence. and These represent the trust score and evidence confidence level of the second segment of trust evidence, respectively. This operator, by multiplying the trust score and evidence confidence level, reflects the natural characteristic of trust gradually decaying with the number of hops as it propagates along the path: the longer the path, the smaller the cumulative product of the trust weights of each connecting edge and the evidence confidence level, resulting in a lower overall path trust evidence. The trust propagation operator satisfies the associative law, meaning that when this operator is executed sequentially on three or more segments of path trust evidence, the order of operations does not affect the final result, thus supporting sequential recursive calculation of valid trust paths of arbitrary length. For a valid trust path composed of multiple context-aware connecting edges, path trust evidence is obtained by sequentially executing the trust propagation operator on the trust tuples corresponding to each connecting edge in the path, ultimately forming a comprehensive path trust evidence containing both the overall path trust score and the overall path evidence confidence level.

[0062] Trust evidence fusion computation is used to perform overall fusion of trust evidence from multiple valid trust paths or multiple source scenarios. Under the confidence-weighted fusion method, for two comprehensive path trust evidence... , ,when and When the sum is greater than zero, the trust composition operator can be expressed as:

[0063]

[0064] in, and These represent the trust score and the confidence level of the evidence from the first path of trust, respectively. and These are the trust score and evidence confidence level of the second path's trust evidence, respectively. The final trust score after synthesis is obtained by weighted summation of the confidence levels of the trust evidence from each integrated path, reflecting the principle that the higher the confidence level of the path, the greater its contribution to the synthesis result. The final evidence confidence level after synthesis is calculated by accumulating the confidence levels of each path and adding a preset upper bound constraint. The upper bound constraint ensures that the confidence level after synthesis does not exceed 1, avoiding confidence overflow caused by evidence superposition.

[0065] This trust synthesis operator satisfies the commutative law when combining two trust evidences, that is... This ensures that the synthesis result of trust evidence from any two paths is unaffected by the order of their inputs. In a distributed heterogeneous network environment, trust evidence from different paths may arrive asynchronously in a non-deterministic order. The commutative law guarantees that the final synthesis result remains consistent regardless of the order in which the evidence arrives. When performing multi-path batch fusion, the system sequentially executes the trust synthesis operator on the comprehensive path trust evidence of all valid paths, gradually aggregating the multi-path trust evidence into a single trust tuple, which serves as the input for subsequent cross-scenario attenuation calculations. In other implementations, for the normalized trust value obtained by reducing the trust tuple, the trust synthesis operator can also employ at least one of the following methods: maximum value fusion, minimum value fusion, or harmonic average fusion, to adapt to the trust evidence synthesis requirements under different security policies or different business scenarios.

[0066] Cross-scenario trust decay calculation is used to perform decay mapping on trust evidence after multi-path fusion based on the semantic differences between the source and target scenarios, reflecting the credibility constraints of trust evidence when crossing different business scenarios. Scenario distance is determined based on differences in at least one attribute among interaction type, interaction goal, governance rules, and privacy level in the scenario six-tuple. In this embodiment, scenario distance can be expressed as:

[0067]

[0068] in, Indicates the source scenario, Indicates the target scenario. Iterate through the four attribute dimensions: interaction type I, interaction goal O, governance rule R, and privacy level P. This represents the weight corresponding to the k-th attribute, used to reflect the importance of that attribute in the semantic difference measurement of the scene. System administrators can adjust the weight of each attribute according to specific business security policies. Customize the configuration; Z represents the difference function of the k-th attribute; Z is the normalization parameter, equal to each The sum of the weighted sums ensures that the calculated scene distance is normalized to the [0,1] interval; the outer min function further truncates the weighted sum to 1 to prevent the scene distance from exceeding the effective range in extreme cases.

[0069] For categorical attributes such as interaction type I, interaction goal O, and governance rule R, the degree of local attribute difference is determined based on whether the attribute classification is consistent or a preset semantic difference matrix. When two scenarios have the same value on a certain categorical attribute, the local difference is 0, indicating that there is no semantic distance between scenarios in that dimension. When the values ​​are different, the local difference is assigned a corresponding positive value based on the semantic distance between attributes. For the privacy level P attribute, the degree of local attribute difference is determined based on a preset normalized sequence level difference. The privacy level attribute difference can be expressed as:

[0070]

[0071] in, and Let x and y represent the corresponding privacy levels in the privacy level sequence, respectively. Q is a preset normalization parameter, which is the difference between the highest and lowest levels in the privacy level sequence, ensuring that the difference is normalized to the [0,1] interval. The privacy level sequence strictly increases in the order of public, internal, confidential, and secret. The larger the difference in levels, the more significant the difference in data confidentiality requirements between the source and target scenarios, and the higher the degree of cross-scenario attenuation.

