A data trust processing method and system integrating trusted computing and blockchain

Abstract

This application relates to the field of information security and discloses a data trust processing method and system that integrates trusted computing and blockchain. The method includes: collecting batch data through distributed blockchain nodes and updating and generating an initial fusion result using a hash ring binding mechanism; verifying the consistency of the initial fusion result and obtaining the corrected fusion state; introducing a differential privacy framework to generate perturbation parameters and randomizing data points to generate a privacy-enhanced dataset; adjusting the noise distribution according to the set operation type to generate a perturbation dataset adapted for the operation; generating a balanced perturbation level through a noise adjustment algorithm; fusing and assessing privacy risks in a trusted execution environment to obtain a secure computation output; and feeding the secure output back to the nodes, updating the state, and synchronously generating a consistent fusion result. This application improves the trustworthiness of data processing by solving the problem of balancing distributed data consistency, privacy protection, and efficiency.

Description

Technical Field

[0001] This application relates to the field of information security technology, and in particular to a data trust processing method and system that integrates trusted computing and blockchain. Background Technology

[0002] Distributed data processing is widely used in various fields, requiring multi-node collaboration to support data computation and decision-making. Blockchain, with its distributed architecture, can provide a foundation for data traceability and node collaboration, while trusted computing can ensure the security of the processing environment. However, in a distributed environment, data is scattered across blockchain nodes, making the balance between maintaining data consistency, protecting privacy, and balancing computational efficiency a critical requirement in collaborative processing.

[0003] Existing distributed data processing solutions combining blockchain suffer from several shortcomings. First, data consistency maintenance is insufficient. While some solutions use blockchain nodes to collect data, they lack efficient synchronization mechanisms. For example, in conventional allocation methods, batch data updates can easily lead to mapping deviations, and the lack of targeted verification and correction mechanisms makes it difficult to guarantee the accuracy of the initial fusion results. Second, privacy protection capabilities are inadequate. Most solutions do not dynamically generate perturbation parameters using differential privacy frameworks, or their noise distribution is fixed, making it impossible to adapt and adjust according to the type of set operation, thus failing to meet computational demands while protecting privacy. Third, trusted processing and risk management are lacking. Fusion processing is not conducted using trusted execution environments, and privacy inference risk assessments of the fusion results are not performed, easily leading to privacy leaks. Furthermore, after node state updates, the lack of a synchronization mechanism combining hash ring binding and consistency protocols makes it difficult to achieve global state uniformity, hindering the effectiveness of trusted data processing.

[0004] This application collects data through blockchain nodes, uses hash ring binding to achieve data updates and initial fusion, combines consistency verification to ensure data accuracy, introduces differential privacy and dynamic noise adjustment to enhance privacy protection, and leverages a trusted execution environment to manage fusion risks and synchronize node states. This solves the problems of data consistency deviation, lack of adaptability to privacy protection, and insufficient trusted processing capabilities in existing distributed data processing solutions. Summary of the Invention

[0005] This application provides a data trust processing method and system that integrates trusted computing and blockchain, which solves the problems of data consistency deviation, lack of adaptability of privacy protection, and insufficient trusted processing capabilities in existing distributed data processing schemes.

[0006] Firstly, this application provides a data trust processing method that integrates trusted computing and blockchain, the method comprising:

[0007] Step S101: Collect batch data through blockchain distributed nodes, update the batch data using a hash ring binding mechanism, and generate initial fusion results;

[0008] Step S102: Perform a consistency check on the initial fusion result to obtain the corrected fusion state;

[0009] Step S103: Introduce a differential privacy framework to the corrected fusion state to generate perturbation parameters, and randomize the data points in the initial fusion result using the perturbation parameters to generate a privacy-enhanced dataset;

[0010] Step S104: Adjust the noise distribution according to the type of set operation performed on the privacy-enhanced dataset, and adjust the privacy-enhanced dataset according to the noise distribution to generate a perturbation dataset adapted for the operation;

[0011] Step S105: After processing the perturbation dataset of the adaptation operation with a noise adjustment algorithm, an equalized perturbation level is generated;

[0012] Step S106: Perform fusion processing on the equalized perturbation level in a trusted execution environment, and conduct privacy inference risk assessment on the fusion result to obtain secure computational output;

[0013] Step S107: Feed the secure computation output back to the blockchain distributed node, update the status of each node through the hash ring binding mechanism, and generate a fusion result for consistency maintenance after synchronization through the inter-node consistency protocol.

