A cross-border digital certificate mutual recognition method based on dynamic trust anchor

Abstract

This application relates to a method for cross-border digital certificate mutual recognition based on dynamic trust anchors, comprising: collecting and preprocessing multi-dimensional trust assessment data of cross-border certificate authorities; collaboratively training a global model using federated XGBoost combined with a scenario-based attention mechanism to output a real-time trust score; constructing a hierarchical dynamic trust chain based on the score and writing the evidence data into the blockchain; generating proofs based on the Groth16 algorithm to complete certificate verification under privacy protection; deploying smart contracts to execute trust warnings, dispute arbitration, and rule updates; monitoring score changes, adaptively adjusting, and triggering an abnormal circuit breaker mechanism. This invention achieves real-time updating of trust weights and dynamic reconstruction of the trust chain, improving the efficiency and security of cross-border mutual recognition while ensuring data privacy compliance.

Description

Technical Field

[0001] This application relates to the fields of digital certificate authentication, federated learning, blockchain and zero-knowledge proof technology, and in particular to a cross-border digital certificate mutual recognition method based on dynamic trust anchors. Background Technology

[0002] In the context of a globalized digital economy, cross-border digital certificate mutual recognition serves as a crucial infrastructure supporting trusted interactions in international trade, finance, and government affairs. Its technological implementation directly impacts the security and efficiency of cross-border transactions. Currently, cross-border digital certificate mutual recognition is primarily based on the Public Key Infrastructure (PKI) system, constructing a trust chain through static trust anchors and employing cross-certification or trust list mechanisms to achieve cross-domain trust transfer between certificate authorities in different countries or regions. Cross-certification establishes a two-way trust binding by mutually issuing certificates to each other, incorporating the other party into its own root certificate trust system to achieve cross-domain certificate recognition. Trust lists, on the other hand, provide limited dynamic control over the institutions included in the mutual recognition by periodically publishing the trust status information of certificate authorities.

[0003] However, the existing solutions mentioned above have several technical shortcomings in practical applications. The cross-certification mechanism suffers from poor scalability of its trust network. With n certificate authorities, n(n-1) cross-certification links need to be established. As the number of participating institutions increases, the complexity of link management and communication costs rise sharply. Furthermore, once a trust relationship is established, it is difficult to dynamically adjust. When the trust level of an institution changes, all relevant cross-certificates need to be renegotiated and updated, resulting in low efficiency and compatibility issues. The trust list mechanism is limited by a fixed update cycle and cannot respond promptly to instantaneous changes in the trust status of certificate authorities. Each verification also requires checking both certificate validity and the trust list status simultaneously, increasing computational overhead. In addition, existing solutions often need to transmit or query sensitive compliance data of certificate authorities during cross-border verification, facing challenges in data privacy protection and cross-border regulatory compliance.

[0004] Therefore, how to achieve real-time dynamic adjustment of certificate authority trust weights and adaptive reconstruction of cross-domain trust chains while ensuring data privacy compliance, and improve the efficiency and reliability of cross-border certificate verification, is a technical problem that urgently needs to be solved in the field of cross-border digital certificate mutual recognition. Summary of the Invention

[0005] To address the issues of lagging trust assessment, rigid trust chains, and high privacy compliance risks in existing cross-border digital certificate mutual recognition technologies, this application proposes a cross-border digital certificate mutual recognition method based on dynamic trust anchors. This method enables real-time updates of trust weights of cross-border digital certificate issuing authorities, adaptive construction of hierarchical trust chains, and privacy and security verification under the Groth16 algorithm.

[0006] To achieve the above objectives, this application provides a cross-border digital certificate mutual recognition method based on dynamic trust anchors, which adopts the following technical solution:

[0007] A method for cross-border digital certificate mutual recognition based on dynamic trust anchors, applied to a cross-border trusted interaction system including blockchain and smart contracts, is characterized by comprising:

[0008] Periodically collect multi-dimensional trust assessment data from cross-border digital certificate authorities, preprocess the multi-dimensional trust assessment data, and generate a standardized trust assessment dataset;

[0009] Based on the preprocessed trust assessment dataset, and with preset parameters such as learning rate, number of iterations, batch size, and convergence threshold, the federated XGBoost algorithm combined with a contextual attention mechanism is used to collaboratively train a global dynamic trust assessment model without leaving the country, and periodically outputs real-time trust scores for cross-border digital certificate authorities.

[0010] A hierarchical dynamic trust chain is constructed based on real-time trust scores. Cross-border digital certificate authorities with different new trust levels are included in the fast trust chain and the enhanced trust chain according to the preset trust score range, and the evidence data is written into the blockchain.

[0011] Certificate verification proof is generated based on the Groth16 algorithm, and the certificate validity is verified without obtaining sensitive data through bilinear pairing operations.

[0012] Deploy smart contracts on cross-chain platforms, listen to preset events through preset function interfaces, and automatically execute contract logic under preset trigger conditions to perform trust weight warnings, automatic dispute arbitration, and dynamic updates of mutual recognition rules.

[0013] Periodically monitor changes in the trust scores of cross-border digital certificate authorities and make adaptive adjustments to the dynamic trust anchor;

[0014] Based on a pre-set confidence threshold, anomalies in the trust behavior of cross-border digital certificate authorities are detected and a circuit breaker mechanism is triggered.

[0015] Preferably, multi-dimensional trust assessment data from cross-border digital certificate authorities is collected periodically, and the dimensional trust assessment data is preprocessed, including:

[0016] Periodically collect multi-dimensional trust assessment data, including basic operational data, compliance data, transaction scenario data, and technical capability data;

[0017] The 3σ principle was used to clean the collected multidimensional trust assessment data and identify outliers.

[0018] The cleaned multidimensional trust assessment data were normalized using min-max standardization.

[0019] Differential privacy is used to add Laplace noise to de-identify the normalized multidimensional trust assessment data.

[0020] Preferably, a federated XGBoost algorithm combined with a contextual attention mechanism is used to collaboratively train a global dynamic trust assessment model without leaving the data's domain, including:

[0021] Deploy the model parameters of the federated XGBoost algorithm locally on each participating party and set the preset learning rate;

[0022] The sub-model is trained using multidimensional trust assessment data preprocessed locally by each participant, with a preset batch size set.

[0023] An improved weighted FedAvg aggregation algorithm is adopted, and the aggregation weights are assigned according to the quality of the sub-model training based on the local multidimensional trust evaluation data, and the training is iterated for a preset number of iterations.

