Cross-data island consensus verification method and system based on credibility score

Abstract

The invention belongs to the technical field of cross-data island consensus, and particularly relates to a cross-data island consensus verification method and system based on credibility scoring. According to the invention, by dynamically evaluating the trust degree of each data island and combining network state fluctuation, real-time verification and abnormal early warning of a consensus result are realized, the stability and safety of cross-data island cooperation are effectively improved, the data consistency is ensured, the data risk caused by network fluctuation is reduced, the self-adaptive capability of the system is enhanced, and the method is suitable for large-scale popularization and application. In the consensus process, historical interaction events and a current cooperation state of a data island are considered, efficient alignment of cross-island features is achieved through a gradient exchange mechanism, a standardized consensus data view is generated, in addition, self-adaptive adjustment of a consensus verification rule is ensured by pre-deploying a trusted execution environment, and the consensus verification efficiency is improved. And the consensus accuracy and efficiency are further improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of cross-data island consensus, and specifically relates to a cross-data island consensus verification method and system based on credibility scoring. Background Art

[0002] With the rapid development of information technology, the amount of data in all walks of life has shown explosive growth. However, due to the existence of data silos, these data are often scattered in different systems, platforms or organizations, making it difficult to achieve effective integration and utilization. The information barriers between data silos not only limit the circulation and sharing of data, but also lead to the waste of data resources and the failure to fully tap the potential value. The credit field is one of the areas seriously affected by data silos. With the popularization of online business, the amount of data in the credit and financial field has shown explosive growth, but the data silo problem between banks, consumer finance companies, credit reporting agencies and other entities is particularly prominent. The user credit data of different institutions (such as repayment records, borrowing frequency, asset information, etc.) are stored in a dispersed manner, forming an information barrier and lacking cross-institutional data collaboration. It is impossible to timely discover potential risks such as multiple borrowings and credit deterioration of users. Based on this, the need for effective integration and consensus recognition of data between data silos is becoming increasingly urgent.

[0003] In the existing technology, credit assessment methods based on statistical learning have problems of high computational complexity and low consensus efficiency when processing heterogeneous data across institutions, and are unable to dynamically adapt to real-time fluctuations in the financial network environment. For example, when the network delay between different financial institutions suddenly increases, traditional methods find it difficult to verify data consistency in real time, which may lead to misjudgment of user qualifications (such as misjudging "normal users" as "high-risk users"). Based on this, the cross-data island consensus verification method proposed in this solution can effectively solve the problems of cross-institutional data integration and real-time risk control in the credit and financial field through dynamic trust evaluation and network fluctuation verification. Summary of the Invention

[0004] The purpose of this invention is to provide a consensus verification method and system across data silos based on credibility scoring, which can optimize the consensus process and improve recognition accuracy and efficiency by dynamically evaluating the trustworthiness of each data silo.

[0005] The technical solutions adopted by the present invention are as follows:

[0006] A cross-data-island consensus verification method based on credibility scoring, including:

[0007] Quantify the historical interaction events between the collected data islands to obtain the collaboration credibility index;

[0008] Assign corresponding execution weights to each data island based on the collaboration credibility index;

[0009] Through the pre-deployed trusted execution environment, the consensus verification rules are adaptively adjusted according to the distribution of execution weights to obtain the consensus result;

[0010] Verify the validity of the consensus results based on network status fluctuations;

[0011] When the verification result of the consensus result is invalid, the abnormal warning mechanism is triggered, and the abnormal data source is traced based on the abnormal warning mechanism.

[0012] In a preferred solution, before quantifying the collected historical interaction events between the data islands to obtain the collaboration credibility index, the method further includes:

[0013] By pre-building a collaborative processing channel between the various data islands, deploying heterogeneous data interfaces for converting data formats of different data islands in the collaborative processing channel, and adopting a unified communication protocol to achieve the exchange of heterogeneous data between the various data islands;

[0014] A distributed feature extraction framework is built based on a deep learning model to perform parallel feature extraction on heterogeneous data in various data silos, add timestamp alignment tags, and add unique traceability identifiers to data sources.

[0015] Among them, the deep learning model is used to learn and represent the features of heterogeneous data in each data island, and to extract key features; multiple parallel feature extraction nodes are set up in the distributed feature extraction framework to simultaneously extract the heterogeneous data of each data island.

[0016] In a preferred solution, the collaboration credibility index obtained by quantifying the collected historical interaction events between the data islands includes:

[0017] Obtain interaction events between data islands and extract data interaction frequency, response delay, and task success rate from the interaction events;

[0018] A sliding window mechanism is used to segment historical interaction events and output the interaction frequency, response delay, and task success rate in each time window.

[0019] Extract abnormal records from historical interaction events, and count the abnormal number and abnormal deviation of interaction frequency, response delay, and task success rate in the abnormal records. Then, calculate the trust attenuation factor of each data island based on the abnormal number and abnormal deviation.

[0020] The interaction frequency, response delay and task success rate are corrected according to the trust attenuation factor to generate a dynamically adjusted collaboration credibility index.