[0014] Optionally, step S101 includes:

[0015] Batch data is collected through blockchain distributed nodes;

[0016] A hash ring binding mechanism is used to allocate the batch data and determine the data mapping relationship of each node;

[0017] The batch data is progressively updated according to the data mapping relationship to generate local fused data for each node;

[0018] The locally fused data is synchronized using an inter-node communication protocol to obtain a set of hash values;

[0019] The initial fusion result is generated by integrating the local fusion data of each node based on the hash value set.

[0020] Optionally, step S102 includes:

[0021] Obtain the set of hash values ​​from the initial fusion result;

[0022] The hash value set is compared using a consistency check algorithm to determine whether the hash values ​​between batches match.

[0023] If a match is found, the initial fusion result is retained as the corrected fusion state.

[0024] If there is a mismatch, the hash ring binding parameters are adjusted to redetermine the data mapping relationship, and the local fusion data is updated to generate the corrected fusion state.

[0025] Optionally, step S103 includes:

[0026] The modified fusion state is analyzed using a differential privacy framework to determine the sensitivity parameters for privacy protection.

[0027] The perturbation parameter is generated based on the sensitivity parameter, and the updated batch data is randomized using the perturbation parameter to generate a dataset with added noise.

[0028] Verify whether the dataset with added noise meets the privacy constraints, and generate the privacy-enhanced dataset.

[0029] Optionally, step S104 includes:

[0030] Determine the type of set operation to be performed on the privacy-enhanced dataset;

[0031] If the set operation is of the type of intersection operation, then the noise reduction algorithm is used to reduce the noise amplitude and generate the first noise distribution;

[0032] If the set operation is a non-intersection operation, then a noise enhancement algorithm is used to increase the noise amplitude and generate a second noise distribution;

[0033] The privacy-enhanced dataset is adjusted according to the first noise distribution or the second noise distribution to generate the perturbation dataset for adaptation computation.

[0034] Optionally, step S105 includes:

[0035] The noise distribution of the disturbance dataset is analyzed using a noise adjustment algorithm to determine the dynamic adjustment parameters of the noise amplitude.

[0036] The noise in the disturbance dataset is optimized based on the dynamic control parameters to generate a balanced noise distribution;

[0037] The disturbed dataset is resampled using the equalized noise distribution to generate an equalized dataset;

[0038] Verify whether the noise level of the equalized dataset meets the preset equalization conditions to obtain the equalized disturbance level.

[0039] Optionally, step S106 includes:

[0040] The equalized perturbation levels are fused using a trusted execution environment to generate a fusion calculation result.

[0041] The privacy inference risk value is determined by analyzing the fusion operation results using a privacy inference risk assessment algorithm.

[0042] If the privacy inference risk value is lower than the preset fusion threshold, the fusion operation result is output as a safe operation output.

[0043] Optionally, step S107 includes:

[0044] The secure computation output is fed back to the blockchain distributed node, and a new mapping relationship is determined through a hash ring binding mechanism;

[0045] The local state of each node is updated according to the new mapping relationship, and an updated set of node states is generated.

[0046] The updated set of node states is synchronized using an inter-node consistency protocol to determine the hash value of the global state.

[0047] The states of each node are integrated based on the hash value of the global state to generate a fusion result that maintains consistency.

[0048] Secondly, this application provides a data trust processing system that integrates trusted computing and blockchain, the system comprising:

[0049] The data acquisition and fusion module is used to collect batch data through blockchain distributed nodes, update the batch data using a hash ring binding mechanism, and generate an initial fusion result.

[0050] The verification and correction module is used to perform consistency verification on the initial fusion result and obtain the corrected fusion state.

[0051] The privacy enhancement module is used to introduce a differential privacy framework into the corrected fusion state, generate perturbation parameters, and randomize the updated batch data using the perturbation parameters to generate a privacy-enhanced dataset.

[0052] The noise adaptation module is used to adjust the noise distribution according to the privacy-enhanced dataset, and adjust the privacy-enhanced dataset according to the noise distribution to generate a perturbation dataset for adaptation operation.

[0053] The perturbation equalization module is used to process the perturbation dataset of the adaptation operation through a noise adjustment algorithm to generate an equalized perturbation level.

[0054] The trusted output module is used to perform fusion processing on the equalized perturbation level in a trusted execution environment, and to perform privacy inference risk assessment on the fusion result to obtain secure computational output.

[0055] The synchronization maintenance module is used to feed back the secure computation output to the blockchain distributed nodes, update the status of each node through the hash ring binding mechanism, and generate a fusion result for consistency maintenance after synchronization through the inter-node consistency protocol.