[0024] Training is stopped when the change in loss of the multidimensional trust assessment data training sub-model is less than a preset convergence threshold. The parameters of the aggregated dynamic trust assessment model are periodically distributed to each participant to update their local multidimensional trust assessment data training sub-model.

[0025] Preferably, the method employs a federated XGBoost algorithm combined with a contextual attention mechanism, and also includes:

[0026] Dynamically set feature weight coefficients according to different application scenarios in finance, e-commerce, and government affairs;

[0027] Attention scores for each feature dimension are calculated using an attention network;

[0028] Attention weights for each feature dimension are generated by normalization using the softmax function;

[0029] The attention weights and feature data are weighted and fused together and then input into the dynamic trust assessment model to output a real-time trust score.

[0030] Preferably, a hierarchical dynamic trust chain is constructed based on real-time trust scores, including:

[0031] Determine the preset trust score range to which the real-time trust score of a cross-border digital certificate authority belongs;

[0032] Cross-border digital certificate authorities in high-trust intervals are included in the fast trust chain, and a three-level jump structure from terminal digital certificate authorities to regional root digital certificate authorities to dynamic trust anchors is adopted, with preset weight coefficients mapped.

[0033] Cross-border digital certificate authorities in the middle trust interval are included in the fast trust chain, and a five-level jump structure is adopted, which is: terminal digital certificate authority - third-party audit digital certificate authority - international alliance digital certificate authority - regional root digital certificate authority - dynamic trust anchor, and a preset weight coefficient is mapped.

[0034] Cross-border digital certificate authorities with low trust levels are included in the enhanced trust chain and mapped with preset weight coefficients.

[0035] Preferably, the evidence data is written to the blockchain, including:

[0036] Encode the evidence storage fields of each cross-border digital certificate authority as Merkel leaf nodes;

[0037] Merkle root hash is generated by layer-by-layer hash calculation;

[0038] Write the Merkel root hash, version number, and timestamp into the blockchain;

[0039] Synchronize the evidence storage data to all participating nodes.

[0040] Preferably, the certificate verification proof is generated based on the Groth16 algorithm, and the certificate validity verification is completed without obtaining sensitive data through bilinear pairing operations, including:

[0041] Generate a common reference string (CRS);

[0042] The certifier generates a Groth16 proof π = (A, B, C) based on the certificate compliance statement.

[0043] The verifier receives the proof π and the common input;

[0044] The validity of the certificate is verified by bilinear pairing operations; if the verification passes, the certificate is confirmed to be legitimate.

[0045] Preferably, deploying smart contracts on a cross-chain platform includes:

[0046] Deploy trust weight triggering contracts, dispute arbitration contracts, and rule update contracts on the BSN relay chain, and provide preset function interfaces;

[0047] Define preset event types;

[0048] Listen for preset trigger conditions;

[0049] By using Hash Time Locked Contract (HTLC), cross-chain atomic swaps are performed, and the logic of the smart contract is executed according to preset rules while synchronizing the cross-chain state.

[0050] Preferably, the changes in the trust scores of cross-border digital certificate authorities are periodically monitored, and the dynamic trust anchor is adaptively adjusted, including:

[0051] Periodically monitor changes in the trust scores of cross-border digital certificate authorities;

[0052] Determine whether the change in trust score exceeds a preset threshold. If the change in trust score exceeds the preset threshold, trigger the reconstruction of the hierarchical dynamic trust chain.

[0053] Periodically perform health checks on the hierarchical dynamic trust chain;

[0054] Update the global dynamic trust anchor configuration based on the health check results and synchronize all nodes.

[0055] Preferably, based on a pre-set confidence threshold, anomaly detection and circuit breaker mechanisms are implemented for the trust behavior of cross-border digital certificate authorities, including:

[0056] The isolated forest algorithm is used to calculate the abnormal behavior score of cross-border digital certificate authorities, and a pre-set reliability threshold is set.

[0057] Determine whether the abnormal behavior score is less than a preset reliability threshold. If the abnormal behavior score is less than the preset reliability threshold, it is marked as abnormal trust behavior.

[0058] Determine whether the trust score of a cross-border digital certificate authority has been tested less than a preset reliability threshold for a consecutive preset number of times;

[0059] If the trust score of a cross-border digital certificate authority is found to be less than a preset trust threshold for a certain number of consecutive preset tests, the mutual recognition qualification of the cross-border digital certificate authority will be suspended and an early warning notice will be sent, and a circuit breaker recovery observation period will be set.

[0060] Compared with existing technologies, this invention provides a method for cross-border digital certificate mutual recognition based on dynamic trust anchors, which has the following beneficial effects:

[0061] 1. By combining multi-dimensional data collection with a federated learning model, global trust collaborative assessment is achieved without exchanging original sensitive data, resolving the contradiction between cross-border data compliance and real-time trust. By using a hierarchical dynamic trust chain and blockchain notarization, the traditional static root certificate chain is replaced, improving the flexibility and traceability of the trust system. By combining zero-knowledge proofs and smart contracts, the automated execution of mutual recognition rules and risk management are achieved while ensuring verification privacy.

[0062] 2. By using the improved weighted FedAvg aggregation algorithm, aggregation weights are dynamically allocated based on local data quality, avoiding negative interference from low-quality nodes on the global model and improving model convergence speed and evaluation accuracy. At the same time, the local training mode, which keeps data within the country, fundamentally blocks the risk path of cross-border transmission of sensitive data.

[0063] 3. The introduction of the isolated forest algorithm for unsupervised anomaly detection can effectively identify malicious behavior or systemic risks with unknown patterns. Through continuous detection and circuit breaker recovery mechanisms, it can avoid erroneous bans caused by occasional false alarms and isolate problematic nodes in a timely manner when risks are confirmed, thus ensuring the overall robustness of the cross-border mutual recognition system. Attached Figure Description

[0064] Figure 1 This is a flowchart illustrating the steps of a cross-border digital certificate mutual recognition method based on dynamic trust anchors, according to an embodiment of this application.

[0065] Figure 2 This is a flowchart illustrating the steps involved in periodically collecting multi-dimensional trust assessment data from cross-border digital certificate authorities and preprocessing this data.

[0066] Figure 3 This is a flowchart illustrating the steps of constructing a hierarchical dynamic trust chain based on real-time trust scores.

[0067] Figure 4 This is a flowchart illustrating the steps of this method to detect anomalies in the trust behavior of cross-border digital certificate authorities and trigger a circuit breaker mechanism based on a preset confidence threshold. Detailed Implementation

[0068] The following is in conjunction with the appendix Figure 1-4 This application will be described in further detail.