[0021] In a preferred solution, the calculation of the trust attenuation factor of each data island based on the number of anomalies and the amount of anomaly deviation includes:

[0022] Perform nonlinear coupling processing on the number of anomalies and the amount of anomaly deviation, and calculate the comprehensive anomaly intensity of the interaction event;

[0023] Assigning a corresponding evaluation interval to the comprehensive abnormality intensity through a preset evaluation interval to obtain an interval threshold of the evaluation interval;

[0024] By combining the interval threshold and the comprehensive anomaly intensity for calculation, the trust attenuation factor of each data island is obtained.

[0025] In a preferred solution, allocating a corresponding execution weight to each data island based on the collaboration credibility index includes:

[0026] Normalize the collaboration credibility index, map the normalized collaboration credibility index to the preset weight range, and output the initial weight of each data island;

[0027] Based on the current consensus request node, reversely construct the backtracking period and collect the data update frequency and abnormal fluctuation amplitude in each data island during the backtracking period;

[0028] The dynamic correction factor is calculated based on the data update frequency and the abnormal fluctuation amplitude, and the initial weight is dynamically adjusted based on the dynamic correction factor, and the dynamically adjusted initial weight is output as the execution weight of the data island.

[0029] In a preferred solution, after outputting the execution weights of the data islands, the method further includes:

[0030] Aggregate the execution weight and collaboration credibility index through linear combination and output the aggregated index;

[0031] Collect interactive data flows between data islands in real time and extract interaction integrity and data consistency indicators of interactive data flows;

[0032] The interaction completeness and consistency indicators are weighted and fused with the aggregation indicators to obtain the credibility score;

[0033] When the credibility score reaches the preset threshold, the consensus mechanism is triggered, and the consensus achievement timestamp is recorded. A consensus verification code is generated and broadcast to all nodes in the network for verification and confirmation. After the consensus verification code is verified, the trust status of all nodes in the network is updated. If the public verification code fails, the trust status is restored to the most recent valid consensus status.

[0034] When the credibility score does not reach the preset threshold, it is determined that there is a risk in the current collaboration between data silos, and the trust attenuation factor of each data silo is re-evaluated, and the corresponding execution weight is adjusted until the credibility score meets the standard.

[0035] In a preferred solution, after the weight output is executed, a gradient exchange mechanism is used to perform cross-island feature alignment to generate a standardized consensus data view. The specific process is as follows:

[0036] Initialize local gradient parameters based on the execution weights of each data island and generate a global gradient matrix based on the trusted execution environment;

[0037] The local gradient parameters are fused with the global gradient matrix through iterative processing until the difference between the local gradient parameters and the global gradient matrix converges to a preset allowable threshold range, generating an aligned global feature representation;

[0038] The aligned global feature representation is standardized to generate a standardized consensus data view, where the standardization process includes data cleaning, format unification, and feature normalization.

[0039] In a preferred embodiment, the step of adaptively adjusting the consensus verification rules according to the distribution of execution weights through a pre-deployed trusted execution environment to obtain a consensus result includes:

[0040] Generate dynamic verification thresholds based on the execution weight of each data island, and use threshold segmentation to divide the interaction nodes between data islands into trusted nodes and nodes to be verified;

[0041] Perform lightweight verification on trusted nodes to confirm the consensus status between data silos;

[0042] The nodes to be verified are subjected to multiple rounds of cross-validation until they meet the trust level requirements and are included in the trusted node range. If the trust level does not meet the requirements, the nodes to be verified will be marked as risky nodes and isolated.

[0043] In a preferred solution, the step of tracing the abnormal data source based on the abnormal warning mechanism includes:

[0044] Encode the consensus results into executable protocols and deploy the executable protocols to the execution queues of various data islands based on timestamp serialization;

[0045] Real-time monitoring of network fluctuation parameters, including bandwidth fluctuation, node latency, and packet loss rate;

[0046] Calculate the network stability score based on the network fluctuation parameters;

[0047] If the network stability score is lower than the preset stability threshold, the consensus result is deemed invalid, an abnormality warning is triggered, the corresponding data flow is frozen, and the traceability mechanism is activated to locate the abnormal data source.

[0048] The present invention also provides a cross-data island consensus verification system based on credibility scoring, comprising:

[0049] The initialization module is used to quantify the historical interaction events between the collected data islands and obtain the collaboration credibility index;

[0050] The weight allocation module is used to assign corresponding execution weights to each data island based on the collaboration credibility index;

[0051] The verification module is used to adaptively adjust the consensus verification rules based on the distribution of execution weights through a pre-deployed trusted execution environment to obtain the consensus result;

[0052] Verification module, used to verify the validity of consensus results based on network status fluctuations;

[0053] The alarm module is used to trigger the abnormal warning mechanism when the verification result of the consensus result is invalid, and trace the abnormal data source according to the abnormal warning mechanism.

