Blockchain-based talent background verification method and system

Abstract

This invention relates to the field of blockchain technology, and in particular to a blockchain-based method and system for verifying talent backgrounds. The method includes acquiring multiple background credentials of the talent to be verified, constructing a spatiotemporal constraint graph, identifying temporal and spatial conflicts, generating conflict feature vectors, constructing a directed acyclic graph of credential dependencies, traversing descendant credentials when a credential is revoked, detecting the existence of substitute credentials, allocating decay coefficients, updating trust, and generating a trust time series. It also involves extracting the trust time series of multiple talents, calculating a weighted cross-correlation coefficient using the decay coefficient as weight, and generating an anomaly verification identifier when the mean of the conflict feature vectors or the weighted cross-correlation coefficient exceeds a threshold, which is then written into the blockchain. This method effectively identifies conflicts and dependency anomalies in background credentials, improving verification accuracy and tamper resistance.

Description

Technical Field

[0001] This invention relates to the field of blockchain technology, and in particular to a blockchain-based method and system for verifying talent backgrounds. Background Technology

[0002] Currently, the field of talent background verification generally adopts centralized or semi-centralized verification models. Institutions such as schools, employers, or third-party background investigation companies act as issuers or verifiers of credentials, confirming the authenticity of credentials through database comparison, manual verification, or API calls. Credentials are usually stored in the form of electronic documents or scanned paper copies, lacking automatic analysis of the spatiotemporal correlation between credentials. Some solutions attempt to introduce blockchain technology to record the hash value or digital signature of credentials on the chain to enhance tamper resistance, but the core verification logic still relies on centralized entities reading the original credentials and manual judgment.

[0003] A significant drawback of existing practices is the difficulty in effectively detecting spatiotemporal inconsistencies between vouchers. Due to the inefficiency of manual verification and the fact that vouchers are scattered across databases in different systems, when multiple vouchers have overlapping time periods or significant geographical conflicts (such as appearing in two different locations at the same time), traditional methods often rely on the subjective experience of the verifier or post-event investigations, failing to automatically identify such inconsistencies. This allows forged, time-inconsistent, or spatially contradictory vouchers to pass through the review process, resulting in distorted background information.

[0004] Another shortcoming lies in the lack of a trust transmission mechanism after credential revocation. Existing blockchain-based evidence storage solutions typically only record the existence of credentials. When a credential is revoked by the issuer, the trust level of other credentials that depend on it is not automatically adjusted. Verifiers, in subsequent queries, still arrive at overly high trust judgments based on the corrupted credential chain. The lack of a dynamic decay mechanism makes the entire verification system insufficiently robust, unable to promptly reflect global trust changes caused by local failures in the credential chain. Summary of the Invention

[0005] This invention provides a blockchain-based method and system for verifying talent backgrounds, which can solve the problems in the prior art.

[0006] A first aspect of this invention provides a blockchain-based method for verifying talent backgrounds, comprising:

[0007] Obtain multiple background credentials for the talent to be verified, each credential including a time range, geographical location, and credential type identifier;

[0008] The credentials are constructed as a spatiotemporal constraint graph. By calculating the proportion of time interval intersection between nodes and the ratio of geographical distance to time interval, temporal conflicts and spatial conflicts are identified, and conflict feature vectors are generated.

[0009] On the blockchain, the hash identifier of the preceding dependent certificate is recorded for each background certificate, and a directed acyclic graph of certificate dependencies is constructed. When a certificate is revoked, all descendant certificates are traversed along the outgoing edges of the directed acyclic graph, the existence of alternative certificates of the same type is detected, and the trust level of the descendant certificate is updated according to the existence by assigning a differentiated decay coefficient. A trust level time series containing the correspondence between timestamps and trust levels is generated, and the corresponding decay coefficient is recorded for each time point in the series.

[0010] For multiple talents queried by the same verifier within a time window, the trust time series of each talent is extracted, and the weighted cross-correlation number between the trust time series is calculated using the corresponding decay coefficient as the weight. When the mean of the conflict feature vector or the weighted cross-correlation number exceeds the corresponding threshold, an abnormal verification identifier is generated, and the abnormal verification identifier and the associated talent identity hash are written into the blockchain.

[0011] The credentials are constructed as a spatiotemporal constraint graph. By calculating the proportion of time interval intersection between nodes and the ratio of geographical distance to time interval, temporal and spatial conflicts are identified, and conflict feature vectors are generated, including:

[0012] Obtain multiple background credentials for the talent to be verified, each credential including a time range, geographical location, and credential type identifier;

[0013] The multiple background credentials are constructed as nodes, and connection edges are established based on the temporal dependencies, temporal mutual exclusions, and spatial reachability relationships between the credentials to form a spatiotemporal constraint graph.

[0014] For each background voucher, the time interval is labeled with a start timestamp and an end timestamp, and a time interval vector is constructed. The ratio of the intersection length of the time intervals of any two voucher nodes to the shorter time interval length of the two voucher nodes is calculated as the temporal conflict intensity. When the temporal conflict intensity exceeds the temporal conflict threshold, it is marked as a temporal conflict edge.

[0015] Extract voucher node pairs with temporal adjacency, calculate the geographical distance between the ending geographical location of the preceding voucher and the starting geographical location of the subsequent voucher, divide the geographical distance by the time interval between the two vouchers to obtain the spatial transfer rate, and mark the spatial transfer rate as a spatial conflict edge when it exceeds the physical reachability limit.

[0016] The number of temporal conflict edges, spatial conflict edges, and the number of edges with broken dependencies in the spatiotemporal constraint graph are counted to construct a multidimensional conflict feature vector.

[0017] On the blockchain, record the hash identifier of each background credential's preceding dependent credentials, constructing a directed acyclic graph of credential dependencies, including:

[0018] For each background credential, its dependency rule is determined based on the credential type identifier. The dependency rule defines the set of preceding credential types required for the credential type. Credentials that match the set of preceding credential types in the background credential set of the talent to be verified are retrieved. The hash identifier of the preceding credential is extracted and recorded in the on-chain data structure of the current background credential.

[0019] Traverse the hash identifiers of all background credentials' preceding dependent credentials, treat each background credential as a node in a directed acyclic graph (DAG), and treat the dependency relationship between a background credential and its preceding dependent credentials as directed edges, with the direction of the directed edges pointing from the preceding dependent credentials to the current background credential. Perform topological sorting verification on the DAG to ensure that there are no cycles in the graph, generate a credential dependency DAG, and write the topological structure identifier of the credential dependency DAG into the blockchain.

[0020] The existence of substitute vouchers of the same type is detected, and the trust level of descendant vouchers is updated based on the existence by assigning differentiated decay coefficients. A trust level time series containing the correspondence between timestamps and trust levels is generated, and the corresponding decay coefficient is recorded for each time point in the series, including:

[0021] When a document is revoked, all descendant documents are traversed along the outgoing edges of the directed acyclic graph.

[0022] Calculate the shortest path distance between the descendant certificate and the revoked certificate in the directed acyclic graph, count the number of all distinct paths from the revoked certificate to the descendant certificate, and use the product of the inverse of the shortest path distance and the number of paths as the dependency strength weight.

[0023] For each descendant voucher, retrieve other vouchers with the same voucher type identifier as the descendant voucher, calculate the ratio of the intersection length to the union length of the time intervals of the other vouchers and the time intervals of the descendant vouchers as the time overlap. When the time overlap is greater than zero, select the other voucher with the largest time overlap as the substitute voucher of the same type and use the largest time overlap as the substitution compensation coefficient. When the time overlap is zero or there are no other vouchers, set the substitution compensation coefficient to zero.

[0024] The differential decay coefficient is calculated based on the dependency strength weight and the substitution compensation coefficient. The current trust level of the descendant certificate is multiplied by the differential decay coefficient to obtain the updated trust level. The update timestamp, the updated trust level, and the differential decay coefficient are recorded to construct a trust level time series.

[0025] The differential attenuation coefficient is calculated based on the dependence strength weight and the substitution compensation coefficient, including:

[0026] The base decay value corresponding to the voucher type is obtained by querying the preset voucher importance level table based on the voucher type identifier of the revoked voucher, and multiplying the base decay value by the dependency strength weight to obtain the dependency influence component.

[0027] Extract the issuing authority identifier of the same type of substitute certificate, query the institution's credit rating based on the issuing authority identifier, map the institution's credit rating to a compensation adjustment factor, and multiply the substitute compensation coefficient by the compensation adjustment factor to obtain the substitute offset component.