[0069] This application discloses a method for cross-border digital certificate mutual recognition based on dynamic trust anchors. (Refer to...) Figure 1 A cross-border digital certificate mutual recognition method based on dynamic trust anchors, applied to a cross-border trusted interaction system including blockchain and smart contracts, comprising:

[0070] S1. Periodically collect multi-dimensional trust assessment data from cross-border digital certificate authorities, preprocess the multi-dimensional trust assessment data, and generate a standardized trust assessment dataset.

[0071] Specifically, this addresses the issue of multi-source, heterogeneous cross-border data containing sensitive information, providing high-quality, comparable, and compliant input data for subsequent trust assessment models. The data collection process is carried out by data collection agents deployed locally on each participating party's premises. These agents connect to the business systems of cross-border digital certificate authorities via standardized RESTful API interfaces, automatically retrieving various types of trust assessment data according to a preset collection cycle. It should be understood that the collection cycle can be flexibly configured according to actual business needs. For scenarios with high real-time requirements, such as financial transactions, it can be shortened to the minute level; for compliance audit data, it can be relaxed to the daily or weekly level.

[0072] Reference Figure 2 :

[0073] S11. Periodically collect multi-dimensional trust assessment data, including basic operational data, compliance data, transaction scenario data, and technical capability data.

[0074] Among them, basic operational data reflects the daily operational status of cross-border digital certificate authorities, specifically including:

[0075] Certificate issuance volume, total number of certificates issued, calculated daily, weekly, and monthly;

[0076] Certificate revocation rate: the ratio of the number of revoked certificates to the total number of certificates issued;

[0077] Percentage of certificates still in validity: The ratio of the number of certificates currently still valid to the total number of certificates issued in history;

[0078] Average certificate validity period: the arithmetic mean of the remaining validity periods of all valid certificates.

[0079] Compliance data reflects the compliance qualifications and historical records of cross-border digital certificate authorities. Specifically, it includes the annual audit rating of local regulatory agencies, such as grade A, B, C, etc.; the validity period of international certifications, such as the expiration date of ISO 27001 information security management system certification and WebTrust certification; and historical violation records, that is, the number of times and severity of penalties imposed by regulatory agencies within a certain time window in the past.

[0080] Transaction scenario data reflects the performance of cross-border digital certificates in actual business operations, specifically including certificate verification pass rates in cross-border transactions, the number of transaction failures due to certificate issues, and the number of user complaints.

[0081] Technical capability data reflects the technical infrastructure level of cross-border digital certificate authorities, specifically including key update frequency, anti-attack drill results, and system response latency.

[0082] Among them, the certificate verification pass rate is the ratio of the number of successful verifications to the total number of verification requests; the number of transaction failures due to certificate issues includes failures caused by expired certificates, invalid signatures, broken trust chains, etc.; and the number of user complaints is the number of complaints from end users regarding certificate verification failures or delays.

[0083] The key update frequency is the rotation cycle of the root key and intermediate key; the anti-attack exercise result is the pass rate and response time in the periodic security exercise; the system response latency is the average response time for certificate status query requests, usually in milliseconds.

[0084] S12. To ensure data quality, the 3σ principle is used to clean the collected multidimensional trust assessment data and remove outliers. The application of the 3σ principle is not simply to remove outliers, but to distinguish between transient system failures and real trust risks.

[0085] Specifically, for each dimension of the data sequence, its mean μ and standard deviation σ are first calculated, and data points that deviate from the mean by more than 3 times the standard deviation are marked as candidate outliers.

[0086] Subsequently, the cause of the candidate outlier was determined by multi-dimensional cross-validation: if the revocation rate of a cross-border digital certificate authority suddenly increases by more than three times the standard deviation of the mean at a certain moment, accompanied by an abnormal increase in system response delay, it is determined to be false data caused by system failure, and is removed and filled with the valid value of the previous period; if only the revocation rate is abnormal while other indicators such as system response delay and certificate issuance volume are normal, the data point is retained as a real risk signal, because this situation usually reflects that the institution has indeed experienced a large-scale certificate revocation event.

[0087] This multi-dimensional cross-validation mechanism avoids interference from noisy data introduced by system failures in subsequent model training, while retaining the true trust risk signals.

[0088] S13. Furthermore, min-max normalization is used to normalize the cleaned multidimensional trust assessment data. Since the four types of data differ significantly in scale and value range, directly mixing them into the model would lead to high-scale features dominating the model training process. The normalization calculation formula is as follows:

[0089]

[0090] In the formula,

[0091] The original data value;

[0092] and These are the minimum and maximum values ​​of the data in this dimension within the current statistical window, respectively. The value is the normalized value, ranging from [0, 1]. This process maps all dimensions of data to the same numerical range, eliminating the impact of dimensional differences on model training and ensuring comparability of features across different dimensions in subsequent federated learning. It should be understood that when the maximum and minimum values ​​of a certain dimension are equal (i.e., all participants have completely identical data for that dimension), the normalization result is uniformly set to 0.5 to avoid computational anomalies such as a denominator of zero.

[0093] S14. Furthermore, to meet cross-border compliance requirements such as GDPR and the Data Security Act, differential privacy is used to add Laplace noise to the normalized multidimensional trust assessment data for anonymization. The purpose of the anonymization is to add controllable random perturbations to the precise values ​​before the data leaves the local node, so that even if an attacker intercepts the data in transmission, they cannot deduce the precise operational indicators of a specific cross-border digital certificate authority. The probability density function of Laplace noise is:

[0094]

[0095] In the formula,

[0096] f Let be the probability density function value of the Laplace noise. The output value of the data after adding noise. The location parameter is set to 0, meaning the noise is symmetrically distributed around zero. This is a scaling parameter, and its value is determined based on query sensitivity. With preset privacy budget according to = The relationship is established. This refers to the global sensitivity of the query function, which is the maximum impact of a change in a single data point on the query results. A preset differential privacy budget is used to control the strength of privacy protection. A smaller value indicates greater noise and stronger privacy protection, but lower data availability. In this embodiment, The typical value range is 0.1 to 1.0, and it can be configured differently according to the regulatory requirements of each participating party's region. For example, cross-border digital certificate authorities in the EU can... Setting it to 0.5 meets the stringent requirements of the GDPR, while other regions can be set to 1.0 to retain higher data availability. This embodiment adds noise after normalization and before model training, which both masks the precise sensitive values ​​of individual institutions and preserves the statistical distribution characteristics of the overall data, allowing the subsequent federated learning model to still converge to a valid solution.