[0054] The technical effects achieved by the present invention are:

[0055] The present invention dynamically evaluates the trustworthiness of each data island and combines it with network status fluctuations to achieve real-time verification and abnormal warning of consensus results, effectively improving the stability and security of cross-data island collaboration, ensuring data consistency, reducing data risks caused by network fluctuations, and enhancing the system's adaptability. During the consensus process, not only the historical interaction events and current collaboration status of the data island are taken into account, but also the gradient exchange mechanism is used to achieve efficient alignment of cross-island features and generate a standardized consensus data view. In addition, the present invention also ensures the adaptive adjustment of consensus verification rules by pre-deploying a trusted execution environment, further improving the accuracy and efficiency of the consensus. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 It is a schematic flow chart of the method of the present invention;

[0057] Figure 2 It is a schematic diagram of the system module of the present invention. DETAILED DESCRIPTION

[0058] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0059] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0060] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in a preferred embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it constitute a separate or selective embodiment that is mutually exclusive of other embodiments.

[0061] See also Figure 1 As shown, the present invention provides a consensus verification method across data islands based on credibility scoring, including:

[0062] S1. Quantify the historical interaction events between the collected data islands to obtain the collaboration credibility index;

[0063] Before step S1, when executing the consensus across data silos, a collaborative processing channel is pre-built between each data silo, and a heterogeneous data interface for converting the data formats of different data silos is deployed in the collaborative processing channel, and a unified communication protocol is used to realize the exchange of heterogeneous data between each data silo;

[0064] A distributed feature extraction framework is built based on a deep learning model to perform parallel feature extraction on heterogeneous data in various data silos, add timestamp alignment tags, and add unique traceability identifiers to data sources.

[0065] Among them, the deep learning model is used to learn and represent the features of heterogeneous data in each data island, and to extract key features; multiple parallel feature extraction nodes are set in the distributed feature extraction framework to simultaneously extract the heterogeneous data of each data island;

[0066] The collaborative processing channel in the embodiment of the present application can be specifically implemented by constructing a high-speed data bus. The data bus supports multiple data protocol conversions to ensure that data between different data islands can be effectively transmitted. For example, in a credit and finance application scenario, the high-speed data bus can integrate a JSON protocol parsing module to process the transaction records of banking institutions, while supporting an XML conversion interface to adapt to the historical data format of the credit reporting system and implementing asynchronous data packet transmission through a unified HTTPS communication protocol. In specific implementations, the bus has a built-in data compression algorithm (such as GZIP) to reduce bandwidth usage, and uses the TLS1.3 encryption layer to prevent man-in-the-middle attacks, ensuring the integrity and confidentiality of user loan information during cross-institutional transmission, thereby eliminating information barriers. Since the data formats stored in different data islands may differ, heterogeneous data interfaces deployed in the collaborative processing channel can convert data in different formats to ensure data interoperability. At the same time, the use of a unified communication protocol can further enhance the efficiency and accuracy of data transmission.

[0067] During the feature extraction phase, the feature learning capabilities of deep learning models are leveraged to perform parallel feature extraction on heterogeneous data in each data island. Deep learning models can be convolutional neural networks, recurrent neural networks, or attention mechanism networks, and the specific model to be selected depends on actual needs. For example, convolutional neural networks are more effective when processing image data, while recurrent neural networks are more effective when processing sequence data. By building a distributed feature extraction framework, not only can the efficiency of feature extraction be improved, but the representativeness of the extracted features can also be ensured, laying the foundation for subsequent consensus verification.

[0068] Multiple feature extraction nodes are also set up in the distributed feature extraction framework. Feature extraction nodes can work in parallel, thereby shortening the feature extraction time. Each feature extraction node has the ability to process data independently and can work in collaboration with other nodes to ensure the integrity and accuracy of the data.

[0069] In addition, in order to further improve the efficiency of data processing, timestamp alignment marks and data source unique traceability marks are also used. The timestamp alignment mark is used to ensure that when processing data from different data islands, it can be aligned based on a unified time base, so as to eliminate data inconsistencies caused by time differences. The data source unique traceability mark is used to quickly locate the source of the data when problems occur in the data, facilitating data tracing and repair. For example, a unique identifier is added to each credit data (such as user ID number + institution code + encryption information) to facilitate subsequent abnormal data tracing (for example, when a user's credit record is found to be abnormal, the specific institution data source can be quickly located).

[0070] In step S1, the historical interaction events between the collected data islands are quantified to obtain the following collaboration credibility indicators:

[0071] Obtain interaction events between data islands and extract data interaction frequency, response delay, and task success rate from the interaction events;

[0072] A sliding window mechanism is used to segment historical interaction events and output the interaction frequency, response delay, and task success rate in each time window.

[0073] Extract abnormal records from historical interaction events, and count the abnormal number and abnormal deviation of interaction frequency, response delay, and task success rate in the abnormal records. Then, calculate the trust attenuation factor of each data island based on the abnormal number and abnormal deviation.

[0074] Based on the trust decay factor, the interaction frequency, response delay and task success rate are corrected to generate a dynamically adjusted collaboration credibility index.