[0072] After obtaining the scene distance δ, the cross-scene attenuation operator performs attenuation mapping on the trust tuples after multipath fusion. In this embodiment, the cross-scene attenuation operator can be expressed as:

[0073]

[0074] in, This is a decay factor used to control the strictness of cross-scenario isolation, and is configured by the system administrator according to the security policy. The distance to the above scene is defined, with a value ranging from [0,1]. This is due to the exponential function... about Strictly monotonically decreasing, when When the value is 0, the attenuation factor is 1, meaning that cross-scene attenuation is not performed within the same scene; when... As the confidence score increases, and confidence level of evidence All values ​​are adjusted synchronously at the same exponential ratio, ensuring that the greater the semantic difference between the source and target scenarios, the lower the final trust assessment result. This monotonically non-increasing characteristic algebraically ensures that cross-scenario trust transfer is strictly constrained by business semantic boundaries, preventing high trust values ​​in low-security scenarios from being migrated to high-security scenarios without attenuation, thereby reducing the risks of trust overreach and privacy leaks.

[0075] In another implementation, in the attenuation factor Distance from the scene When the product is within a relatively small preset range, a first-order approximation function of the exponential decay function can be used for calculation, i.e., using... Approximate substitution Furthermore, boundary constraints are added to the calculation results to ensure that the attenuated trust score and evidence confidence are not lower than zero. The above two implementations of the cross-scenario attenuation operator can cover both attenuation calculations based on trust tuples and attenuation calculations based on normalized trust values ​​obtained by reducing trust tuples, and can be flexibly selected according to actual computing resources and accuracy requirements.

[0076] It should be noted that the trust propagation operator, trust synthesis operator, and cross-scenario attenuation operator can be combined and executed according to the type of the target trust assessment request. In single-path trust assessment in a single scenario, path trust evidence is obtained only based on the trust propagation operator; in multi-path trust assessment, a comprehensive trust result is obtained based on the trust propagation operator and the trust synthesis operator; in cross-scenario trust assessment, a trust result in the target scenario is obtained based on the trust propagation operator and the cross-scenario attenuation operator; in multi-source scenario or multi-path cross-scenario trust assessment, trust evidence from multiple source scenarios or multiple paths is propagated and attenuated across scenarios separately, and then trust synthesis is further performed. Cross-scenario attenuation is only executed once in the final cross-scenario mapping stage and is not repeatedly applied during the edge-by-edge propagation process within the path, to ensure that the attenuation operation is strictly defined as a projection transformation from the source scenario to the target scenario, rather than an edge-by-edge accumulated penalty term. The above operator combination mechanism enables this invention to adapt to various practical application needs such as direct assessment within the same scenario, indirect assessment across scenarios, and comprehensive assessment of multiple paths.

[0077] To improve the system's response efficiency in large-scale heterogeneous networks, the trust calculation step also includes a trust result caching mechanism: using the combined features of the source node, target node, source scenario, and target evaluation scenario as a cache index, the calculated comprehensive trust result is stored in the cache unit. During the graph state update process, when a new operation, deletion operation, or corresponding trust weight update occurs on a context-aware connection edge in the context-aware trust graph, the system retrieves and clears all associated historical cache results based on the feature information of that connection edge. The granular design of the cache index ensures that only the cache entries corresponding to the affected paths are cleared, while the cached trust evidence of paths unaffected by this update is retained, thus balancing computational efficiency and result consistency in dynamic graph update scenarios. Furthermore, this invention exhibits semantic orthogonality during scenario expansion: when injecting context-aware connection edges corresponding to new business scenarios into the graph, the connection edges and their trust evidence under existing scenarios are unaffected, and the trust evaluation results of existing trust paths remain unchanged, enabling the system to dynamically adapt to new business scenarios without recalculating the entire graph's trust value.

[0078] In the security control step, the system rigorously matches and verifies the comprehensive trust result obtained from the trust calculation step with the security access threshold and privacy constraint policy corresponding to the target evaluation scenario to generate a node evaluation result for the source node and execute corresponding access control actions. The comprehensive trust result includes trust tuples, or normalized trust values ​​obtained by reducing trust tuples. The normalized trust value is constrained to the range of [0,1], and the higher the value, the higher the comprehensive trust level of the target node in the target evaluation scenario.