[0097] S2. Based on the preprocessed trust assessment dataset, and with preset parameters including learning rate, number of iterations, batch size, and convergence threshold, a federated XGBoost algorithm combined with a contextualized attention mechanism is used to collaboratively train a global dynamic trust assessment model, periodically outputting real-time trust scores for cross-border digital certificate authorities, all while keeping the original data of each participant locally and without cross-border transmission. The core objective of this step is to collaboratively construct a globally shared trust assessment model through a federated learning framework, thereby balancing data privacy compliance and model evaluation accuracy, while ensuring that the original data of each participant remains locally and is not transferred across borders.

[0098] S21. Deploy the model parameters of the federated XGBoost algorithm locally on each participating party and set a preset learning rate. XGBoost is a high-efficiency machine learning algorithm based on decision tree ensemble. It gradually builds multiple decision trees and fits the prediction residuals of each tree, finally obtaining the final output by weighted summation of the prediction results of all trees. In this embodiment, the maximum depth of each decision tree is set to 4 to 8 layers, the typical value range of the learning rate is 0.01 to 0.3, and the total number of trees is set to 100 to 500.

[0099] S22. After each participant initializes the model parameters locally, they train the sub-model using the locally preprocessed multidimensional trust evaluation data, setting a preset batch size. During training, the construction of each new tree is based on the ensemble prediction residuals of all trees in the previous round. An optimal split point is selected using a greedy algorithm to maximize the decrease in the loss function of the child nodes after the split.

[0100] S23. Based on this, an improved weighted FedAvg aggregation algorithm is adopted. Aggregation weights are allocated according to the quality of the sub-model training based on the local multidimensional trust assessment data, and the training is iterated for a preset number of iterations. Unlike the traditional FedAvg algorithm which allocates weights proportionally based on the amount of data of each participant, this embodiment dynamically calculates the aggregation weights based on the Shannon entropy or validation set accuracy of the local data of each participant.

[0101] Specifically, Shannon entropy measures the information richness of the local dataset; a higher entropy value indicates a more uniform data distribution and greater information content, corresponding to a higher aggregation weight. Validation set accuracy directly reflects the prediction performance of the local sub-model on the independent validation set; nodes with higher accuracy receive greater aggregation weights. After receiving the encrypted model gradients uploaded by each participant, the federated coordination node calculates the aggregation weight for each participant according to the aforementioned quality metrics, and then performs weighted average aggregation to generate global model parameters. After aggregation, the global model parameters are encrypted and distributed to each participant via the SecureAggregation protocol. Each participant updates its local sub-model using the global parameters and enters the next round of iterative training.

[0102] S24. When the change in loss of the multidimensional trust assessment data training sub-model is less than a preset convergence threshold, training is stopped. The parameters of the aggregated dynamic trust assessment model are periodically distributed to each participant to update their local multidimensional trust assessment data training sub-model. Through this quality-aware weighted aggregation strategy, nodes with higher data quality contribute more to the global model, thereby accelerating global model convergence and effectively suppressing the negative interference of low-quality data on model accuracy.

[0103] In addition, to adapt to differentiated business needs, the use of the federated XGBoost algorithm combined with a scenario-based attention mechanism also includes:

[0104] S231. Dynamically set feature weight coefficients based on different application scenarios in finance, e-commerce, and government affairs. The emphasis on each dimension of trust assessment varies significantly across different application scenarios. For example, in financial scenarios, compliance data is far more important than transaction data because financial institutions have extremely strict requirements for compliance qualifications; in e-commerce scenarios, the weight of transaction verification pass rate and system response latency is significantly increased because e-commerce transactions are highly sensitive to verification speed and success rate; in government affairs scenarios, the weight of technical capability data is prioritized because government systems have higher requirements for security and stability.

[0105] S232. Calculate the attention score for each feature dimension using an attention network. The attention network consists of a fully connected layer and an activation function. The input is the concatenation of the current scene identifier vector and the feature vector, and the output is the original attention score for each feature dimension.

[0106] S233. Subsequently, attention weights for each feature dimension are generated by normalization using the softmax function. The softmax normalization formula is as follows:

[0107]

[0108] In the formula:

[0109] For the first Attention weights normalized to each feature dimension The first output of the attention network Attention scores for each feature dimension For the first Attention scores for each feature dimension To sum the index variables and iterate from 1 to All feature dimensions, The total number of feature dimensions. For the natural constant An exponential function with base 0.

[0110] The normalized attention weights satisfy the constraint that the sum of the weights of all dimensions is 1.

[0111] S234. The attention weights and feature data are weighted and fused together, then input into the dynamic trust assessment model to output a real-time trust score. The trust score ranges from 0 to 100, with a higher score indicating a higher overall trust level for the cross-border digital certificate authority. This mechanism allows the same underlying model to output customized trust scores for different scenarios, solving the problem of poor adaptability of general-purpose models in specific domains. It should be understood that the scenario types in the scenario-based attention mechanism are not limited to finance, e-commerce, and government affairs. In actual deployment, more scenario types can be expanded according to business needs. The feature weight coefficients corresponding to each scenario can obtain initial values ​​through offline training with historical data and be continuously fine-tuned during online operation.

[0112] S3. Construct a hierarchical dynamic trust chain based on the real-time trust score. According to the preset trust score range, cross-border digital certificate authorities with different new trust levels are included in the fast trust chain and the enhanced trust chain, respectively, and the evidence data is written to the blockchain. The core objective of this step is to automatically match a trust chain structure with the corresponding security level for each cross-border digital certificate authority based on the real-time trust score output in step S2, optimizing verification efficiency while ensuring a minimum level of security.

[0113] Reference Figure 3 S31. Determine the preset trust score range to which the real-time trust score of the cross-border digital certificate authority belongs. This embodiment presets three trust score ranges: high trust range, medium trust range (70 ≤ score < 90), and low trust range (score < 70). It should be understood that the above range thresholds are only illustrative examples and can be dynamically adjusted according to the overall trust level of the participating cross-border digital certificate authorities in actual deployment. For example, when the overall trust level of participating institutions is high, the lower limit of the high trust range can be lowered from 90 points to 85 points to expand the coverage of the fast trust chain.

[0114] The high trust interval is defined as a score ≥ 90; the medium trust interval is defined as a score ≤ 70 < 90; and the low trust interval is defined as a score < 70.