[0075] Specifically, when calculating the collaboration credibility index, we first capture the interaction events between data silos and extract key information such as the number of data interactions, response delay, and task success rate within the interaction events as the basis for subsequent evaluation. Then, we use a sliding window mechanism to divide past interaction events into multiple segments and calculate the average number of interactions, response speed, and task success rate in each time period. The size of the sliding window can be adjusted according to actual needs to ensure that the changing trends of historical data can be clearly expressed. Comparing the changes in the number of data interactions, response delay, and task success rate in different time periods can help determine whether the collaboration between data silos is stable or unstable. When calculating the trust decay factor, we also consider the number of anomalies and the size of the anomaly values. The number of anomalies and the size of the anomaly values ​​are also averaged over each time period, which can more comprehensively reflect the historical situation and avoid misjudgments due to accidental factors. After the trust decay factor is calculated, it is used to correct the number of interactions, response delay, and task success rate. Finally, the adjusted value is recorded as the collaboration credibility index to provide preliminary data support for the subsequent analysis process.

[0076] It should be noted that the trust attenuation factor for each data island calculated based on the number of anomalies and the amount of anomaly deviation includes:

[0077] Perform nonlinear coupling processing on the number of anomalies and the amount of anomaly deviation, and calculate the comprehensive anomaly intensity of the interaction event;

[0078] Assigning a corresponding evaluation interval to the comprehensive abnormality intensity through a preset evaluation interval to obtain an interval threshold of the evaluation interval;

[0079] By combining the interval threshold and the comprehensive anomaly intensity, the trust attenuation factor of each data island is obtained;

[0080] In the above, after the number of anomalies and the amount of anomaly deviation are output, nonlinear coupling processing is performed on the number of anomalies and the amount of anomaly deviation to calculate the comprehensive anomaly intensity of the interaction event. The calculation formula for the comprehensive anomaly intensity is:

[0081]

[0082] Among them, E represents the comprehensive abnormal intensity, N represents the number of abnormalities, and D i Indicates the amount of abnormal deviation;

[0083] After calculating the comprehensive anomaly intensity, it will be compared with the preset evaluation interval, matching the corresponding interval threshold, and combining the interval threshold to calculate and output the trust attenuation factor of each island. The calculation formula of the piecewise function is as follows:

[0084]

[0085] Among them, γ represents the trust decay factor, k represents the adjustment factor, and the specific value is adjusted according to the actual application situation. T represents the interval threshold. Based on the output results of the above piecewise function, the trust decay factor of each data island can be determined, and then it can be corrected according to the trust decay factor and the interaction frequency, response delay and task success rate. Specifically, the interaction frequency, response delay and task success rate can be adjusted by multiplying them by the trust decay factor to ensure that the corrected indicators are closer to the actual collaboration performance, so that the corrected indicators can more accurately reflect the real-time trust status between data islands, providing a more reliable basis for the subsequent execution weight allocation.

[0086] S2. Assign corresponding execution weights to each data island based on the collaboration credibility index;

[0087] In step S2, after the collaborative credibility index is output, the execution weights of the data islands are assigned. Then, a gradient interaction mechanism is used to align cross-island features to optimize the data fusion effect and improve the accuracy of collaborative processing, thereby ensuring the consistency of information transmission between data islands. A unified data fusion framework is constructed in the form of a standardized consensus data view to further strengthen the trust chain between islands. The execution weights assigned to each data island based on the collaborative credibility index include:

[0088] Normalize the collaboration credibility index, map the normalized collaboration credibility index to the preset weight range, and output the initial weight of each data island;

[0089] Based on the current consensus request node, reversely construct the backtracking period and collect the data update frequency and abnormal fluctuation amplitude in each data island during the backtracking period;

[0090] The dynamic correction factor is calculated based on the data update frequency and abnormal fluctuation amplitude, and the initial weight is dynamically adjusted based on the dynamic correction factor. The dynamically adjusted initial weight is output as the execution weight of the data island;

[0091] And, after outputting the execution weights of the data islands, the method further includes:

[0092] Aggregate the execution weight and collaboration credibility index through linear combination and output them as aggregated index;

[0093] Collect interactive data flows between data islands in real time and extract interaction integrity and data consistency indicators of interactive data flows;

[0094] The interaction completeness and consistency indicators are weighted and fused with the aggregation indicators to obtain the credibility score;

[0095] When the credibility score reaches the preset threshold, the consensus mechanism is triggered, and the consensus achievement timestamp is recorded. A consensus verification code is generated and broadcast to all nodes in the network for verification and confirmation. After the consensus verification code is verified, the trust status of all nodes in the network is updated. If the public verification code fails, the system will revert to the most recent valid consensus status.