[0028] Calculate the time difference between the revocation timestamp of the revoked certificate and the current update timestamp, map the time difference to a time decay factor, subtract the substitution offset component from the dependent influence component, and then multiply by the time decay factor to obtain the differential decay coefficient.

[0029] Extract the trust level time series of each talent, and use the corresponding decay coefficient as weight to calculate the weighted cross-correlation coefficient between the trust level time series, including:

[0030] For multiple talents queried by the same verification party within a time window, a time alignment operation is performed on the trust time series of each talent. Taking the start time of the time window as the reference time point, the timestamps in the trust time series of each talent are converted into relative time offsets relative to the reference time point. Time resampling is performed on the converted multiple trust time series to obtain an aligned trust time series with a uniform time interval.

[0031] For each time sampling point in the alignment trust time series, extract the trust value of each talent at that time point and its corresponding decay coefficient, calculate the reciprocal of the decay coefficient of each talent at that time point as a weight factor and perform normalization processing to obtain normalized weights.

[0032] For any two talents' alignment trust time series, at each sampling time, the corresponding trust value is multiplied by the normalized weight to obtain a weighted trust value. Based on the weighted trust value, the similarity measure between the weighted trust time series of the two talents is calculated as a weighted cross-correlation number. The weighted cross-correlation number is used to characterize the degree of correlation between the trust evolution patterns of the two talents in the time dimension.

[0033] When the mean of the conflict feature vector or the weighted cross-correlation coefficient exceeds the corresponding threshold, an anomaly verification identifier is generated, and the anomaly verification identifier and the associated talent identity hash are written into the blockchain, including:

[0034] The arithmetic mean of each dimension component in the conflict feature vector is calculated as the conflict intensity index, and the maximum value of the weighted cross-correlation coefficient is calculated as the collusion similarity index. The conflict intensity index is compared with the conflict determination threshold, and the collusion similarity index is compared with the collusion determination threshold. Based on the comparison results, the index type that triggers the anomaly is determined, and the talent identity identifiers associated with the conflict feature vector or weighted cross-correlation coefficient that triggers the anomaly are extracted to form a talent identity identifier set.

[0035] Obtain the verifier identifier corresponding to the verifier that triggered the anomaly determination; set an anomaly type field according to the indicator type; calculate an anomaly degree field based on the magnitude by which the indicator value that triggered the anomaly exceeds the corresponding threshold; construct an anomaly verification identifier by combining the current timestamp; and combine each talent identity identifier in the talent identity identifier set with the verifier identifier to form an anomaly association pair.

[0036] Perform a hash operation on the talent identity identifier in the abnormal association pair to generate a talent identity hash, and encapsulate the abnormal verification identifier and the talent identity hash into an abnormal verification transaction and write it into the blockchain.

[0037] Based on the anomaly verification identifier, the spatiotemporal constraint graph of each talent in the talent identity identifier set is re-verified and the conflict feature vector is recalculated. When the number of non-zero components in the recalculated conflict feature vector is greater than the number of non-zero components in the initial judgment, an anomaly confirmation identifier is generated and written into the blockchain.

[0038] A second aspect of this invention provides a blockchain-based talent background verification system, comprising:

[0039] The credential acquisition unit is used to acquire multiple background credentials for the talent to be verified. Each background credential includes a time range, geographical location, and credential type identifier.

[0040] The conflict identification unit is used to construct the credentials into a spatiotemporal constraint map. By calculating the proportion of time interval intersection between nodes and the ratio of geographical distance to time interval, it identifies temporal conflicts and spatial conflicts and generates conflict feature vectors.

[0041] Dependency building blocks are used to record the hash identifiers of the preceding dependent credentials for each background credential on the blockchain, constructing a directed acyclic graph of credential dependencies.

[0042] The trust update unit is used to traverse all descendant vouchers along the outgoing edges of the directed acyclic graph when a voucher is revoked, detect the existence of alternative vouchers of the same type, update the trust level of the descendant vouchers by assigning a differentiated decay coefficient according to the existence, generate a trust level time series containing the correspondence between timestamps and trust levels, and record the corresponding decay coefficient for each time point in the series.

[0043] The weighted correlation unit is used to extract the trust time series of each talent for multiple talents queried by the same verifier within a time window, and to calculate the weighted cross-correlation number between the trust time series using the corresponding decay coefficient as the weight.

[0044] An anomaly recording unit is used to generate an anomaly verification identifier when the mean of the conflict feature vector or the weighted cross-correlation coefficient exceeds the corresponding threshold, and to write the anomaly verification identifier and the associated talent identity hash into the blockchain.

[0045] A third aspect of the present invention provides an electronic device, comprising:

[0046] processor;

[0047] Memory used to store processor-executable instructions;

[0048] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0049] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0050] This method significantly improves the accuracy and reliability of talent background verification. By constructing a spatiotemporal constraint graph and calculating the ratio of the intersection of time intervals between nodes to the geographical distance time interval, it can intelligently identify temporal and spatial conflicts and automatically generate conflict feature vectors. This effectively identifies forged or contradictory work experience and educational background credentials, avoiding oversights and subjective biases in manual verification. By associating the hash identifiers of background credentials to construct a dependent directed acyclic graph, a dynamic trust update mechanism is implemented. This mechanism traverses all descendant credentials along the outgoing edges when a credential is revoked and assigns a differentiated decay coefficient based on the existence of similar substitute credentials. This generates a trust time series containing the correspondence between timestamps and trust levels, allowing the trust level to be adjusted in real time as credential changes occur, eliminating the drawback of the entire verification result becoming invalid due to the revocation of a single credential.

[0051] This method significantly enhances the tamper-proof capabilities and transparency of the verification process. All background credential data is anchored to the blockchain, and the dependencies between credentials are irreversibly recorded in the form of hash identifiers. Any modification to historical credentials will cause a break in the directed acyclic graph structure and trigger abnormal fluctuations in trust levels, which will be automatically captured by the system. Simultaneously, the trust level at each point in time and its corresponding decay coefficient are fully recorded in the trust time series, making the verification process completely traceable. The verifier can review the trust calculation basis at any point in time at any time, fundamentally guaranteeing the authenticity and non-repudiation of the talent's background information.

[0052] This method significantly improves the efficiency of identifying abnormal behavior across talent pools and time windows. For multiple talents queried by the same verifier within a time window, the trust time series of each talent are extracted, and a weighted cross-correlation coefficient is calculated using the corresponding decay coefficient as weights. This accurately identifies abnormal association patterns between talent background verification requests, such as repeated queries by the same institution for different but similar talents, or a cluster of sudden changes in the trust sequence of a particular talent. The mean of the conflict feature vector and the weighted cross-correlation coefficient are compared with corresponding thresholds. Once the threshold is exceeded, an abnormal verification identifier is immediately generated and written to the blockchain, achieving automated and real-time risk warnings and significantly reducing fraud risks and operational costs during the verification process. Attached Figure Description

[0053] Figure 1 A flowchart illustrating a blockchain-based talent background verification method;

[0054] Figure 2 A flowchart for calculating the differential decay coefficient of trust in descendant documents based on document revocation and constructing a trust time series. Detailed Implementation

[0055] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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.

[0056] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes will not be repeated in some embodiments.

[0057] Figure 1 This is a flowchart illustrating the blockchain-based talent background verification method according to an embodiment of the present invention.

[0058] Blockchain-based methods for verifying talent backgrounds include:

[0059] Obtain multiple background credentials for the talent to be verified, each credential including a time range, geographical location, and credential type identifier;

[0060] The credentials are constructed as a spatiotemporal constraint graph. By calculating the proportion of time interval intersection between nodes and the ratio of geographical distance to time interval, temporal conflicts and spatial conflicts are identified, and conflict feature vectors are generated.

[0061] On the blockchain, the hash identifier of the preceding dependent certificate is recorded for each background certificate, and a directed acyclic graph of certificate dependencies is constructed. When a certificate is revoked, all descendant certificates are traversed along the outgoing edges of the directed acyclic graph, the existence of alternative certificates of the same type is detected, and the trust level of the descendant certificate is updated according to the existence by assigning a differentiated decay coefficient. A trust level time series containing the correspondence between timestamps and trust levels is generated, and the corresponding decay coefficient is recorded for each time point in the series.

[0062] For multiple talents queried by the same verification party within a time window, the trust time series of each talent is extracted, and the weighted cross-relationship number between the trust time series is calculated using the corresponding decay coefficient as the weight.

[0063] When the mean of the conflict feature vector or the weighted cross-correlation coefficient exceeds the corresponding threshold, an anomaly verification identifier is generated, and the anomaly verification identifier and the associated talent identity hash are written into the blockchain.