[0115] S32. Incorporate cross-border digital certificate authorities (DCAs) in high-trust intervals into a fast trust chain, employing a three-tiered hop structure from the terminal DCA to the regional root DCA to the dynamic trust anchor, mapped with preset weight coefficients. This three-tiered structure means that when verifying a certificate issued by this authority, the verifier only needs to complete the legitimacy confirmation through three trust transfers: first, verifying the signature validity of the terminal DCA; then, verifying the endorsement of the terminal DCA by the regional root DCA; and finally, verifying the authorization of the regional root by the global dynamic trust anchor. This streamlined link significantly reduces the number of network round trips and computational overhead during the verification process, making it suitable for institutions with high trust levels and good historical records.

[0116] S33. Incorporate cross-border digital certificate authorities (DCAs) within the intermediate trust interval into the fast trust chain, adopting a five-level jump structure: Endpoint DCA - Third-Party Audited DCA - International Consortium DCA - Regional Root DCA - Dynamic Trust Anchor, mapped with a preset weight coefficient (e.g., 0.8). This five-level jump structure adds two intermediate verification nodes to the three-level jump structure: the Third-Party Audited DCA and the International Consortium DCA. The Third-Party Audited DCA is responsible for independently auditing the compliance qualifications of the institution and issuing audit certificates, while the International Consortium DCA acts as a cross-regional trust bridge, providing additional trust endorsement. The purpose of adding these two intermediate nodes is to enhance the reliability of trust transmission by introducing independent third-party audit verification when the trust level has not yet reached its optimal level. The weight coefficient of 0.8 indicates that certificates issued by this institution may require additional cross-validation steps during verification.

[0117] S34. Incorporate cross-border digital certificate authorities (DCs) in the low-trust range into the enhanced trust chain, mapping them to preset weight coefficients and mandating additional manual review or multi-factor authentication. The enhanced trust chain further strengthens security constraints on top of the five-level hop structure: on the one hand, every cross-border certificate issued by this institution must trigger a manual review process during verification, with operations personnel confirming the institution's latest compliance status; on the other hand, the verifier must ensure the certificate's validity through multi-factor authentication. A weight coefficient of 0.3 indicates a low trust level for the institution, and the acceptance of its issued certificates in cross-border transactions is strictly limited. This tiered structure, while ensuring a minimum level of security, significantly reduces the number of verification hops for high-trust institutions, improving overall mutual recognition efficiency.

[0118] S341. To ensure the traceability of changes to the trust chain, the process of writing the evidence data into the blockchain includes: encoding the evidence fields of each cross-border digital certificate authority into Merkel leaf nodes.

[0119] The evidence storage fields include the institution's unique identifier, current trust score, type of trust chain to which it belongs, weight coefficient, and score update time.

[0120] S342. Each institution's evidence storage field is hashed using SHA-256 to generate the corresponding leaf node hash value. Merkle root hash is generated through layer-by-layer hash calculation.

[0121] Specifically, all leaf node hash values ​​are paired sequentially, and each pair is concatenated and then subjected to another SHA-256 hash operation to generate the hash value of the node in the next layer. This process is repeated layer by layer upwards until a unique Merkle root hash is generated. The Merkle root hash, version number, and timestamp are then written into the blockchain.

[0122] S343, where the version number is used to identify the sequence number of this trust chain update, and the timestamp records the precise time of this update.

[0123] The blockchain adopts a consortium blockchain architecture, in which consensus nodes are jointly maintained by regional root digital certificate authorities that participate in mutual recognition, ensuring the immutability of the stored data.

[0124] S344. Synchronize the evidence storage data to all participating nodes. After receiving the evidence storage data, each participating node independently verifies the correctness of the Merkle root hash. After confirming consistency, it writes the update to its local ledger.

[0125] Leveraging the efficient verification properties of Merkle trees, even if a large number of institutions' trust states are stored on-chain, the verifier can still verify... This technology allows for rapid verification of the authenticity of an institution's status with minimal time complexity, providing immutable, judicial-grade evidence for subsequent dispute arbitration. For example, when a cross-border transaction dispute requires tracing the trust status of an institution at a specific point in time, the arbitrator can quickly locate and verify the institution's trust score and trust chain affiliation at that point in time using Merkle Proof, without having to traverse the entire Merkle tree.

[0126] S4. Generate certificate verification proof based on the Groth16 algorithm, and complete the certificate validity verification without obtaining sensitive data through bilinear pairing operation.

[0127] The core objective of this step is to address the privacy breach issue in cross-border certificate verification: In traditional solutions, the verifier needs to obtain sensitive data such as audit reports from cross-border digital certificate authorities, user identity information, and private key association information to complete the certificate validity verification. This poses serious data privacy protection and regulatory compliance risks in cross-border scenarios.

[0128] The Groth16 algorithm, as an efficient zero-knowledge proof protocol, allows the prover to demonstrate to the verifier the authenticity of the proposition that the certificate meets all compliance requirements without exposing any sensitive raw data.

[0129] Specifically, this step begins with a trusted third party generating a Public Reference String (CRS). The trusted third party can be a neutral organization participating in the mutual recognition alliance or a key management service jointly authorized by multiple parties.

[0130] The generation process of CRS is based on elliptic curve cryptography, specifically using paired friendly curves such as BN254 or BLS12-381. During the generation process, a trusted third party first selects a random secret parameter, and then calculates a series of group elements based on this secret parameter as the components of the CRS, including a proof key and a verification key.

[0131] Once the CRS is generated, the trusted third party must completely destroy the secret parameters to prevent anyone from using them to forge legitimate proofs. The CRS is publicly available as a public parameter to all participants, accessible to both the proving and verifying parties. Subsequently, the proving party (i.e., the cross-border digital certificate authority) generates a Groth16 proof based on the certificate compliance statement. .

[0132] A certificate compliance statement is a set of structured assertions, including but not limited to: "This certificate has passed a compliance audit," "This certificate's revocation status is normal," "This certificate's signature is valid," and "This certificate has not expired." The proving party transforms these assertions into an arithmetic circuit and calculates and generates a proof based on the proof key portion in the CRS and its own private witness data.

[0133] The proof consists of three group elements, with a very small total size, based on the BN254 curve, typically only 192 bytes, negligible compared to the size of the original sensitive data. The proof is then received by the verifier. The validity of the proof is verified using bilinear pairing operations on the common input. The common input includes the publicly disclosed portion of the certificate compliance statement. Verification requires only three bilinear pairing operations and a small number of group operations, with constant computational overhead independent of the complexity of the original data. The verification equation is as follows:

[0134]

[0135] In the formula:

[0136] This represents the bilinear pairing operation defined on the elliptic curve group. , , Proofs for Groth16 The three group elements in the data are calculated and generated by the proving party based on the certificate compliance statement. For the verification key portion of the Public Reference String (CRS), These are common input parameters in the certificate compliance statement. This is a public reference string generated by a trusted third party during the Setup phase.