[0056] Thirdly, this application provides a data trusted processing device that integrates trusted computing and blockchain, including a memory and a processor. The memory stores a computer program that can run on the processor. When the processor executes the computer program, it implements any of the data trusted processing methods that integrate trusted computing and blockchain as described in the application.

[0057] This application proposes a data trust processing method and system integrating trusted computing and blockchain, applicable to distributed data processing scenarios in the field of information security. It addresses the problems of data consistency instability, lack of adaptability for privacy protection, and insufficient trust processing capabilities in existing distributed data processing solutions. Compared with existing technologies, the beneficial effects of this application's technical solution are at least as follows:

[0058] First, batch data is collected through distributed nodes of the blockchain. A hash ring binding mechanism is used to determine the data mapping relationship of each node to achieve gradual data updates and local data fusion synchronization. Then, a consistency verification algorithm is used to compare the hash value set in the initial fusion result. If there is a mismatch, the hash ring binding parameters are adjusted to redetermine the data mapping relationship and update the local fusion data. This ensures the accuracy of the initial fusion result and improves the data consistency maintenance capability.

[0059] Second, the sensitivity parameters are determined by analyzing the corrected fusion state through the differential privacy framework, and the perturbation parameters are generated to randomize the data points to generate a privacy-enhanced dataset. The noise distribution is adjusted according to the type of set operation (intersection operation reduces the noise amplitude, non-intersection operation increases the noise amplitude), and then the perturbation dataset is processed by the noise adjustment algorithm to generate a balanced perturbation level. This achieves the adaptation of privacy protection and computational needs, and improves the pertinence and effectiveness of privacy protection.

[0060] Third, the equilibrium perturbation level is fused through a trusted execution environment, and the fusion result is analyzed by combining a privacy inference risk assessment algorithm. Secure operation output is only output when the risk value is lower than a preset threshold. After the secure operation output is fed back to the node, the local state of the node is updated through a hash ring binding mechanism, and then the global state hash value is generated synchronously and integrated through the inter-node consistency protocol. This reduces the risk of privacy leakage and unifies the global node state, thereby improving the trusted data processing effect. Attached Figure Description

[0061] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0062] Figure 1 This is a flowchart illustrating a data trust processing method integrating trusted computing and blockchain according to this application.

[0063] Figure 2 This is a diagram showing the distribution of virtual nodes and load distribution in the hash ring in this application;

[0064] Figure 3 This is a summary diagram of the results of the distributed data trusted fusion processing in this application;

[0065] Figure 4 This is a schematic diagram of the structure of a data trust processing system integrating trusted computing and blockchain according to this application;

[0066] Figure 5 This is a schematic diagram of the structure of a data trust processing device that integrates trusted computing and blockchain according to this application. Detailed Implementation

[0067] This application provides a data trust processing method and system that integrates trusted computing and blockchain. The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0068] For ease of understanding, the specific process of the embodiments of this application is described below. Please refer to [link / reference]. Figure 1 One embodiment of a data trust processing method integrating trusted computing and blockchain in this application includes:

[0069] Step S101: Collect batch data through blockchain distributed nodes, update the batch data using a hash ring binding mechanism, and generate initial fusion results.

[0070] In one specific embodiment, step S101 may specifically include the following steps:

[0071] Batch data is collected through blockchain distributed nodes;

[0072] A hash ring binding mechanism is used to allocate the batch data and determine the data mapping relationship of each node;

[0073] The batch data is progressively updated according to the data mapping relationship to generate local fused data for each node;

[0074] The locally fused data is synchronized using an inter-node communication protocol to obtain a set of hash values;

[0075] The initial fusion result is generated by integrating the local fusion data of each node based on the hash value set.

[0076] Specifically, in distributed data processing scenarios, blockchain distributed nodes first conduct batch data collection. This data originates from business data acquired by the nodes within a preset period, such as transaction flow data in the financial sector or device sensor data in the Internet of Things (IoT) sector. Nodes divide the data into batches according to time slices or business modules generated by the data, and temporarily store the collected raw data in local storage units, providing a data foundation for subsequent processing. This process addresses the problem of data being scattered across nodes and lacking a unified collection mechanism in distributed environments. By leveraging the distributed collection capabilities of blockchain nodes, it achieves effective aggregation of multi-source batch data, avoiding processing delays caused by data dispersion.