[0096] When the credibility score does not reach the preset threshold, it is determined that there is a risk in the collaboration between the current data silos, and the trust attenuation factor of each data silo is re-evaluated and the corresponding execution weight is adjusted until the credibility score meets the standard;

[0097] Specifically, when assigning execution weights to each data island, the first step is to normalize their collaboration credibility indicators so that different collaboration credibility indicators can be compared under the same dimension. The normalized collaboration credibility indicators will be mapped to the preset weight range. In this way, the initial weight of each data island can be matched. Then, starting from the node that currently initiates the consensus request, the time period is reversed for a period of time, and this reversed time period is recorded as the backtracking period. The length of the backtracking period will be set according to actual needs. During the backtracking period, the data update frequency and abnormal fluctuation amplitude of each data island will be collected. A high data update frequency indicates that the data is active, and a small fluctuation amplitude indicates that it is more stable during collaboration. Then, a dynamic correction factor is calculated using the collected data update frequency and abnormal fluctuation amplitude. The calculation formula of the dynamic correction factor is dynamic correction factor = data update frequency / baseline frequency + abnormal fluctuation amplitude / baseline fluctuation amplitude. The dynamic correction factor takes into account both data activity and stability. Its purpose is to adjust the initial weight. The method is to multiply the correction factor by the initial weight to obtain the adjusted weight, which is the final execution weight. Re-output, of course, the algorithm of the correction factor can be optimized, such as using exponential smoothing or logarithmic transformation. The specific selection depends on actual needs. I will not elaborate on it here. After the weight output is executed, the collaborative credibility evaluation will be performed synchronously. When evaluating the collaborative credibility, the execution weight and the collaborative credibility index are first combined in a linear combination to obtain an aggregation index. At the same time, the data flow of interactions between islands is collected in real time, and the interaction integrity and data consistency indicators are extracted from them. After that, the interaction integrity, consistency index and aggregation index are weighted and integrated to obtain the final credibility score. This credibility score integrates multiple aspects such as collaboration credibility, data activity, and stability, and serves as an important basis for deciding whether to start the consensus mechanism. When the credibility score reaches the preset threshold, the consensus mechanism is triggered, and the time point of consensus is recorded. A consensus verification code is generated and broadcast to all nodes in the entire network for verification and confirmation to ensure that the consensus is valid and secure. When the credibility score does not meet the standard, a risk warning is activated. At this time, the trust attenuation of each data island must be re-evaluated and their execution weights adjusted until the score meets the standard, thereby ensuring that the collaboration between islands is stable and secure.

[0098] Secondly, after executing the weight output, the gradient exchange mechanism is used to perform cross-island feature alignment to generate a standardized consensus data view. The specific process is as follows:

[0099] Initialize local gradient parameters based on the execution weights of each data island and generate a global gradient matrix based on the trusted execution environment;

[0100] The local gradient parameters are fused with the global gradient matrix through iterative processing until the difference between the local gradient parameters and the global gradient matrix converges to a preset allowable threshold range, generating an aligned global feature representation;

[0101] Standardize the aligned global feature representation to generate a standardized consensus data view. Standardization includes data cleaning, format unification, and feature normalization.

[0102] In the above, after the execution weight of the data island is output, the local gradient parameters are first initialized to ensure the rationality of the initial state of each island feature vector, and then the global gradient matrix is ​​generated based on the trusted execution environment. The global gradient matrix serves as the benchmark for cross-island feature alignment and contains the global gradient information of each data island feature. The fusion process of the local gradient parameters and the global gradient matrix is ​​achieved through an iterative algorithm. Each iteration calculates the difference between the local gradient parameters and the global gradient matrix, and adjusts the local gradient parameters according to the difference until the difference converges to the preset allowable threshold. This process ensures the alignment and consistency of each data island feature from a global perspective. The aligned global feature representation not only retains the uniqueness of each island data, but also realizes the feature fusion across islands, enhancing the integrity and synergy of the data. The aligned global feature representation will be standardized later. Standardization is a data preprocessing The key steps include data cleaning, format unification and feature normalization. Data cleaning can remove redundant and noisy data and improve data quality. Format unification ensures the consistency and comparability of data from different sources. Feature normalization scales the eigenvalues ​​to the same order of magnitude. After standardization, a standardized consensus data view is generated. The consensus data view serves as the basis for collaborative processing across data silos and provides data support for subsequent consensus verification. In addition, the construction of a standardized consensus data view further strengthens the trust chain between silos, enhances the collaboration capabilities between data silos and the security of data sharing. Taking banks and credit reporting agencies as an example, banks focus on repayment record gradients, while credit reporting agencies focus on overdue record gradients. Through the gradient exchange mechanism, the two can share feature gradients to generate a consensus data view that includes both repayment record features and overdue record features, which not only meets the data needs of both parties but also ensures the security and privacy of the data.

[0103] S3. Through the pre-deployed trusted execution environment, the consensus verification rules are adaptively adjusted according to the distribution of execution weights to obtain the consensus result;