[0137] When the above equation holds true, the verification passes, confirming the certificate's legitimacy. This means the verifier is confident that the certificate meets all compliance requirements without needing to obtain any sensitive raw data from the proving party. The Groth16 algorithm's advantage lies in its extremely small proof size and constant verification overhead. A single verification can be compressed to milliseconds, fully meeting the performance requirements of cross-border real-time transactions while completely avoiding the cross-border transmission of sensitive raw data such as audit reports and user identities.

[0138] It should be understood that the Groth16 algorithm is only one feasible implementation of zero-knowledge proofs. In other embodiments, other zero-knowledge proof protocols such as PLONK, Bulletproofs, or STARK can also be used, as long as the functional requirement of "the prover can prove the truth of the proposition to the verifier without exposing sensitive data" is met. Different protocols have their own advantages and disadvantages in terms of proof size, proof generation time, and verification time, and can be selected according to the performance requirements of specific application scenarios.

[0139] S5. Deploy smart contracts on the cross-chain platform, listen for preset events through preset function interfaces, and automatically execute contract logic under preset trigger conditions to perform trust weight warnings, automatic dispute arbitration, and dynamic updates of mutual recognition rules. The core objective of this step is to encode trust management rules into on-chain executable smart contracts, realize the automated execution of cross-border mutual recognition rules and cross-domain collaboration, and eliminate the problems of rule execution lag and cross-domain coordination difficulties caused by reliance on manual intervention in traditional solutions.

[0140] Specifically, the BSN relay chain deploys trust weight trigger contracts, dispute arbitration contracts, and rule update contracts, providing pre-defined function interfaces. As a hub for cross-chain communication, the BSN relay chain connects consortium chains or public chains in the regions of participating parties, enabling the execution results of smart contracts to be synchronized across chains.

[0141] The three types of contracts each perform different automated management functions: the Trust Weight Trigger Contract monitors changes in the trust scores of cross-border digital certificate authorities and automatically issues warnings and adjusts eligibility when preset conditions are met; the Dispute Arbitration Contract automatically retrieves on-chain evidence and adjudicates according to preset rules when cross-border certificate verification disputes occur; and the Rule Update Contract automatically updates the feature weight configuration of the trust assessment model when regional mutual recognition agreements are revised. Preset event types are defined, and preset trigger conditions are monitored.

[0142] Preset event types include, but are not limited to: trust score update events, certificate verification dispute events, and protocol revision events. The smart contract uses an event listening mechanism to capture these events in real time and automatically executes the corresponding contract logic when preset trigger conditions are met.

[0143] For example, if a cross-border digital certificate authority's trust score falls below 60 for three consecutive hours, the trust weight trigger contract automatically sends an alert to related regional nodes and suspends its Fast Trust Chain qualification. Simultaneously, it migrates the authority to the Enhanced Trust Chain and records this status change in the blockchain ledger for subsequent auditing and traceability. In the event of a verification dispute, the dispute arbitration contract automatically retrieves the on-chain stored Trust Chain version and certificate verification record, and makes a ruling within 15 minutes according to the "highest trust weight priority" rule. Specifically, when verification results from two different regions conflict, the contract automatically accepts the verification result of the cross-border digital certificate authority with the higher trust weight and writes the ruling to the blockchain. When regional mutual recognition agreements are revised, the rule update contract automatically updates the model feature weight configuration. For example, if new mutual recognition clauses require an increased weighting of compliance data, the contract will automatically adjust the feature weight coefficients of the attention mechanism in step S2, without requiring manual model redeployment.

[0144] To ensure the atomicity of cross-chain operations, Hash Time Locking Contracts (HTLCs) are used for atomic cross-chain swaps. The smart contract logic is executed according to preset rules, and the cross-chain state is synchronized. The HTLC mechanism works as follows: The initiator first generates a random secret value and calculates its hash value as the locking condition. Then, contracts with this hash value as the locking condition are created on each relevant chain, and a uniform timeout is set. Before the timeout expires, each participant must submit their secret value sequentially to unlock the contract and complete the state change. If all participants submit their secret values ​​within the specified time, the cross-chain state change takes effect simultaneously on all chains; if any participant fails to complete the operation within the timeout period, the contracts on all chains automatically roll back to their state before the change.

[0145] The hash time-locked contract mechanism ensures that even in the event of network partitioning or partial node failure, changes to the trust state either take effect successfully on all relevant chains or are rolled back entirely, thus preventing trust crises caused by inconsistencies in the state.

[0146] S6. Periodically monitor changes in the trust scores of cross-border digital certificate authorities and make adaptive adjustments to the dynamic trust anchor.

[0147] The core objective of this step is to establish the adaptive adjustment capability of the trust system, enabling the entire mutual recognition system to respond promptly to drastic changes in the trust status of cross-border digital certificate authorities, and avoid verification failures or security risks caused by delayed trust assessments.

[0148] S61. Periodically monitor the changes in trust scores of cross-border digital certificate issuing authorities. The monitoring period can be configured to be once per hour or once every two hours. During each monitoring, the system reads the current score of each institution from the latest trust score output in step S2, and calculates the difference between it and the score of the previous period to obtain the change in trust score.

[0149] S62. Determine if the change in trust score exceeds a preset threshold. If the change exceeds the preset threshold, trigger the reconstruction of the hierarchical dynamic trust chain. The typical value of the preset threshold is 20 points, meaning that when an institution's trust score drops by more than 20 points within a monitoring period, the system determines that the institution has experienced a significant deterioration in its trust status. For example, if an institution's score drops from 92 points to 68 points within one hour, the system immediately migrates it from the fast trust chain to the enhanced trust chain, recalculates the verification strategies for all verification paths involving the institution, and sends status change notifications to associated nodes that rely on the institution's certificates for cross-border transactions.

[0150] S63. Periodically perform health checks on the hierarchical dynamic trust chain to detect issues such as broken chains, expired certificates, or invalid signatures. The health check execution cycle can be configured to be once daily or once weekly. The check content includes: verifying whether the certificate of each node in the trust chain is still valid, whether the signature relationship between nodes is complete and valid, and whether there is a break in the trust chain due to a node going offline.

[0151] S64. For issues discovered during the inspection, the system automatically generates repair suggestions and triggers corresponding repair processes. For example, when it detects that the certificate of an intermediate node is about to expire, the system sends a certificate update reminder to that node in advance; when it detects a broken trust chain, the system automatically attempts to rebuild the trust chain through an alternative path. The system updates the global dynamic trust anchor configuration based on the health check results and synchronizes it with all nodes.