[0064] In one optional implementation, the credentials are constructed as a spatiotemporal constraint graph. By calculating the proportion of time interval intersections between nodes and the ratio of geographical distance to time interval, temporal conflicts and spatial conflicts are identified, and conflict feature vectors are generated, including:

[0065] Obtain multiple background credentials for the talent to be verified, each credential including a time range, geographical location, and credential type identifier;

[0066] The multiple background credentials are constructed as nodes, and connection edges are established based on the temporal dependencies, temporal mutual exclusions, and spatial reachability relationships between the credentials to form a spatiotemporal constraint graph.

[0067] For each background voucher, the time interval is labeled with a start timestamp and an end timestamp, and a time interval vector is constructed. The ratio of the intersection length of the time intervals of any two voucher nodes to the shorter time interval length of the two voucher nodes is calculated as the temporal conflict intensity. When the temporal conflict intensity exceeds the temporal conflict threshold, it is marked as a temporal conflict edge.

[0068] Extract voucher node pairs with temporal adjacency, calculate the geographical distance between the ending geographical location of the preceding voucher and the starting geographical location of the subsequent voucher, divide the geographical distance by the time interval between the two vouchers to obtain the spatial transfer rate, and mark the spatial transfer rate as a spatial conflict edge when it exceeds the physical reachability limit.

[0069] The number of temporal conflict edges, spatial conflict edges, and the number of edges with broken dependencies in the spatiotemporal constraint graph are counted to construct a multidimensional conflict feature vector.

[0070] For example, before constructing the spatiotemporal constraint graph, it is necessary to structure the multiple background credentials of the talent to be verified. Each background credential carries three key attributes: time interval, geographical location, and credential type identifier. The time interval records the start and end times of the activity corresponding to the credential, the geographical location describes the physical spatial coordinates associated with the credential (usually represented by latitude and longitude or administrative division codes), and the credential type identifier distinguishes different categories such as academic credentials, work experience credentials, and qualification certification credentials. After uniformly formatting these attributes, each background credential is mapped to a node in the graph, and the node attributes contain the above three types of information.

[0071] After the nodes are established, connecting edges are created based on the three types of relationships between credentials, thus forming a spatiotemporal constraint graph. Temporal dependency edges describe the logical requirement that one credential must exist before another, for example, a bachelor's degree credential is a prerequisite for a master's degree credential. Temporal mutual exclusion edges describe potential conflicts where two credentials overlap in time, such as the same person having two full-time job credentials within the same time period. Spatial reachability edges describe whether switching the geographical location of two credentials within adjacent time periods is physically feasible. The introduction of these three types of edges allows the graph to express both the logical constraints between credentials and conflict detection information in both temporal and spatial dimensions.

[0072] To calculate time-series conflicts, it is necessary to label the start and end timestamps of each voucher node's time interval, constructing a time interval vector. For any two voucher nodes... and Let their time intervals be respectively and ,in , vouchers The start and end timestamps, , vouchers The start and end timestamps. The intersection length of the two time intervals is... ,in This indicates the overlap length of the time intervals between the two vouchers. The length of the shorter time interval between the two vouchers is... ,in This represents the smaller of the two voucher time interval lengths. The timing conflict strength is defined as... ,in This represents the intensity of timing conflicts, and its value range is... .when Exceeding the preset timing conflict threshold At that time, and Edges between these edges are marked as time-series conflict edges. The value can be configured differently based on the voucher type. For example, for full-time employment vouchers of the same type, It can be set to a lower value (such as 0.05), while for voucher types that allow part-time work, The restrictions can be relaxed appropriately.

[0073] Spatial conflict detection is based on pairs of voucher nodes with temporal adjacency. Temporal adjacency means that the interval between the end timestamp of a preceding voucher and the start timestamp of a subsequent voucher is within a reasonable range, i.e., the two vouchers are closely connected on the timeline. For pairs of temporally adjacent voucher nodes... Extract preceding vouchers Ending geographical coordinates and subsequent vouchers Given the starting geographic coordinates, calculate the geographic distance between the two points. ,in This represents the actual distance (in kilometers) between the geographical locations of the two vouchers. Geographical distance is calculated using the Haversine formula or route estimation based on administrative divisions to ensure the accuracy of the distance measurement. Divide by the time interval between the two documents (in The spatial transfer rate is calculated by representing the time difference between the end time of the preceding document and the start time of the subsequent document (in hours). ,in This represents the equivalent speed (in kilometers per hour) at which a person moves between two vouchers. Exceeding the physical reach limit At that time, and The edges between them are marked as spatial conflict edges. The time interval Δt is used to set segments: when Δt is less than or equal to 2 hours... Set to 1200 km / h; when Δt is greater than 2 hours and less than or equal to 24 hours, The speed is set to 900 km / h; when Δt is greater than 24 hours, The speed is set to 600 km / h. This segmented setting mechanism is based on the following functional relationship: when Δt is within each interval, the determination is made directly according to the above fixed threshold; when more precise adjustments are needed, a linear interpolation function can be used within each interval. (Δt) = - k·Δt, where This represents the upper limit of the speed corresponding to the starting point of this interval, where k is the attenuation coefficient, ranging from 5 to 10 km / h², to ensure... It decreases monotonically as Δt increases.

[0074] The detection logic for broken dependency edges is combined with that for temporal dependency edges. On the blockchain, each background credential records the hash identifier of its preceding dependent credentials. If a credential declares a dependency on another credential, but the preceding credential node does not exist in the graph, has been revoked, or its time interval does not satisfy the constraint of being prior to the current credential, then the corresponding dependency edge is marked as a broken dependency edge. These broken edges reflect a logical incompleteness in the credential chain and are an important signal for identifying forged or missing credentials.

[0075] After marking the three types of conflict edges, count the number of temporally conflicting edges in the spatiotemporal constraint graph. The number of spatial conflict edges and the number of dependency break edges ,in This represents the total number of edges in the graph that are marked as temporal conflicts. This represents the total number of edges in the graph that are marked as spatial conflicts. This represents the total number of edges in the graph marked as dependency breaks. The three statistics are concatenated into a multidimensional conflict feature vector. ,in This represents a feature vector describing the overall conflict state of the current talent background credentials. This vector serves as input for subsequent anomaly detection; by calculating its mean or comparing it with a threshold, it is determined whether there are systematic anomalies in the current talent background credentials.

[0076] In practical applications, when the number of background credentials for candidates to be verified is large, the number of edges in the graph increases significantly with the number of nodes, leading to a rise in the computational cost of conflict detection. To address this, a pruning strategy can be introduced when constructing the spatiotemporal constraint graph: for credential node pairs whose time intervals do not overlap and whose time intervals exceed a set threshold, the calculation of temporal conflict intensity is skipped; for adjacent credential node pairs with the same geographical location or belonging to the same city, the calculation of spatial transfer rate is skipped. These optimization measures effectively reduce the computational overhead of graph construction and conflict identification while ensuring detection accuracy, making the overall verification process highly scalable in large-scale talent background verification scenarios.

[0077] In one alternative implementation, the hash identifier of the preceding dependent credentials is recorded on the blockchain for each background credential, constructing a directed acyclic graph of credential dependencies, including:

[0078] For each background credential, its dependency rule is determined based on the credential type identifier. The dependency rule defines the set of preceding credential types required for the credential type. Credentials that match the set of preceding credential types in the background credential set of the talent to be verified are retrieved. The hash identifier of the preceding credential is extracted and recorded in the on-chain data structure of the current background credential.

[0079] Traverse the hash identifiers of all background credentials' preceding dependent credentials, treat each background credential as a node in a directed acyclic graph (DAG), and treat the dependency relationship between a background credential and its preceding dependent credentials as directed edges, with the direction of the directed edges pointing from the preceding dependent credentials to the current background credential. Perform topological sorting verification on the DAG to ensure that there are no cycles in the graph, generate a credential dependency DAG, and write the topological structure identifier of the credential dependency DAG into the blockchain.

[0080] For example, in constructing a directed acyclic graph of credential dependencies, for each background credential, a predefined dependency rule base is first queried based on its credential type identifier. The dependency rule base stores credential types as keys and the required set of preceding credential types as values. For instance, the dependency rule for an academic qualification credential type requires the existence of an admission notice or prior academic qualification credential, while the dependency rule for a professional qualification credential type requires the existence of a corresponding academic qualification credential or training completion credential. By retrieving the set of preceding credential types required for the current credential type from the dependency rule base using the credential type identifier, credentials matching this set are then selected from the full set of background credentials for the talent to be verified. The selected results are used as the candidate set of preceding dependency credentials for the current credential.