[0104] In step S3, within the trusted execution environment, the consensus verification rules will be adaptively adjusted according to the distribution of execution weights. First, the trusted execution environment will receive a standardized consensus data view as input, and then dynamically adjust the parameters in the consensus algorithm according to the execution weight of each data island. For example, data islands with higher weights will be given more opportunities to participate in consensus verification, and the data they provide will account for a larger proportion in the consensus process, thereby ensuring that the consensus results can more accurately reflect the actual collaboration between data islands. At the same time, the consensus verification rules will also take into account the historical interaction performance between data islands, that is, the collaboration credibility index. For data islands with higher collaboration credibility, more trust weight will be given, reducing The difficulty of verification in the consensus verification process is reduced, thereby speeding up the consensus and improving overall efficiency. In addition, the consensus verification rules will be fine-tuned according to the interaction integrity and data consistency indicators in the real-time collected interaction data stream to ensure that when there are abnormalities in the information transmission between data islands, they can be discovered and remedied in time to avoid the impact of erroneous data on the consensus results. During the consensus verification process, the trusted execution environment will also use advanced encryption technology and security protocols to ensure the security and privacy of data during transmission and processing, and prevent data leakage or tampering. Among them, the consensus verification rules are adaptively adjusted according to the distribution of execution weights through the pre-deployed trusted execution environment to obtain the consensus results. The steps include:

[0105] Generate dynamic verification thresholds based on the execution weight of each data island, and use threshold segmentation to divide the interaction nodes between data islands into trusted nodes and nodes to be verified;

[0106] Perform lightweight verification on trusted nodes to confirm the consensus status between data silos;

[0107] The nodes to be verified will be subject to multiple rounds of cross-validation until they meet the trustworthiness standard and are included in the trusted node range. If the trustworthiness standard is not met, the nodes to be verified will be marked as risky nodes and isolated.

[0108] Specifically, in the collaborative confirmation process, a dynamic verification threshold will first be generated based on the execution weight of each data island. The dynamic verification threshold can be dynamically adjusted based on the real-time status of each island and the overall network load to ensure the flexibility and efficiency of the verification process. For example, when the execution weight of a data island is high, the dynamic verification threshold will be increased accordingly, increasing the difficulty of passing the verification to ensure the high accuracy of the consensus result. Conversely, for data islands with lower execution weights, the dynamic verification threshold will be appropriately lowered to give them more opportunities to participate in consensus verification to promote the overall collaboration of the network. Specifically, the corresponding mapping relationship can be pre-set based on expert experience or historical data to generate a dynamic verification threshold that adapts to the current network status. After the dynamic verification threshold is determined, the node trust level will be identified through threshold segmentation to achieve stratification. Verification ensures the efficiency of the collaborative confirmation process. Specifically, nodes are divided into two categories: trusted nodes and nodes to be verified. Lightweight verification is performed on trusted nodes. For example, hash verification can be used to quickly confirm their data consistency, while multiple rounds of cross-validation are performed on nodes to be verified. The trustworthiness is comprehensively evaluated based on historical behavior and real-time performance to ensure the reliability of the node before it is included in the trusted range. Of course, other machine learning algorithms, such as random forests or deep learning models, can be used to further improve the accuracy and efficiency of verification, and ultimately form an efficient and secure collaborative confirmation mechanism. Of course, there may still be risk nodes that cannot pass verification. In order to ensure the overall security of the data island, the risk nodes will be isolated to prevent the spread of their potential risks. At the same time, their abnormal behavior will be recorded and sent to the management end for subsequent corresponding analysis and processing.

[0109] S4. Verify the validity of the consensus result based on network status fluctuations;

[0110] In step S4, when the network status fluctuates, the validity of the consensus result will be verified to ensure the stability and reliability of the consensus result. In actual applications, network status fluctuations may be caused by various factors, such as network delays, data loss or node failures, all of which may affect the accuracy of the consensus result. Therefore, after the consensus result is reached, the network status will be monitored in real time. Once the network status fluctuation is found, the verification mechanism will be immediately triggered to evaluate the validity of the consensus result to ensure that the collaboration status between data islands is not affected.

[0111] S5. When the verification result of the consensus result is invalid, the abnormal warning mechanism is triggered, and the abnormal data source is traced according to the abnormal warning mechanism;

[0112] In step S5, when an abnormal network state is found or the consensus result is inconsistent with the actual situation, the abnormal warning mechanism will be immediately triggered, and the abnormal data source will be traced to determine the source of the problem so as to facilitate targeted repair and optimization. The step of tracing the abnormal data source according to the abnormal warning mechanism includes:

[0113] Encode the consensus results into executable protocols and deploy the executable protocols to the execution queues of various data islands based on timestamp serialization;

[0114] Real-time monitoring of network fluctuation parameters, including bandwidth fluctuation, node latency, and packet loss rate;

[0115] Calculate the network stability score based on the network fluctuation parameters;

[0116] If the network stability score is lower than the preset stability threshold, the consensus result is deemed invalid, an abnormality warning is triggered, the corresponding data flow is frozen, and the traceability mechanism is activated to locate the abnormal data source;