[0079] Specifically, the system compares and verifies the overall trust result with the security allowance threshold and the downgrade processing threshold corresponding to the target evaluation scenario: If the overall trust result is determined to meet the security allowance threshold and the privacy constraint policy, the corresponding authentication allowance or permission granting instruction is triggered, allowing the source node to access the target resource or service with the corresponding permissions in the target evaluation scenario; If the overall trust result is determined to be lower than the security allowance threshold but higher than the downgrade processing threshold and the privacy constraint policy is met, a downgrade verification instruction is triggered, including restricting access permissions, secondary authentication, supplementary verification elements, or manual review, and a downgrade authorization policy is adopted for the source node; If the overall trust result is determined to be lower than the downgrade processing threshold or does not meet the privacy constraint policy, an access interception instruction is directly triggered, and the access request of the source node is rejected.

[0080] The verification of privacy constraint policies includes reviewing the privacy levels of the connection edges traversed by the source path of the comprehensive trust result to ensure that the trust evidence actually used in the trust assessment process does not exceed the privacy permission scope allowed by the target assessment scenario. The security allowance threshold and downgrade processing threshold are configured by the system administrator according to the security level requirements of the target assessment scenario. Thresholds under different business scenarios, different privacy levels, and different governance rules can be set independently, supporting fine-grained dynamic access control policies. By directly connecting the trust assessment results with the scenario security constraint policies, this invention enables the output of trust assessment to directly drive access control, identity authentication, and permission management operations, improving the engineering feasibility and security control effectiveness of the trust assessment method.

[0081] This invention also provides a multi-level trust assessment system for heterogeneous networks. The system includes system modules for executing the above-described methods, specifically: a scenario modeling module, used to acquire historical interaction records of nodes in the heterogeneous network and their business context information, and construct a scenario six-tuple containing the interaction subject and multi-dimensional scenario semantic attributes; a graph construction module, used to construct a context-aware trust graph reflecting the multi-scenario association relationship between nodes based on the scenario six-tuple, historical interaction records of nodes, and corresponding trust weights; a path filtering module, used to receive a trust assessment request containing the source node to be assessed, the target node, and the target assessment scenario, and search and filter out effective trust paths by combining the scenario labels of the connecting edges, path hop count limits, and privacy permission conditions; a trust calculation module, used to quantify the trust evidence corresponding to the effective trust path using a preset set of trust assessment operators, including performing trust transfer calculation, trust evidence fusion calculation, and cross-scenario trust decay calculation, to obtain a comprehensive trust result; and an assessment decision module, used to strictly match and verify the comprehensive trust result with the security access threshold and privacy constraint policy of the target assessment scenario, generate a node assessment result for the source node, and execute the corresponding permission control action.

[0082] The scene modeling module is responsible for acquiring raw interaction records from multiple data sources, extracting the interaction subjects, interaction types, interaction goals, governance rules, and privacy levels from the records in a structured manner, and generating a corresponding scene six-tuple for each interaction record. As new interaction records are continuously added to the system, the scene modeling module transforms them into scene six-tuples in real time, providing the graph construction module with incremental updates to the context-aware trust graph.

[0083] After receiving the scene six-tuple and trust weight, the graph construction module uses the source node, target node, and scene six-tuple triple combination as the search key to determine whether a corresponding context-aware connection edge already exists in the context-aware trust graph. If it exists, the trust weight of the edge is updated according to a preset weight update strategy, such as using a sliding window weighted average or time decay weighting to fuse the trust evaluations of new and old interaction records. If it does not exist, a corresponding context-aware connection edge is created in the graph. The graph construction module also maintains a scene label mapping function Λ to ensure that the scene six-tuple of each connection edge corresponds one-to-one with its storage location in the graph, supporting scene label filtering during the path selection stage.

[0084] Upon receiving a trust assessment request, the path filtering module initiates a path search constrained by the maximum hop count from the context-aware trust graph. It then sequentially performs scene label filtering and privacy permission constraint checks, outputting a set of valid trusted paths that satisfy all filtering criteria. The path filtering module restricts the path search space to the scene namespaces that match the source scene set, ensuring that the search process only traverses local subgraphs semantically relevant to the current assessment request, rather than the complete global graph, thereby reducing the computational complexity of the path search.