[0081] For each preceding dependent credential in the candidate set, its hash identifier on the blockchain is extracted. This hash identifier is calculated using a digest algorithm (such as SHA-256) of the credential content, and is unique and immutable, serving as a globally unique pointer to the credential entity on the blockchain. The hash identifiers of all preceding dependent credentials are written in list form into the on-chain data structure corresponding to the current background credential. The on-chain data structure includes fields such as credential hash, credential type identifier, time interval, geographical location, preceding dependent hash list, and record timestamp. In this way, each background credential not only stores its own content digest on the chain but also explicitly records the on-chain anchor point of its dependent preceding credentials, forming a traceable chain of dependency relationships.

[0082] After recording the pre-dependency hash identifiers for all background credentials, the entire set of background credentials is traversed, mapping each background credential to a node in a directed acyclic graph (DAG). The unique identifier of each node is its hash value. For each credential node, the list of pre-dependency hashes stored in its on-chain data structure is read. For each pre-dependency hash in the list, a directed edge is added to the graph from the pre-dependency credential node to the current credential node. The direction of the directed edge is defined as "pre-dependency credential → current background credential," semantically indicating that the existence of the pre-dependency credential is a logical prerequisite for the validity of the current credential. Taking academic qualifications and professional qualification credentials as an example, if a professional qualification credential depends on an academic qualification credential, then there is a directed edge in the graph from the academic qualification credential node to the professional qualification credential node, intuitively expressing the logical order of "first the academic qualification, then the qualification."

[0083] After constructing the set of nodes and edges of the directed graph, a topological sorting verification is required to ensure that there are no cycles in the graph, thus guaranteeing the logical consistency of dependencies. Topological sorting employs an iterative algorithm based on in-degree: calculate the in-degree of each node in the graph, add all nodes with an in-degree of zero to a queue, remove a node from the queue each time, and decrement the in-degree of all its successor nodes by 1. If the in-degree of a successor node thus drops to zero, add it to the queue. Repeat this process until the queue is empty. If the total number of nodes processed equals the total number of nodes in the graph, there are no cycles, and the topological sorting is successful. If the total number of nodes processed is less than the total number of nodes in the graph, a cycle exists, indicating a logical contradiction in the cyclic dependency relationship. In this case, an anomaly flag is triggered, the set of credential nodes forming the cycle is recorded, and the count of broken edges in the dependency relationship is included in the statistics during the subsequent conflict feature vector construction stage.

[0084] Once the topological sorting verification is successful, the generated topological sequence itself has significant value: it determines the legitimate processing order of credential dependencies, providing a traversal basis for subsequent credential revocation propagation and trust decay calculations. Credential nodes at the beginning of the topological sequence form the foundation for other credentials, and their trust changes have a more profound impact on downstream credentials; credential nodes at the end of the dependency chain have a relatively limited scope of revocation impact. When the topological structure identifier of the directed acyclic graph of credential dependencies is written into the blockchain, the topological structure identifier is generated jointly by the graph's adjacency digest and the topological sorting sequence, ensuring that any tampering with the dependencies will invalidate the topological structure identifier, thus leaving a verifiable proof of integrity on the chain.

[0085] The on-chain stored topology identifiers organize all credential nodes and their dependent edge information in the form of a Merkle tree. The Merkle root hash is written as a unique summary of the entire directed acyclic graph (DAG) into a specific field of the blockchain. Any third-party verifier can independently verify the integrity of the credential dependency graph by recalculating the Merkle root hash and comparing it with the on-chain record, without trusting any intermediary. This design enables the credential dependency DAG to possess both on-chain immutable evidence storage capabilities and efficient local verification. Verifiers only need to provide the Merkle path from the node to be verified to the root node to complete the verification of the authenticity of a single dependency relationship in logarithmic time complexity.

[0086] In practical applications, the content of the dependency rule base can be configured and expanded according to the requirements of different industries or institutions. For example, the medical industry's physician licensing certificate relies on a medical undergraduate degree certificate and a certificate of completion of standardized residency training, while the financial industry's professional qualification certificate relies on relevant professional academic qualifications and qualification examination passing certificates. Dependency rules are deployed on the blockchain in the form of smart contracts. Adding or modifying rules requires consensus from multiple parties to ensure the authority and consistency of the rules themselves. When a new certificate type is introduced, the corresponding dependency rule is registered at the smart contract level, and all subsequent dependency relationships for that type of certificate automatically follow the new rule, without requiring retrospective modifications to historical certificates, ensuring the system's backward compatibility.

[0087] When multiple similar credentials exist in the candidate set of prerequisite dependencies, the dependency rules can further specify matching strategies, such as selecting the prerequisite credential with the most recent time interval, selecting the prerequisite credential with the highest trust level, or establishing dependency edges for all eligible credentials. Different matching strategies affect the edge density and structural complexity of the directed acyclic graph, thereby affecting the computational overhead of subsequent reversal propagation. In the system implementation, the principle of closest time interval is used by default for single prerequisite credential matching to maintain the simplicity of the graph structure; for high-risk credential types that require multiple dependency verification, multiple dependency strategies can be explicitly configured in the dependency rules to include all eligible prerequisite credentials in the directed edge set, thereby obtaining more comprehensive dependency chain coverage in the trust assessment stage.

[0088] In one optional implementation, the existence of substitute credentials of the same type is detected, and the trust level of descendant credentials is updated based on the existence by assigning a differentiated decay coefficient. A trust level time series containing the correspondence between timestamps and trust levels is generated, and the corresponding decay coefficient is recorded for each time point in the series, including:

[0089] When a document is revoked, all descendant documents are traversed along the outgoing edges of the directed acyclic graph.

[0090] Calculate the shortest path distance between the descendant certificate and the revoked certificate in the directed acyclic graph, count the number of all distinct paths from the revoked certificate to the descendant certificate, and use the product of the inverse of the shortest path distance and the number of paths as the dependency strength weight.

[0091] For each descendant voucher, retrieve other vouchers with the same voucher type identifier as the descendant voucher, calculate the ratio of the intersection length to the union length of the time intervals of the other vouchers and the time intervals of the descendant vouchers as the time overlap. When the time overlap is greater than zero, select the other voucher with the largest time overlap as the substitute voucher of the same type and use the largest time overlap as the substitution compensation coefficient. When the time overlap is zero or there are no other vouchers, set the substitution compensation coefficient to zero.

[0092] The differential decay coefficient is calculated based on the dependency strength weight and the substitution compensation coefficient. The current trust level of the descendant certificate is multiplied by the differential decay coefficient to obtain the updated trust level. The update timestamp, the updated trust level, and the differential decay coefficient are recorded to construct a trust level time series.

[0093] Combination Figure 2 This document describes the calculation of the differential decay coefficient of trust in descendant certificates based on certificate revocation and the construction of a trust time series. When a certificate revocation event is triggered, a breadth-first traversal is performed along all outgoing edges of the directed acyclic graph (DAG) starting from the revoked certificate, visiting all descendant certificate nodes layer by layer. During the traversal, the node identifier of each descendant certificate is recorded, and the complete path information from the revoked certificate to that descendant certificate is retained for subsequent calculation of dependency strength weights. The traversal terminates when the current node has no outgoing edges, i.e., a leaf node of the DAG has been reached.

[0094] For each descendant document that is traversed, the dependency strength weight between it and the revoked document needs to be calculated. Let the shortest path distance from the revoked document to a certain descendant document be... The number of all distinct paths from the revoked certificate to that descendant certificate is Then it depends on the strength weight. Defined as: Among them, the shortest path distance This reflects the number of layers in the dependency chain between descendant certificates and revoked certificates; a larger gap indicates a weaker direct dependency. (Number of paths) This reflects how many different dependency paths a descendant credential is affected by a revoked credential; more paths indicate a higher degree of dependency. The number of paths... Divide by the shortest path distance ,Right now This results in descendant credentials that are close in distance and have more dependent paths receiving higher dependency strength weights, thus experiencing greater attenuation in subsequent trust updates. In actual calculations, the shortest path distance... The number of paths is obtained by performing Dijkstra's algorithm or BFS level-order traversal on a directed acyclic graph. The results were obtained by counting and statistically analyzing all paths originating from the revoked voucher using dynamic programming.