[0117] Specifically, the consensus result will first be encoded into an executable protocol. Blockchain technology or other technologies can be used to ensure that the consensus result cannot be tampered with, and timestamp information can be embedded to ensure its uniqueness and traceability. It will then be deployed to the execution queue of each data island in a serialized order to wait for execution. At the same time, network fluctuation parameters will be monitored in real time. Network fluctuation parameters include key indicators such as bandwidth fluctuation, node latency, and data packet loss rate, and the network stability score will be calculated based on key indicators. It can be calculated through weighted fusion or other algorithms. If the network stability score is lower than the preset stability threshold, the consensus result will be judged invalid and the abnormal warning mechanism will be triggered immediately. At this time, the corresponding data stream will be immediately frozen to prevent the further spread of abnormal data, and the traceability mechanism will be started simultaneously to locate the abnormality. For common data sources, the traceability mechanism will extract the abnormal data features in the consensus results and compare them with the historical abnormal data to determine whether there are known abnormal patterns. For known abnormal patterns, the corresponding processing strategy will be matched immediately, and the repair process will be automatically triggered to quickly restore the normal collaboration status between data islands. For unknown abnormal patterns, they will be submitted for manual review. At the same time, the execution weight and collaboration credibility indicators of each data island will be combined to comprehensively evaluate the data islands where the abnormal data may come from. Then, through cross-validation and multi-source data comparison, the scope of the abnormal data source will be further narrowed until the specific abnormal data source is located. After the abnormal data source is determined, it can be isolated and processed, and its abnormal behavior pattern can be recorded for subsequent targeted repair and optimization.

[0118] See also Figure 2, a cross-data-island consensus verification system based on credibility scoring, including:

[0119] The initialization module is used to quantify the historical interaction events between the collected data islands and obtain the collaboration credibility index;

[0120] The weight allocation module is used to assign corresponding execution weights to each data island based on the collaboration credibility index;

[0121] The verification module is used to adaptively adjust the consensus verification rules based on the distribution of execution weights through a pre-deployed trusted execution environment to obtain the consensus result;

[0122] Verification module, used to verify the validity of consensus results based on network status fluctuations;

[0123] The alarm module is used to trigger the abnormal warning mechanism when the verification result of the consensus result is invalid, and trace the abnormal data source according to the abnormal warning mechanism.

[0124] In the above, the initialization module is mainly responsible for collecting and quantifying historical interaction events between various data islands. Historical interaction events can reflect the collaboration history and behavior patterns between data islands. By quantifying the interaction events, a collaboration credibility index can be generated. The collaboration credibility index will be recorded as an indicator to measure the trust relationship between data islands. The weight allocation module is based on the collaboration credibility index. It assigns a reasonable execution weight to each data island. The execution weight reflects the importance and influence of the data island in the consensus verification process. The verification module uses the pre-deployed trusted execution environment to adaptively adjust the consensus verification rules according to the distribution of execution weights to ensure the accuracy and reliability of the consensus results. The verification module monitors network status fluctuations in real time, verifies the validity of the consensus results, and promptly discovers and handles potential anomalies. The alarm module triggers the abnormal warning mechanism when the consensus result verification is invalid, traces and processes the abnormal data source, and ensures the stability and security of the collaboration between data islands.

[0125] It should be noted that, in this document, the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, apparatus, article, or method comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, apparatus, article, or method. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, apparatus, article, or method comprising the element.

[0126] The foregoing is merely a preferred embodiment of the present invention. It should be noted that those skilled in the art may make various improvements and modifications without departing from the principles of the present invention, and such improvements and modifications are also within the scope of protection of the present invention. Structures, devices, and operating methods not specifically described or explained herein shall, unless otherwise specified or limited, be implemented in accordance with conventional means in the art.

Claims

1. A cross-data island consensus verification method based on credibility scoring, characterized by: include: Quantify the historical interaction events between the collected data islands to obtain the collaboration credibility index; Assign corresponding execution weights to each data island based on the collaboration credibility index; Through the pre-deployed trusted execution environment, the consensus verification rules are adaptively adjusted according to the distribution of execution weights to obtain the consensus result; Verify the validity of the consensus results based on network status fluctuations; When the verification result of the consensus result is invalid, the abnormal warning mechanism is triggered, and the abnormal data source is traced based on the abnormal warning mechanism.

2. The cross-data island consensus verification method based on credibility scoring according to claim 1 is characterized by: Before quantifying the collected historical interaction events between the data islands to obtain a collaboration credibility index, the method further includes: By pre-building a collaborative processing channel between the various data islands, deploying heterogeneous data interfaces for converting data formats of different data islands in the collaborative processing channel, and adopting a unified communication protocol to achieve the exchange of heterogeneous data between the various data islands; A distributed feature extraction framework is built based on a deep learning model to perform parallel feature extraction on heterogeneous data in various data silos, add timestamp alignment tags, and add unique traceability identifiers to data sources. Among them, the deep learning model is used to learn and represent the features of heterogeneous data in each data island, and to extract key features; multiple parallel feature extraction nodes are set up in the distributed feature extraction framework to simultaneously extract the heterogeneous data of each data island.

3. The cross-data island consensus verification method based on credibility scoring according to claim 1 is characterized by: The collaboration credibility indicators obtained by quantifying the historical interaction events between the collected data islands include: Obtain interaction events between data islands and extract data interaction frequency, response delay, and task success rate from the interaction events; A sliding window mechanism is used to segment historical interaction events and output the interaction frequency, response delay, and task success rate in each time window. Extract abnormal records from historical interaction events, and count the abnormal number and abnormal deviation of interaction frequency, response delay, and task success rate in the abnormal records. Then, calculate the trust attenuation factor of each data island based on the abnormal number and abnormal deviation. The interaction frequency, response delay and task success rate are corrected according to the trust attenuation factor to generate a dynamically adjusted collaboration credibility index.