[0152] The configuration update is completed through the blockchain notarization mechanism described in step S3, ensuring that all participating nodes obtain the latest and consistent configuration. This adaptive adjustment mechanism enables the trust system to cope with sudden risks and avoids large-scale service interruptions caused by the failure of trust anchors in traditional static public key infrastructure systems.

[0153] S7. Based on the preset confidence threshold, perform anomaly detection on the trust behavior of cross-border digital certificate authorities and trigger the circuit breaker mechanism.

[0154] This step serves as the final security line in the entire mutual recognition method. It aims to identify potential malicious behavior or systemic risks among cross-border digital certificate authorities through unsupervised anomaly detection algorithms, and to promptly isolate problematic nodes after confirming risks to prevent the risks from spreading to the entire mutual recognition system.

[0155] Reference Figure 4 S71. The isolated forest algorithm is used to calculate the abnormal behavior score of cross-border digital certificate authorities, and a pre-set reliability threshold is set.

[0156] The core idea of ​​the Isolation Forest algorithm is that outlier data points, because they deviate from most normal data points, are more easily isolated by a small number of segmentation operations when randomly partitioning the feature space, while normal data points require more segmentation operations to be isolated.

[0157] The algorithm first constructs multiple isolated trees. Each tree recursively segments the data by randomly selecting feature dimensions and split values ​​until each data point is isolated or the maximum depth of the tree is reached. Then, for each cross-border digital certificate authority, its average path length in all isolated trees is calculated. The shorter the path length, the easier it is for the data point to be isolated, and the more likely it is to be an outlier.

[0158] The behavioral anomaly score is calculated based on the normalized average path length and ranges from 0 to 1. A lower score indicates a higher degree of anomaly. The typical preset reliability threshold is 0.3 to 0.5, which can be adjusted according to the risk tolerance of the actual business scenario.

[0159] S72. Determine whether the abnormal behavior score is less than the preset reliability threshold. If the abnormal behavior score is less than the preset reliability threshold, mark it as abnormal trust behavior.

[0160] For example, if a cross-border digital certificate authority's behavior anomaly score is 0.2, the system will mark it as abnormal trust behavior and initiate a continuous detection and confirmation process. To prevent false alarms caused by network fluctuations or temporary malfunctions,

[0161] S73. Determine whether the trust score of the cross-border digital certificate authority has been detected less than the preset trust threshold for a consecutive preset number of times.

[0162] The time interval for continuous detection is consistent with the monitoring cycle in step S6, meaning that the circuit breaker will only be triggered if the detection results show abnormalities for three consecutive hours. This continuous detection mechanism effectively filters out occasional abnormal scores caused by non-malicious factors such as instantaneous network fluctuations and data acquisition delays, thus avoiding unnecessary false bans.

[0163] S74. The mutual recognition qualification of a cross-border digital certificate authority will be suspended and an early warning notification will be sent only when the trust score of the cross-border digital certificate authority is detected less than a preset trust threshold for a certain number of consecutive preset times, and a circuit breaker recovery observation period will be set. After the circuit breaker is executed, the system will automatically complete the following operations: remove the authority from all trust chains, broadcast the circuit breaker notification to all participating nodes, write the latest status of the authority to the blockchain for evidence storage, and send an early warning report to the regulatory agency.

[0164] The typical duration of the observation period for circuit breaker recovery is 7 to 30 days. During this period, the institution's mutual recognition qualification is suspended, and its issued certificates are not accepted by other participating parties. The circuit breaker status can only be lifted if the institution's various indicators return to normal and it passes a manual or automated review during the observation period. The review includes: confirming the root cause of the abnormal behavior, assessing the adequacy of the corrective measures taken by the institution, and verifying whether its trust score has recovered to above the safety threshold. After the review is passed, the system automatically reintegrates the institution into the corresponding trust chain and synchronizes it with all participating nodes. This mechanism ensures the overall robustness of the system while minimizing the impact on normal business operations.

[0165] Through the coordinated execution of steps S1 to S7, this embodiment constructs a data-driven, privacy-preserving, and adaptively adjustable cross-border digital certificate mutual recognition system. There is a close logical causal relationship between each step: the high-quality standardized data output by S1 forms the basis for the accurate training of the S2 model; the real-time score output by S2 directly drives the trust chain grading in S3 and the adaptive adjustment in S6; the on-chain evidence storage in S3 provides a credible basis for the smart contract arbitration in S5; the zero-knowledge proof in S4 ensures privacy compliance throughout the verification process; and the abnormal circuit breaker in S7 serves as the last line of defense, ensuring the security of the entire system under extreme conditions. This end-to-end closed-loop design enables this invention not only to achieve breakthroughs in a single technology but also to realize an overall leap in the efficiency of cross-border trust interaction at the system level.

[0166] The federated XGBoost algorithm used in the foregoing embodiments can be replaced with other privacy-preserving federated learning frameworks. The Groth16 zero-knowledge proof scheme can also be equivalently replaced with algorithms such as PLONK or Bulletproofs. The number of jump levels and the scoring interval threshold of the hierarchical dynamic trust chain can be adaptively adjusted according to actual cross-border business needs. The BSN relay chain can also be replaced with other consortium chains or public chain platforms with cross-chain capabilities. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this invention, based on the core concepts of dynamic trust anchor construction, privacy compliance verification, and smart contract collaborative execution, should be included within the protection scope of this invention. Therefore, the protection scope of this invention should be determined by the scope of the claims.

Claims

1. A method for cross-border digital certificate mutual recognition based on dynamic trust anchors, applied to a cross-border trusted interaction system including blockchain and smart contracts, characterized in that, include: Periodically collect multi-dimensional trust assessment data from cross-border digital certificate authorities, preprocess the multi-dimensional trust assessment data, and generate a standardized trust assessment dataset; Based on the preprocessed trust assessment dataset, and with preset parameters such as learning rate, number of iterations, batch size, and convergence threshold, the federated XGBoost algorithm combined with a contextual attention mechanism is used to collaboratively train a global dynamic trust assessment model without leaving the country, and periodically outputs real-time trust scores for cross-border digital certificate authorities. A hierarchical dynamic trust chain is constructed based on real-time trust scores. Cross-border digital certificate authorities with different new trust levels are included in the fast trust chain and the enhanced trust chain according to the preset trust score range, and the evidence data is written into the blockchain. Certificate verification proof is generated based on the Groth16 algorithm, and the certificate validity is verified without obtaining sensitive data through bilinear pairing operations. Deploy smart contracts on cross-chain platforms, listen to preset events through preset function interfaces, and automatically execute contract logic under preset trigger conditions to perform trust weight warnings, automatic dispute arbitration, and dynamic updates of mutual recognition rules. Periodically monitor changes in the trust scores of cross-border digital certificate authorities and make adaptive adjustments to the dynamic trust anchor; Based on a pre-set confidence threshold, anomalies in the trust behavior of cross-border digital certificate authorities are detected and a circuit breaker mechanism is triggered.