[0095] After obtaining the dependency strength weights, alternative credentials of the same type are retrieved for each descendant credential. Specifically, using the credential type identifier of the descendant credential as the search key, other credentials with the same credential type identifier are searched in the full set of credentials stored on the blockchain, excluding the descendant credential itself and revoked credentials. For each other credential retrieved, the Jaccard similarity between its time interval and the time interval of the descendant credential is calculated, which is the ratio of the intersection length to the union length of the two time intervals, as the time overlap. Let the time interval of the descendant credential be... The time interval for candidate alternative vouchers is Then the intersection length Union length They are respectively: , Time overlap for: ;when When both voucher time intervals degenerate into a single point and overlap, Set to 1; when hour, A value of zero indicates that the two time intervals are completely disjoint. Calculate this separately for all candidate substitute vouchers. Afterwards, if there is at least one If the candidate credentials are not found, then select them. The largest candidate certificate is selected as a substitute certificate of the same type, and this maximum value is used as the substitution compensation coefficient. If all candidate credentials If all are zero, or if no other vouchers of the same type exist, the compensation coefficient will be replaced. Set to zero. Substitution compensation coefficient. The physical meaning is that if there are alternative vouchers with high time coverage, it means that the function of the revoked voucher has been effectively replaced in the time dimension, and the actual impact on the descendant vouchers is relatively small. Therefore, the degree of trust decay should be appropriately reduced.

[0096] Based on dependency strength weight and substitution compensation coefficient Calculate the basic differential decay coefficient corresponding to each descendant voucher. The design principle for the basic differential attenuation coefficient is: the higher the dependence strength ( The larger the value, the more severe the attenuation; the more sufficient the substitution compensation ( The closer to 1, the less severe the attenuation. The specific calculation method is as follows: ;in, This indicates the suppressive effect of substitution compensation on attenuation, when The time decay coefficient is zero, meaning that the substitute voucher completely covers the time interval of the descendant voucher, and the trust level does not decay. To normalize the mapping of dependency strength weights, from Mapped to This ensures that the decay coefficient remains within a reasonable range, preventing it from exceeding the effective range in extreme cases. The basic differentiated decay coefficient is applicable to simplified scenarios that do not consider the impact of time decay factors and institutional credit ratings. Let the current trust level of the descendant certificate before the certificate revocation event be... The updated trust level for: This formula shows that when the differential decay coefficient The larger the value, the greater the decrease in trust; when At that time, the level of trust remained unchanged.

[0097] After the trust level update is completed, the relevant information for this update event is written into the trust level time series. Each record contains three fields: update timestamp. (i.e., the block timestamp of the confirmed credential revocation event on the blockchain), and the updated trust level. And the differential attenuation coefficient calculated in this study. The trust time series uses the hash identifier of the descendant credential as the index key and is arranged in ascending order of timestamp, forming a complete historical record of the evolution of the credential's trust level over time. During subsequent queries by the verifier, the latest record in the trust time series can be directly retrieved. As the current level of trust, and using the corresponding time points in the sequence. As a weight input for calculating the weighted cross-correlation coefficient, it enables historical changes in trust level to support anomaly detection.

[0098] In scenarios with multiple reversal events, the trust time series will contain multiple update records, with each reversal event triggering an update of the previous record. As a new By substituting the values ​​into the calculation, the cumulative decay of trust is achieved. This design ensures that multiple anomalous events on the credential dependency chain can be fully tracked, and an auditable chain of evidence for the evolution of trust is formed under the guarantee of the blockchain's immutability.

[0099] In one optional implementation, calculating the differential attenuation coefficient based on the dependency strength weight and the substitution compensation coefficient includes:

[0100] The base decay value corresponding to the voucher type is obtained by querying the preset voucher importance level table based on the voucher type identifier of the revoked voucher, and multiplying the base decay value by the dependency strength weight to obtain the dependency influence component.

[0101] Extract the issuing authority identifier of the same type of substitute certificate, query the institution's credit rating based on the issuing authority identifier, map the institution's credit rating to a compensation adjustment factor, and multiply the substitute compensation coefficient by the compensation adjustment factor to obtain the substitute offset component.

[0102] Calculate the time difference between the revocation timestamp of the revoked certificate and the current update timestamp, map the time difference to a time decay factor, subtract the substitution offset component from the dependent influence component, and then multiply by the time decay factor to obtain the differential decay coefficient.

[0103] The calculation of the differential decay coefficient needs to comprehensively consider the importance of the revoked certificate itself, the quality level of the substitute certificate, and the time decay effect after the revocation event. To this end, the entire calculation process is decomposed into three interrelated components, which respectively capture the influence strength of certificate dependence, the degree of compensation for loss by the substitute certificate, and the dynamic decay characteristics in the time dimension, and finally integrate the three into a unified differential decay coefficient.

[0104] For example, based on the voucher type identifier of the revoked voucher, an exact match query is performed in a preset voucher importance level table to obtain the basic attenuation value corresponding to that voucher type. The credential importance ranking table is a mapping table pre-defined by domain experts and stored in an on-chain smart contract. Each credential type identifier (such as academic credentials, professional qualification credentials, work experience credentials, etc.) corresponds to a normalized interval. The base decay value indicates the degree of disruption to the dependency chain caused by the revocation of that type of certificate. For example, academic credentials, as prerequisites for many subsequent professional qualification certificates, are typically assigned a higher base decay value, while auxiliary training records correspond to a lower base decay value. Then, it is combined with the dependency strength weight. Multiply to obtain the dependent influence component. The calculation relationship is as follows . The product of these two factors, calculated based on path distance and path number during the document revocation propagation phase, reflects the degree of structural dependency between descendant documents and the revoked document. It comprehensively characterizes the combined effect of the importance of voucher type and the strength of topological dependency, and is the core driving component for the calculation of attenuation coefficient.

[0105] In the calculation of the substitution offset component, the issuing authority identifier of the same type of substitution certificate is extracted. This identifier is stored in the certificate metadata in a standardized institution code format. Based on this institution identifier, the on-chain institution credit rating database is queried to obtain the corresponding institution's credit rating value. Its value range is normalized to an interval. A higher value indicates greater authority and credibility of the issuing organization. Subsequently, a pre-defined monotonically increasing mapping function is used to... Converted to compensation regulation factor The mapping relationship is ,in The sensitivity parameter can be adjusted by the system administrator according to the actual business scenario; the default value is [value to be filled in]. This allows substitute certificates issued by high-credit institutions to achieve a stronger compensation effect, while substitute certificates issued by low-credit institutions only have a limited offsetting effect. The substitution compensation coefficient... With compensation regulator Multiplying them together yields the alternative offsetting components. The calculation relationship is as follows . The time overlap has already been determined in the previous steps. The calculations show that the substitute document's ability to compensate for descendant documents in terms of time coverage is reflected. This further introduces a modification to the quality dimension of the issuing authority, so that the substitution offset component not only considers the degree of time matching, but also takes into account the substantial impact of the authority of the institution on the effectiveness of the compensation.

[0106] The time decay factor is introduced to reflect an intuitively reasonable phenomenon: the influence of a revoked certificate gradually weakens over time, and the longer the interval after the revocation event, the more stable its impact on the trust level of descendant certificates becomes. The revocation timestamp of the revoked certificate is calculated. Compared to the current update timestamp Time difference between The unit is uniformly set to hours, and the calculation relationship is as follows: .Will Mapped to a time decay factor via an exponential decay function. Specifically ,in This is the time decay rate constant, controlling the steepness of the decay curve; the default value is [value to be filled in]. (The corresponding time unit is hours), which can be adjusted according to the validity period characteristics of background certificates in different industries. When When approaching zero, Approaching This indicates that the impact is most intense immediately after the event is revoked; as... Increase A monotonically decreasing trend approaching zero indicates that the impact of the cancellation gradually dissipates over time.

[0107] Combining the above three components, the complete differential attenuation coefficient The calculation relationship is as follows The dependent influence component. Subtract the substitute offset component The net attenuation intensity is obtained, and then multiplied by the time decay factor. This yields the final, complete differential attenuation coefficient. Based on the differential attenuation coefficient Based on the computational logic, this paper further introduces three dimensions of correction: document importance level, institutional credit rating, and time decay factor. This is suitable for complete verification scenarios that require refined consideration of document type importance, the quality of alternative document issuing institutions, and the time distance of revocation events. In actual system implementation, when a rapid response to document revocation events is required, this paper can be used... Perform an initial trust level update; when conducting a full audit or verifying high-risk credentials, adopt... To ensure assessment accuracy, both decay coefficients are recorded in the trust time series and distinguished by a type identifier field within the series records. In the simplified verification scenario, λ is set to... In the complete verification scenario, λ takes the value of .