[0085] The trust calculation module performs trust quantification calculations based on a set of valid trust paths, following a combination of trust propagation operators, trust synthesis operators, and cross-scenario attenuation operators. After calculation, the results are written to the trust result cache unit. When the cache is hit, the trust calculation module directly returns the trust result from the cache, eliminating the need to re-execute the complete path traversal and operator operations, thus reducing the response latency of repeated evaluation requests.

[0086] The assessment and decision-making module is responsible for matching the comprehensive trust results output by the trust calculation module with the security policy parameters corresponding to the target assessment scenario, and selecting the appropriate access control action instructions to send to the execution layer based on the matching results. The assessment and decision-making module supports multi-level threshold configuration, allowing system administrators to independently set allow and degrade thresholds according to the privacy level and governance rules of different target assessment scenarios, thereby achieving fine-grained permission control for different business scenarios.

[0087] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principle of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A multi-level trust management evaluation method for heterogeneous network nodes, characterized in that, include: Scene modeling steps: Obtain historical interaction records of nodes in the heterogeneous network and their business context information, and construct a scene six-tuple containing the interaction subject and multi-dimensional scene semantic attributes accordingly; The graph construction steps are as follows: Based on the scenario six-tuple, the node historical interaction records and the corresponding trust weights, a context-aware trust graph reflecting the multi-scenario association between nodes is constructed. For the interaction of the same source node and target node in different business scenarios, it is mapped as independent connection edges with different scenario labels. Path filtering step: Receive a trust evaluation request containing the source node to be evaluated, the target node, and the target evaluation scenario. Within the context-aware trust graph, combine the scenario labels of the connecting edges, the path hop limit, and the privacy permission conditions to search and filter out the valid trust paths from the source node to the target node. Trust calculation steps: The trust evidence corresponding to the effective trust path is quantitatively calculated using a preset set of trust evaluation operators. The calculation process includes accumulated trust transfer calculation along a single path, trust evidence fusion calculation for multiple paths or multiple source scenarios, and cross-scenario trust decay calculation based on the semantic attribute differences between the source scenario and the target scenario, thereby obtaining the comprehensive trust result under the target evaluation scenario. Security control steps: The calculated comprehensive trust result is strictly matched and verified with the security access threshold and privacy constraint policy corresponding to the target evaluation scenario to generate a node evaluation result for the source node, and the corresponding access authorization, downgrade verification or access interception control actions are executed based on the evaluation result.

2. The method for evaluating multi-level trust management of heterogeneous network nodes according to claim 1, characterized in that, In the trust calculation step, the specific process of cross-scenario trust decay calculation is as follows: compare the interaction type, interaction target, governance rules and privacy level attributes in the six-tuples corresponding to the source scenario and the target scenario, and calculate the semantic scenario distance between the two scenarios; For categorical attributes, the degree of local attribute difference is determined based on whether the attribute classification is consistent or a preset semantic difference matrix; for privacy level attributes, the degree of local attribute difference is determined based on a preset normalized sequence level difference. After obtaining the semantic scene distance by weighted summation and normalization of the local differences, the target attenuation coefficient is calculated using a monotonically non-increasing attenuation function or a first-order approximation function of the attenuation function, and then the attenuation coefficient is mapped to the trust evidence to be transmitted.

3. The method for evaluating multi-level trust management of heterogeneous network nodes according to claim 1, characterized in that, In the graph construction step: the context-aware trust graph includes a node set, a context-aware connection edge set, a trust weight mapping function, and a scene label mapping function. The context-aware connection edge is composed of a source node, a target node, a scene six-tuple, and a trust weight. When an interaction record with the same source node and target node but different scene six-tuples occurs in the network, the combination of the source node, target node, and scene six-tuple is used as the retrieval condition to establish multiple context-aware connection edges with different scene labels for the node pair in the context-aware trust graph.

4. The method for evaluating multi-level trust management of heterogeneous network nodes according to claim 1, characterized in that, In the path filtering step: the trust assessment request further includes a set of source scenarios for extracting initial trust evidence, and the target assessment scenario is used to determine the final business scenario to which the node assessment result applies and its corresponding privacy constraint policy; the system searches for candidate paths from the source node to the target node within a preset maximum hop count range, and performs preliminary filtering on the candidate paths based on the scenario labels of the connecting edges; the privacy level attribute corresponding to the context-aware connecting edge in the candidate path is compared with the permission level corresponding to the trust assessment request for privacy constraints, and when it is determined that there is a connecting edge in the candidate path with a privacy level higher than the permission level, the candidate path containing the connecting edge is blocked and removed to complete the filtering of effective trust paths.