2. The method for cross-border digital certificate mutual recognition based on dynamic trust anchors according to claim 1, characterized in that, Periodically collect multi-dimensional trust assessment data from cross-border digital certificate authorities, and preprocess the dimensional trust assessment data, including: Periodically collect multi-dimensional trust assessment data, including basic operational data, compliance data, transaction scenario data, and technical capability data; The 3σ principle was used to clean the collected multidimensional trust assessment data and identify outliers. The cleaned multidimensional trust assessment data were normalized using min-max standardization. Differential privacy is used to add Laplace noise to de-identify the normalized multidimensional trust assessment data.

3. The method for cross-border digital certificate mutual recognition based on dynamic trust anchors according to claim 1, characterized in that, A global dynamic trust assessment model is collaboratively trained using a federated XGBoost algorithm combined with a contextual attention mechanism, without leaving the data outside the network. This includes: Deploy the model parameters of the federated XGBoost algorithm locally on each participating party and set the preset learning rate; The sub-model is trained using multidimensional trust assessment data preprocessed locally by each participant, with a preset batch size set. An improved weighted FedAvg aggregation algorithm is adopted, and the aggregation weights are assigned according to the quality of the sub-model training based on the local multidimensional trust evaluation data, and the training is iterated for a preset number of iterations. Training is stopped when the change in loss of the multidimensional trust assessment data training sub-model is less than a preset convergence threshold. The parameters of the aggregated dynamic trust assessment model are periodically distributed to each participant to update their local multidimensional trust assessment data training sub-model.

4. The method for cross-border digital certificate mutual recognition based on dynamic trust anchors according to claim 1, characterized in that, It employs a federated XGBoost algorithm combined with a contextual attention mechanism, and also includes: Dynamically set feature weight coefficients according to different application scenarios in finance, e-commerce, and government affairs; Attention scores for each feature dimension are calculated using an attention network; Attention weights for each feature dimension are generated by normalization using the softmax function; The attention weights and feature data are weighted and fused together and then input into the dynamic trust assessment model to output a real-time trust score.

5. A method for cross-border digital certificate mutual recognition based on dynamic trust anchors according to claim 1, characterized in that, A hierarchical dynamic trust chain is constructed based on real-time trust scores, including: Determine the preset trust score range to which the real-time trust score of a cross-border digital certificate authority belongs; Cross-border digital certificate authorities in high-trust intervals are included in the fast trust chain, and a three-level jump structure from terminal digital certificate authorities to regional root digital certificate authorities to dynamic trust anchors is adopted, with preset weight coefficients mapped. Cross-border digital certificate authorities in the middle trust interval are included in the fast trust chain, and a five-level jump structure is adopted, which is: terminal digital certificate authority - third-party audit digital certificate authority - international alliance digital certificate authority - regional root digital certificate authority - dynamic trust anchor, and a preset weight coefficient is mapped. Cross-border digital certificate authorities with low trust levels are included in the enhanced trust chain and mapped with preset weight coefficients.

6. A method for cross-border digital certificate mutual recognition based on dynamic trust anchors according to claim 1, characterized in that, Write the evidence data into the blockchain, including: Encode the evidence storage fields of each cross-border digital certificate authority as Merkel leaf nodes; Merkle root hash is generated by layer-by-layer hash calculation; Write the Merkel root hash, version number, and timestamp into the blockchain; Synchronize the evidence storage data to all participating nodes.

7. A method for cross-border digital certificate mutual recognition based on dynamic trust anchors according to claim 1, characterized in that, Certificate verification proofs are generated based on the Groth16 algorithm. Certificate validity verification is performed through bilinear pairing operations without obtaining sensitive data, including: Generate a common reference string (CRS); The certifier generates a Groth16 proof π = (A, B, C) based on the certificate compliance statement. The verifier receives the proof π and the common input; The validity of the certificate is verified by bilinear pairing operations; if the verification passes, the certificate is confirmed to be legitimate.

8. A method for cross-border digital certificate mutual recognition based on dynamic trust anchors according to claim 1, characterized in that, Deploying smart contracts on cross-chain platforms includes: Deploy trust weight triggering contracts, dispute arbitration contracts, and rule update contracts on the BSN relay chain, and provide preset function interfaces; Define preset event types; Listen for preset trigger conditions; By using Hash Time Locked Contract (HTLC), cross-chain atomic swaps are performed, and the logic of the smart contract is executed according to preset rules while synchronizing the cross-chain state.

9. A method for cross-border digital certificate mutual recognition based on dynamic trust anchors according to claim 1, characterized in that, Periodically monitor changes in the trust scores of cross-border digital certificate authorities and adaptively adjust the dynamic trust anchor, including: Periodically monitor changes in the trust scores of cross-border digital certificate authorities; Determine whether the change in trust score exceeds a preset threshold. If the change in trust score exceeds the preset threshold, trigger the reconstruction of the hierarchical dynamic trust chain. Periodically perform health checks on the hierarchical dynamic trust chain; Update the global dynamic trust anchor configuration based on the health check results and synchronize all nodes.

10. A method for cross-border digital certificate mutual recognition based on dynamic trust anchors according to claim 1, characterized in that, Based on a pre-set confidence threshold, anomalies in the trust behavior of cross-border digital certificate authorities are detected and a circuit breaker mechanism is triggered, including: The isolated forest algorithm is used to calculate the abnormal behavior score of cross-border digital certificate authorities, and a pre-set reliability threshold is set. Determine whether the abnormal behavior score is less than a preset reliability threshold. If the abnormal behavior score is less than the preset reliability threshold, it is marked as abnormal trust behavior. Determine whether the trust score of a cross-border digital certificate authority has been tested less than a preset reliability threshold for a consecutive preset number of times; If the trust score of a cross-border digital certificate authority is found to be less than a preset trust threshold for a certain number of consecutive preset tests, the mutual recognition qualification of the cross-border digital certificate authority will be suspended and an early warning notice will be sent, and a circuit breaker recovery observation period will be set.

Images