[0108] In practical applications, the basic differential attenuation coefficient With complete differential attenuation coefficient They are suitable for different verification scenarios. Considering only the two core factors of dependence strength weight and substitution compensation coefficient α, the computational complexity is low, making it suitable for rapid response scenarios where credential revocation events are triggered in real time. In this scenario, the system needs to complete the batch update of the trust level of all descendant credentials within milliseconds to ensure the timeliness of blockchain transaction confirmation. exist Building upon this foundation, further modifications are introduced using three dimensions: document importance level β, institutional credit rating G, and time decay factor ϕ. This results in high computational complexity and is suitable for scenarios requiring high-precision assessment, such as formal background checks initiated by the verifier, compliance audits conducted by regulatory agencies, or evidence chain reconstruction during dispute arbitration. In each record of the trust time series, the decay coefficient type used for this update is explicitly recorded through a decay coefficient type identifier field, ensuring accurate reconstruction of the calculation logic during subsequent audits. When the same descendant document is repeatedly updated due to multiple revocation events, each update can select the appropriate decay coefficient type based on the application scenario at that time. or The system independently tracks the cumulative effects of the two attenuation coefficients and calculates them separately in the statistical analysis of the trust time series. The cumulative decay path and based on The cumulative decay path provides the verifier with a view of the evolution of trust at different granularities.

[0109] When the substitute voucher is of extremely high quality and has complete time coverage. near This causes the net attenuation strength to approach zero, meaning that the trust level of descendant documents is almost unaffected by revocation events; conversely, when there are no equivalent alternative documents or the alternative documents come from low-credit institutions, The net attenuation intensity is relatively small, mainly due to Under this system, descendant credentials will suffer a significant decrease in trust level. To prevent negative values ​​or results exceeding reasonable ranges in calculations, in actual implementation... Apply a cutoff constraint to ensure that its value always falls within the interval. If the calculation result is negative, it is set to zero; if it exceeds [a certain value], it is set to zero. Then set to .

[0110] At the engineering implementation level, both the credential importance ranking table and the institutional credit rating database are deployed on the blockchain in the form of smart contracts. Query operations are completed through on-chain function calls, ensuring the immutability and global consistency of the ranking and rating data. The calculation result of each differentiated decay coefficient, along with the corresponding timestamp... The block confirmation time, descendant credential identifiers, and values ​​of each intermediate component are all written as structured records into the on-chain trust time series, providing a complete audit trail for subsequent weighted cross-correlation coefficient calculations. This multi-component fusion calculation mechanism allows the attenuation coefficient to finely differentiate the actual impact of revocation events in different scenarios, avoiding the problem of excessive or insufficient penalties caused by applying a uniform fixed attenuation coefficient to all descendant credentials, and significantly improving the accuracy and fairness of talent background verification results.

[0111] In one optional implementation, the trust level time series of each talent is extracted, and the weighted cross-correlation coefficient between the trust level time series is calculated using the corresponding decay coefficient as weight, including:

[0112] For multiple talents queried by the same verification party within a time window, a time alignment operation is performed on the trust time series of each talent. Taking the start time of the time window as the reference time point, the timestamps in the trust time series of each talent are converted into relative time offsets relative to the reference time point. Time resampling is performed on the converted multiple trust time series to obtain an aligned trust time series with a uniform time interval.

[0113] For each time sampling point in the alignment trust time series, extract the trust value of each talent at that time point and its corresponding decay coefficient, calculate the reciprocal of the decay coefficient of each talent at that time point as a weight factor and perform normalization processing to obtain normalized weights.

[0114] For any two talents' alignment trust time series, at each sampling time, the corresponding trust value is multiplied by the normalized weight to obtain a weighted trust value. Based on the weighted trust value, the similarity measure between the weighted trust time series of the two talents is calculated as a weighted cross-correlation number. The weighted cross-correlation number is used to characterize the degree of correlation between the trust evolution patterns of the two talents in the time dimension.

[0115] For example, when multiple talents are queried by the same verifier within a time window, the trust time series of each talent suffers from misalignment in timestamp distribution due to differences in the timing of credential revocation events and trust update frequencies. To eliminate the impact of time base differences on subsequent correlation analysis, a time alignment operation is performed on the trust time series of each talent. This is done using the start time of the time window... As a baseline time point, for the first Each raw timestamp in the time series of trust levels of individual talents Calculate its relative to relative time offset ,in The unit remains consistent with the original timestamp (e.g., seconds or milliseconds). After this transformation, the trust time series of all talents are expanded with the starting point of the time window as the zero point, eliminating the absolute time deviation introduced by different query times.

[0116] After unifying the time base, the timing of trust update events for different talents still differs, and the sampling points of each sequence do not overlap on the relative time axis, making direct point-by-point comparison calculations impossible. Therefore, it is necessary to perform time resampling on the transformed trust time series to obtain aligned trust time series with a unified time interval. Specifically, a unified resampling interval is set. During the time window Generate an equally spaced sampling time sequence ,in The total length of the time window. The total number of sampling points satisfies For each talent's trust score at a given sampling time, if a record exists at that time, it is used directly; otherwise, a piecewise linear interpolation method is used, employing the two most recent known trust scores before and after that sampling time for interpolation estimation, thus ensuring that each talent's trust score sequence is consistent. Each sampling time point has a corresponding confidence value, forming an aligned confidence time series.

[0117] For each time sampling point in the alignment trust time series, at the th At the sampling time, extract the first... Trust value of individual talent and its corresponding attenuation coefficient Due to the attenuation coefficient The range of values ​​is The smaller the value, the greater the impact of the cancellation event on the level of trust at that moment, and the lower the reliability. Therefore, the reciprocal of the attenuation coefficient is used. As initial weighting factors, moments with smaller decay coefficients correspond to larger initial weights, reflecting the design intention to give higher attention to confidence data at high decay moments. To ensure the comparability of weights and eliminate dimensional differences, the first... All sampling times The initial weights of each talent are normalized. The calculation method is as follows: ;in This refers to the total number of talents queried by the same verification party within the same time window. A talent index is created. After normalization, the sum of the weights of all talents at the same sampling time equals 1, ensuring the numerical stability of subsequent weighted calculations.

[0118] For any two talents and The alignment trust time series, in the first At each sampling time point, the respective confidence value is multiplied by the corresponding normalized weight to obtain the weighted confidence value. and The weighted trust value incorporates the influence of the attenuation coefficient into the trust value itself, so that when the revocation event has a strong impact, the corresponding trust contribution is appropriately amplified, thereby highlighting the impact of the voucher revocation event on the trust evolution pattern in the correlation calculation.

[0119] Based on the weighted trust score, calculate the trust scores of the two talents. and The similarity measure between weighted trust time series is used as the weighted cross-relation number. In the calculation process, the mean of the two talent-weighted trust degree sequences is first calculated separately. and Then calculate the weighted cross-correlation coefficient: ; The range of values ​​is ,when When the value is close to 1, it indicates a high positive correlation between the weighted trust evolution patterns of the two individuals, meaning that their trust levels show a highly consistent upward or downward trend over time; when... When the value is close to -1, it indicates that the evolutionary patterns of the two are inversely correlated; when When the value is close to 0, it indicates that the evolution patterns of trust between the two are not significantly correlated over time.

[0120] In practical applications, if the weighted cross-correlation coefficients of multiple talents queried by the same verification party within the same time window are generally high, this phenomenon is statistically difficult to explain by the natural distribution of talent backgrounds, suggesting that these talents' background credentials are being forged in a coordinated manner or through mass order manipulation. Therefore, all talent pairs should be cross-correlation coefficients should be cross-correlation coefficients. After the values ​​are aggregated, they are compared with a preset correlation threshold. When the threshold is exceeded, the generation logic of the abnormal verification identifier is triggered, and the abnormal identifier and the associated talent identity hash are written into the blockchain to realize on-chain evidence storage and traceability of suspicious verification behavior.

[0121] Throughout the process, time alignment eliminates absolute time base differences, time resampling ensures point-by-point comparability of multiple sequences, the inverse weighting mechanism of decay coefficient incorporates the impact of voucher revocation into the correlation calculation, and the weighted cross-correlation coefficient characterizes the synergy of multiple people's trust changes from the perspective of the overall similarity of time evolution patterns, thus providing a quantitative basis for identifying batch abnormal verification behavior.

[0122] In one optional implementation, when the mean of the conflict feature vector or the weighted cross-correlation coefficient exceeds a corresponding threshold, an anomaly verification identifier is generated, and the anomaly verification identifier and the associated talent identity hash are written into the blockchain, including:

[0123] The arithmetic mean of each dimension component in the conflict feature vector is calculated as the conflict intensity index, and the maximum value of the weighted cross-correlation coefficient is calculated as the collusion similarity index. The conflict intensity index is compared with the conflict determination threshold, and the collusion similarity index is compared with the collusion determination threshold. Based on the comparison results, the index type that triggers the anomaly is determined, and the talent identity identifiers associated with the conflict feature vector or weighted cross-correlation coefficient that triggers the anomaly are extracted to form a talent identity identifier set.