4. The cross-data island consensus verification method based on credibility scoring according to claim 3 is characterized by: The trust attenuation factor of each data island is calculated based on the number of anomalies and the amount of anomaly deviation, including: Perform nonlinear coupling processing on the number of anomalies and the amount of anomaly deviation, and calculate the comprehensive anomaly intensity of the interaction event; Assigning a corresponding evaluation interval to the comprehensive abnormality intensity through a preset evaluation interval to obtain an interval threshold of the evaluation interval; By combining the interval threshold and the comprehensive anomaly intensity for calculation, the trust attenuation factor of each data island is obtained.

5. The cross-data island consensus verification method based on credibility scoring according to claim 1 is characterized by: The allocation of corresponding execution weights to each data island based on the collaboration credibility index includes: Normalize the collaboration credibility index, map the normalized collaboration credibility index to the preset weight range, and output the initial weight of each data island; Based on the current consensus request node, reversely construct the backtracking period and collect the data update frequency and abnormal fluctuation amplitude in each data island during the backtracking period; The dynamic correction factor is calculated based on the data update frequency and the abnormal fluctuation amplitude, and the initial weight is dynamically adjusted based on the dynamic correction factor, and the dynamically adjusted initial weight is output as the execution weight of the data island.

6. The cross-data island consensus verification method based on credibility scoring according to claim 5 is characterized by: After outputting the execution weights of the data islands, the method further includes: Aggregate the execution weight and collaboration credibility index through linear combination and output the aggregated index; Collect interactive data flows between data islands in real time and extract interaction integrity and data consistency indicators of interactive data flows; The interaction completeness and consistency indicators are weighted and fused with the aggregation indicators to obtain the credibility score; When the credibility score reaches the preset threshold, the consensus mechanism is triggered, and the consensus achievement timestamp is recorded. A consensus verification code is generated and broadcast to all nodes in the network for verification and confirmation. After the consensus verification code is verified, the trust status of all nodes in the network is updated. If the public verification code fails, the trust status is restored to the most recent valid consensus status. When the credibility score does not reach the preset threshold, it is determined that there is a risk in the current collaboration between data silos, and the trust attenuation factor of each data silo is re-evaluated, and the corresponding execution weight is adjusted until the credibility score meets the standard.

7. The cross-data island consensus verification method based on credibility scoring according to claim 6 is characterized by: After the weight output is executed, the gradient exchange mechanism is used to perform cross-island feature alignment to generate a standardized consensus data view. The specific process is as follows: Initialize local gradient parameters based on the execution weights of each data island and generate a global gradient matrix based on the trusted execution environment; The local gradient parameters are fused with the global gradient matrix through iterative processing until the difference between the local gradient parameters and the global gradient matrix converges to a preset allowable threshold range, generating an aligned global feature representation; The aligned global feature representation is standardized to generate a standardized consensus data view, where the standardization process includes data cleaning, format unification, and feature normalization.

8. The cross-data island consensus verification method based on credibility scoring according to claim 1 is characterized by: The step of adaptively adjusting the consensus verification rules according to the distribution of execution weights through the pre-deployed trusted execution environment to obtain a consensus result includes: Generate dynamic verification thresholds based on the execution weight of each data island, and use threshold segmentation to divide the interaction nodes between data islands into trusted nodes and nodes to be verified; Perform lightweight verification on trusted nodes to confirm the consensus status between data silos; The nodes to be verified are subjected to multiple rounds of cross-validation until they meet the trust level requirements and are included in the trusted node range. If the trust level does not meet the requirements, the nodes to be verified will be marked as risky nodes and isolated.

9. The cross-data island consensus verification method based on credibility scoring according to claim 1 is characterized by: The step of tracing the abnormal data source according to the abnormal warning mechanism includes: Encode the consensus results into executable protocols and deploy the executable protocols to the execution queues of various data islands based on timestamp serialization; Real-time monitoring of network fluctuation parameters, including bandwidth fluctuation, node latency, and packet loss rate; Calculate the network stability score based on the network fluctuation parameters; If the network stability score is lower than the preset stability threshold, the consensus result is deemed invalid, an abnormality warning is triggered, the corresponding data flow is frozen, and the traceability mechanism is activated to locate the abnormal data source.

10. A cross-data island consensus verification system based on credibility scoring, characterized by: include: The initialization module is used to quantify the historical interaction events between the collected data islands and obtain the collaboration credibility index; The weight allocation module is used to assign corresponding execution weights to each data island based on the collaboration credibility index; The verification module is used to adaptively adjust the consensus verification rules based on the distribution of execution weights through a pre-deployed trusted execution environment to obtain the consensus result; Verification module, used to verify the validity of consensus results based on network status fluctuations; The alarm module is used to trigger the abnormal warning mechanism when the verification result of the consensus result is invalid, and trace the abnormal data source according to the abnormal warning mechanism.

Images