5. The method for evaluating multi-level trust management of heterogeneous network nodes according to claim 1, characterized in that, In the trust calculation step: The trust transfer calculation is used to quantify the overall trust evidence on a single valid trust path. The specific calculation process includes: forming a trust evidence group by combining the original trust score and the evidence confidence level of each context-aware connection edge on the valid trust path. Along the effective trust path that connects the beginning and end, the trust score and evidence confidence of each connecting edge are recursively accumulated and multiplied to obtain the comprehensive path trust evidence, which includes the overall path trust score and the overall path evidence confidence.

6. The method for evaluating multi-level trust management of heterogeneous network nodes according to claim 1, characterized in that, In the trust calculation step: the trust evidence fusion calculation is used to perform overall fusion of comprehensive path trust evidence corresponding to multiple valid trust paths or multiple source scenarios; the fusion calculation method includes confidence-weighted fusion, maximum value fusion, minimum value fusion or harmonic average fusion; When performing the confidence-weighted fusion, the comprehensive path trust evidence of each valid path is merged. The final trust score after fusion is obtained by weighted summation of the confidence of each comprehensive path trust evidence. The final evidence confidence is calculated by accumulating the confidence of each path and adding a preset upper bound constraint. This calculation process satisfies the commutative law.

7. The method for evaluating multi-level trust management of heterogeneous network nodes according to claim 1, characterized in that, In the security control step: the calculated comprehensive trust result is compared and verified with the security release threshold and downgrade processing threshold corresponding to the target evaluation scenario, respectively; If the overall trust result is determined to meet the security clearance threshold and the privacy constraint policy, the corresponding authentication clearance or permission granting instruction is triggered; if the overall trust result is determined to be lower than the security clearance threshold but higher than the downgrade processing threshold and the privacy constraint policy is met, a downgrade verification instruction including restricting access permissions, secondary authentication, supplementary verification elements, or manual review is triggered; if the overall trust result is determined to be lower than the downgrade processing threshold or does not meet the privacy constraint policy, an access blocking instruction is directly triggered.

8. The method for evaluating multi-level trust management of heterogeneous network nodes according to claim 1, characterized in that, In the scenario modeling step: the scenario six-tuple includes the interaction source subject, the interaction target subject, the interaction type, the interaction target, the governance rule, and the privacy level attribute; the interaction source subject and the interaction target subject are used to determine the nodes and interaction directions in the context-aware trust graph, and the interaction type, the interaction target, the governance rule, and the privacy level are used to characterize the scenario semantic differences of node interactions; the interaction type includes the interaction form formed by any combination of consumers, enterprises, and governments, and the privacy level is divided into a strictly increasing level sequence according to the degree of data openness and confidentiality.

9. A multi-level trust management evaluation method for heterogeneous network nodes according to any one of claims 1 to 8, characterized in that, The method further includes a result caching step following the security control step: using the combined features of the source node, target node, source scenario, and target evaluation scenario as a cache index, the calculated comprehensive trust result is stored in the cache unit; During the graph state update process, when a new operation, a deletion operation, or a corresponding trust weight update is detected in the context-aware connection edge of the context-aware trust graph, the system retrieves and clears all historical cached results associated with the connection edge based on its feature information.

10. A multi-level trust management and evaluation system for heterogeneous network nodes, characterized in that, The system includes a system module for performing the method as described in any one of claims 1 to 9, specifically including: The scene modeling module is used to obtain the historical interaction records of nodes in the heterogeneous network and their business context information, and to construct a scene six-tuple containing the interaction subject and multi-dimensional scene semantic attributes. The graph construction module is used to construct a context-aware trust graph that reflects the multi-scenario association between nodes based on the scenario six-tuple, the node's historical interaction records, and the corresponding trust weights. The path filtering module is used to receive trust evaluation requests containing the source node to be evaluated, the target node, and the target evaluation scenario. It combines the scenario labels of the connecting edges, the path hop limit, and the privacy permission conditions to search and filter out valid trust paths. The trust calculation module is used to quantify the trust evidence corresponding to the effective trust path using a preset set of trust evaluation operators. The calculation includes performing trust transfer calculation, trust evidence fusion calculation, and cross-scenario trust decay calculation to obtain a comprehensive trust result. The evaluation and decision-making module is used to strictly match and verify the comprehensive trust result with the security access threshold and privacy constraint policy of the target evaluation scenario, generate the node evaluation result for the source node, and execute the corresponding permission control action.

Images