[0124] Obtain the verifier identifier corresponding to the verifier that triggered the anomaly determination; set an anomaly type field according to the indicator type; calculate an anomaly degree field based on the magnitude by which the indicator value that triggered the anomaly exceeds the corresponding threshold; construct an anomaly verification identifier by combining the current timestamp; and combine each talent identity identifier in the talent identity identifier set with the verifier identifier to form an anomaly association pair.

[0125] Perform a hash operation on the talent identity identifier in the abnormal association pair to generate a talent identity hash, and encapsulate the abnormal verification identifier and the talent identity hash into an abnormal verification transaction and write it into the blockchain.

[0126] Based on the anomaly verification identifier, the spatiotemporal constraint graph of each talent in the talent identity identifier set is re-verified and the conflict feature vector is recalculated. When the number of non-zero components in the recalculated conflict feature vector is greater than the number of non-zero components in the initial judgment, an anomaly confirmation identifier is generated and written into the blockchain.

[0127] For example, after constructing the conflict feature vector and calculating the weighted cross-correlation coefficient, it is necessary to compare the two types of anomaly indicators with their corresponding thresholds to determine whether the anomaly verification process is triggered. For the conflict feature vector, the values ​​of each dimension component are extracted, and an arithmetic mean is calculated for all dimension components to obtain the conflict intensity index. The calculation method is as follows ,in The total dimension of the conflict feature vectors. For the first The values ​​of each dimension component. For the weighted cross-correlation coefficient, the combination of all talents is obtained by iterating through all talents. The set is used as the collusion similarity index. ,Right now .Will Conflict determination threshold Compare, and at the same time Collusion detection threshold Compare the indicators and determine the type of indicator that triggered the anomaly based on the comparison results: if only Exceed If only, then the indicator type is conflict type; if only Exceed If both exceed the corresponding threshold, the indicator type is classified as collusive; if both exceed the corresponding threshold, the indicator type is classified as composite.

[0128] After determining the type of indicator that triggers the anomaly, extract the talent identity identifiers directly associated with the triggering anomaly. For conflict-type anomalies, extract the talent node identity identifiers involved in the conflict edges from the spatiotemporal constraint graph nodes corresponding to the conflict feature vectors; for collusion-type anomalies, extract the talent node identity identifiers involved in the conflict edges from the spatiotemporal constraint graph nodes corresponding to the conflict feature vectors. achieve Talented people Extract the corresponding talent identity identifiers; for composite anomalies, merge the talent identity identifiers from the two sources mentioned above, and after deduplication, form a set of talent identity identifiers. The construction of the talent identity set must ensure coverage of all talents directly involved in anomaly detection, avoiding omissions of related entities.

[0129] Get the verifier identifier corresponding to the verifier that triggered this anomaly determination. This identifier originates from the verification party registration record on the blockchain and is globally unique. An anomaly type field is set based on the determined indicator type. The value can be one of conflict type, collusion type, or a combination thereof. The anomaly severity field is calculated based on the magnitude by which the triggering indicator value exceeds the corresponding threshold. For conflict-type, For collusion type, For composite types, the sum of the excess values ​​of the two types is taken as the average. This quantifies the severity of the anomaly. It is combined with the current timestamp. Exception type field Exception level field Verification party identifier and current timestamp Jointly encapsulate and construct an exception verification identifier. The anomaly verification identifier uses a structured coding method to ensure its uniqueness and traceability on the blockchain.

[0130] Set up talent identity identifiers Each talent identity identifier and verification identifier By combining them one by one, a set of abnormal association pairs is formed. Each anomaly association pair is in the form of ,in for Middle Individual talent identification. The purpose of constructing abnormal association pairs is to clearly record which talents were identified as abnormal association subjects in which verification party's query behavior, providing a complete association link for subsequent tracing and auditing.

[0131] For each talent identity identifier in the abnormal association pair Perform a hash operation to generate the corresponding talent identity hash. The hash algorithm uses a standard algorithm consistent with the blockchain platform. Anomaly verification flags are then used. With all talent identity hashes Encapsulated as an exception verification transaction The transaction is then submitted to the blockchain network for broadcast and confirmation. Once the transaction is written to the blockchain, the correspondence between the anomaly verification identifier and the talent's identity hash is permanently recorded. Any participant can verify the authenticity and integrity of this record through the blockchain query interface, and the record cannot be tampered with.

[0132] After the abnormal verification transaction is recorded on the blockchain, it triggers the collection of talent identity identifiers. The process for re-verifying the spatiotemporal constraint graphs of various talents involves: for each talent, reloading all their background credentials, recalculating the proportion of time interval intersections and the ratio of geographical distance to time intervals between nodes according to the construction rules of the spatiotemporal constraint graph, re-identifying temporal and spatial conflict edges, re-counting dependency break edges, and generating new multidimensional conflict feature vectors. During the re-verification process, all recorded credential revocation events and trust update records on the blockchain are fully synchronized to ensure that the data state used for recalculation is consistent with the latest state of the blockchain, thus avoiding duplicate or missed judgments due to data lag.

[0133] The conflict feature vector obtained by statistical recalculation The number of non-zero components Conflicting feature vectors with those from the initial determination The number of non-zero components Comparison. When If this occurs, it indicates that re-verification has uncovered new conflicting dimensions that were not identified during the initial assessment, making the anomaly more severe. In this case, an anomaly confirmation flag is generated. The hash of the individual's identity, along with the corresponding talent identity hash and a recalculated conflict feature vector digest, is written into the blockchain to form an anomaly confirmation transaction. The generation of the anomaly confirmation identifier indicates that the talent's background credentials have undergone secondary verification and confirmation of multi-dimensional conflicts, thus possessing higher credibility and serving as the trigger for subsequent manual review or automated processing. At that time, no abnormal confirmation flag is generated, but the abnormal verification flag initially uploaded to the chain is still valid and retained for subsequent audit reference. The entire abnormal verification and confirmation process ensures the fairness and traceability of the verification results through the immutability of the blockchain, effectively preventing the forgery of talent background certificates and collusion between verification parties.

[0134] A second aspect of this invention provides a blockchain-based talent background verification system, comprising:

[0135] The credential acquisition unit is used to acquire multiple background credentials for the talent to be verified. Each background credential includes a time range, geographical location, and credential type identifier.

[0136] The conflict identification unit is used to construct the credentials into a spatiotemporal constraint map. By calculating the proportion of time interval intersection between nodes and the ratio of geographical distance to time interval, it identifies temporal conflicts and spatial conflicts and generates conflict feature vectors.

[0137] Dependency building blocks are used to record the hash identifiers of the preceding dependent credentials for each background credential on the blockchain, constructing a directed acyclic graph of credential dependencies.

[0138] The trust update unit is used to traverse all descendant vouchers along the outgoing edges of the directed acyclic graph when a voucher is revoked, detect the existence of alternative vouchers of the same type, update the trust level of the descendant vouchers by assigning a differentiated decay coefficient according to the existence, generate a trust level time series containing the correspondence between timestamps and trust levels, and record the corresponding decay coefficient for each time point in the series.

[0139] The weighted correlation unit is used to extract the trust time series of each talent for multiple talents queried by the same verifier within a time window, and to calculate the weighted cross-correlation number between the trust time series using the corresponding decay coefficient as the weight.

[0140] An anomaly recording unit is used to generate an anomaly verification identifier when the mean of the conflict feature vector or the weighted cross-correlation coefficient exceeds the corresponding threshold, and to write the anomaly verification identifier and the associated talent identity hash into the blockchain.

[0141] A third aspect of the present invention provides an electronic device, comprising:

[0142] processor;

[0143] Memory used to store processor-executable instructions;

[0144] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0145] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0146] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

Claims

1. A blockchain-based method for verifying talent background, characterized in that, include: Obtain multiple background credentials for the talent to be verified, each credential including a time range, geographical location, and credential type identifier; The credentials are constructed as a spatiotemporal constraint graph. By calculating the proportion of time interval intersection between nodes and the ratio of geographical distance to time interval, temporal conflicts and spatial conflicts are identified, and conflict feature vectors are generated. On the blockchain, the hash identifier of the preceding dependent certificate is recorded for each background certificate, and a directed acyclic graph of certificate dependencies is constructed. When a certificate is revoked, all descendant certificates are traversed along the outgoing edges of the directed acyclic graph, the existence of alternative certificates of the same type is detected, and the trust level of the descendant certificate is updated according to the existence by assigning a differentiated decay coefficient. A trust level time series containing the correspondence between timestamps and trust levels is generated, and the corresponding decay coefficient is recorded for each time point in the series. For multiple talents queried by the same verifier within a time window, the trust time series of each talent is extracted, and the weighted cross-correlation number between the trust time series is calculated using the corresponding decay coefficient as the weight. When the mean of the conflict feature vector or the weighted cross-correlation number exceeds the corresponding threshold, an abnormal verification identifier is generated, and the abnormal verification identifier and the associated talent identity hash are written into the blockchain.

2. The method according to claim 1, characterized in that, By calculating the proportion of time interval intersection between nodes and the ratio of geographical distance to time interval, temporal conflicts and spatial conflicts are identified, and conflict feature vectors are generated, including: For each background voucher, the time interval is labeled with a start timestamp and an end timestamp, and a time interval vector is constructed. The ratio of the intersection length of the time intervals of any two voucher nodes to the shorter time interval length of the two voucher nodes is calculated as the temporal conflict intensity. When the temporal conflict intensity exceeds the temporal conflict threshold, it is marked as a temporal conflict edge. Extract voucher node pairs with temporal adjacency, calculate the geographical distance between the ending geographical location of the preceding voucher and the starting geographical location of the subsequent voucher, divide the geographical distance by the time interval between the two vouchers to obtain the spatial transfer rate, and mark the spatial transfer rate as a spatial conflict edge when it exceeds the physical reachability limit. The number of temporal conflict edges, spatial conflict edges, and the number of edges with broken dependencies in the spatiotemporal constraint graph are counted to construct a multidimensional conflict feature vector.

3. The method according to claim 1, characterized in that, On the blockchain, record the hash identifier of each background credential's preceding dependent credentials, constructing a directed acyclic graph of credential dependencies, including: For each background credential, its dependency rule is determined based on the credential type identifier. The dependency rule defines the set of preceding credential types required for the credential type. Credentials that match the set of preceding credential types in the background credential set of the talent to be verified are retrieved. The hash identifier of the preceding credential is extracted and recorded in the on-chain data structure of the current background credential. Traverse the hash identifiers of all background credentials' preceding dependent credentials, treat each background credential as a node in a directed acyclic graph, and treat the dependency relationship between a background credential and its preceding dependent credentials as directed edges, with the direction of the directed edges pointing from the preceding dependent credentials to the current background credential. Perform topological sorting verification on the directed acyclic graph to generate a credential dependency directed acyclic graph.

4. The method according to claim 1, characterized in that, The existence of substitute vouchers of the same type is detected, and the trust level of descendant vouchers is updated based on the existence by assigning differentiated decay coefficients. A trust level time series containing the correspondence between timestamps and trust levels is generated, and the corresponding decay coefficient is recorded for each time point in the series, including: Calculate the shortest path distance between the descendant certificate and the revoked certificate in the directed acyclic graph, count the number of all distinct paths from the revoked certificate to the descendant certificate, and use the product of the inverse of the shortest path distance and the number of paths as the dependency strength weight. For each descendant voucher, retrieve other vouchers with the same voucher type identifier as the descendant voucher, calculate the ratio of the intersection length to the union length of the time intervals of the other vouchers and the time intervals of the descendant vouchers as the time overlap. When the time overlap is greater than zero, select the other voucher with the largest time overlap as the substitute voucher of the same type and use the largest time overlap as the substitution compensation coefficient. When the time overlap is zero or there are no other vouchers, set the substitution compensation coefficient to zero. The differential decay coefficient is calculated based on the dependency strength weight and the substitution compensation coefficient. The current trust level of the descendant certificate is multiplied by the differential decay coefficient to obtain the updated trust level. The update timestamp, the updated trust level, and the differential decay coefficient are recorded to construct a trust level time series.

5. The method according to claim 4, characterized in that, The differential attenuation coefficient is calculated based on the dependence strength weight and the substitution compensation coefficient, including: The base decay value corresponding to the voucher type is obtained by querying the preset voucher importance level table based on the voucher type identifier of the revoked voucher, and multiplying the base decay value by the dependency strength weight to obtain the dependency influence component. Extract the issuing authority identifier of the same type of substitute certificate, query the institution's credit rating based on the issuing authority identifier, map the institution's credit rating to a compensation adjustment factor, and multiply the substitute compensation coefficient by the compensation adjustment factor to obtain the substitute offset component. Calculate the time difference between the revocation timestamp of the revoked certificate and the current update timestamp, map the time difference to a time decay factor, subtract the substitution offset component from the dependent influence component, and then multiply by the time decay factor to obtain the differential decay coefficient.

6. The method according to claim 1, characterized in that, Extract the trust level time series of each talent, and use the corresponding decay coefficient as weight to calculate the weighted cross-correlation coefficient between the trust level time series, including: For multiple talents queried by the same verification party within a time window, a time alignment operation is performed on the trust time series of each talent. Taking the start time of the time window as the reference time point, the timestamps in the trust time series of each talent are converted into relative time offsets relative to the reference time point. Time resampling is performed on the converted multiple trust time series to obtain an aligned trust time series with a uniform time interval. For each time sampling point in the alignment trust time series, extract the trust value of each talent at that time point and its corresponding decay coefficient, calculate the reciprocal of the decay coefficient of each talent at that time point as a weight factor and perform normalization processing to obtain normalized weights. For any two talents' alignment trust time series, at each sampling time, the corresponding trust value is multiplied by the normalized weight to obtain a weighted trust value. Based on the weighted trust value, the similarity measure between the two talents' weighted trust time series is calculated as the weighted cross-correlation coefficient.

7. The method according to claim 1, characterized in that, When the mean of the conflict feature vector or the weighted cross-correlation coefficient exceeds the corresponding threshold, an anomaly verification identifier is generated, and the anomaly verification identifier and the associated talent identity hash are written into the blockchain, including: The arithmetic mean of each dimension component in the conflict feature vector is calculated as the conflict intensity index, and the maximum value of the weighted cross-correlation coefficient is calculated as the collusion similarity index. The conflict intensity index is compared with the conflict determination threshold, and the collusion similarity index is compared with the collusion determination threshold. Based on the comparison results, the index type that triggers the anomaly is determined, and the talent identity identifiers associated with the conflict feature vector or weighted cross-correlation coefficient that triggers the anomaly are extracted to form a talent identity identifier set. Obtain the verifier identifier corresponding to the verifier that triggered the anomaly determination; set an anomaly type field according to the indicator type; calculate an anomaly degree field based on the magnitude by which the indicator value that triggered the anomaly exceeds the corresponding threshold; construct an anomaly verification identifier by combining the current timestamp; and combine each talent identity identifier in the talent identity identifier set with the verifier identifier to form an anomaly association pair. Perform a hash operation on the talent identity identifier in the abnormal association pair to generate a talent identity hash, and encapsulate the abnormal verification identifier and the talent identity hash into an abnormal verification transaction and write it into the blockchain. Based on the anomaly verification identifier, the spatiotemporal constraint graph of each talent in the talent identity identifier set is re-verified and the conflict feature vector is recalculated. When the number of non-zero components in the recalculated conflict feature vector is greater than the number of non-zero components in the initial judgment, an anomaly confirmation identifier is generated and written into the blockchain.

8. A blockchain-based talent background verification system, used to implement the method as described in any one of claims 1-7, characterized in that, include: The credential acquisition unit is used to acquire multiple background credentials for the talent to be verified. Each background credential includes a time range, geographical location, and credential type identifier. The conflict identification unit is used to construct the credentials into a spatiotemporal constraint map. By calculating the proportion of time interval intersection between nodes and the ratio of geographical distance to time interval, it identifies temporal conflicts and spatial conflicts and generates conflict feature vectors. Dependency building blocks are used to record the hash identifiers of the preceding dependent credentials for each background credential on the blockchain, constructing a directed acyclic graph of credential dependencies. The trust update unit is used to traverse all descendant vouchers along the outgoing edges of the directed acyclic graph when a voucher is revoked, detect the existence of alternative vouchers of the same type, update the trust level of the descendant vouchers by assigning a differentiated decay coefficient according to the existence, generate a trust level time series containing the correspondence between timestamps and trust levels, and record the corresponding decay coefficient for each time point in the series. The weighted correlation unit is used to extract the trust time series of each talent queried by the same verifier within a time window, and calculate the weighted cross-correlation number between the trust time series using the corresponding decay coefficient as the weight. An anomaly recording unit is used to generate an anomaly verification identifier when the mean of the conflict feature vector or the weighted cross-correlation coefficient exceeds the corresponding threshold, and to write the anomaly verification identifier and the associated talent identity hash into the blockchain.

9. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 7.

Images