[0077] When allocating the collected batch data using a hash ring binding mechanism, the key fields in the batch data (such as transaction ID and device number) are first converted into 256-bit hash values ​​using the SHA-256 hash algorithm. Simultaneously, multiple virtual nodes are allocated to each blockchain distributed node, and their hash values ​​are also calculated using the SHA-256 algorithm. These data hash values ​​and the virtual node hash values ​​are then mapped together to a modulo operation. On a ring-shaped space (hash ring), the system traverses the hash ring clockwise, identifying the physical node associated with the first virtual node pointed to by the hash value of each batch of data. This establishes the data mapping relationship between nodes. Traditional allocation methods are prone to mapping deviations due to updates to batch data. The hash ring binding mechanism, through the multi-mapping characteristic of virtual nodes, reduces the impact of single-node failures on data allocation. Furthermore, when a node is added or removed, only a small amount of data needs to be remapped, reducing the probability of mapping deviations and addressing the problem of insufficient data allocation stability.

[0078] Based on the aforementioned data mapping relationship, each node performs incremental updates to the allocated batch data. Nodes retrieve locally stored historical fusion data, compare it with the newly allocated batch data, identify newly added, modified, or deleted data entries, and only perform operations on the changed data entries. For example, in financial transaction data processing, newly added transaction records are incremented in amount and marked with transaction status; modified transaction records are updated with fields. No repeated operations are performed on unchanged historical data, generating partial fusion data for each node. Traditional technologies sometimes employ full data updates, leading to wasted computing resources and low processing efficiency. Incremental updates reduce computational load and improve data processing efficiency through incremental operations, while avoiding the risk of historical data misoperation that may occur with full updates.

[0079] Locally merged data is synchronized through inter-node communication protocols (such as gossip). Each node calculates a hash value for its locally generated locally merged data using the SHA-256 algorithm and broadcasts this hash value to neighboring nodes randomly at preset time intervals or data volume thresholds. Receiving nodes collect hash values ​​from different nodes and compare them with the hash value of their local locally merged data. If discrepancies exist, they request complete locally merged data from the sending node for verification and correction until all nodes obtain a consistent set of hash values. Traditional technologies lack efficient synchronization mechanisms, leading to data inconsistencies between nodes. Inter-node communication protocols, through distributed broadcasting and hash value verification, ensure the consistency of hash values ​​in the locally merged data across nodes, providing a consistent data benchmark for subsequent integration and resolving the problem of data synchronization lag between nodes.

[0080] Based on the acquired hash value set, the local fusion data from each node is integrated. Using hash values ​​as indexes, the local fusion data fragments corresponding to the hash values ​​in each node are matched and calculated according to pre-defined aggregation rules (such as summing numerical data and deduplicating non-numerical data). For example, in the integration of IoT device sensor data, the average value of temperature data collected from different nodes of the same device is taken, and the device status data is deduplicated and merged to generate an initial fusion result. Traditional technologies lack targeted verification and correction mechanisms, leading to insufficient accuracy in the initial fusion result. This step, however, integrates data based on a consistent hash value set, ensuring the validity of the local fusion data fragments from each node, reducing errors during data integration, and improving the accuracy of the initial fusion result. This provides a reliable foundation for subsequent data consistency verification and privacy protection processing.

[0081] Step S102: Perform a consistency check on the initial fusion result to obtain the corrected fusion state.

[0082] In one specific embodiment, step S102 may specifically include the following steps:

[0083] Obtain the set of hash values ​​from the initial fusion result;

[0084] The hash value set is compared using a consistency check algorithm to determine whether the hash values ​​between batches match.

[0085] If a match is found, the initial fusion result is retained as the corrected fusion state.

[0086] If there is a mismatch, the hash ring binding parameters are adjusted to redetermine the data mapping relationship, and the local fusion data is updated to generate the corrected fusion state.

[0087] Specifically, in distributed data processing scenarios, obtaining the set of hash values ​​in the initial fusion result is the basis for consistency verification. These hash value sets come from the hash calculations of the local fusion data by each distributed blockchain node in the early stage. Each hash value corresponds to a unique identifier of a set of local fusion data. For example, in a financial transaction scenario, the local fusion data generated by 500 transaction data processed by a certain node corresponds to a SHA-256 hash value. The hash values ​​of all nodes together constitute the set of hash values ​​in the initial fusion result. This set is directly related to the local fusion data of each node and provides data basis for subsequent comparison.

[0088] When comparing hash value sets using a consistency check algorithm, a Merkle tree algorithm is used to construct a tree structure for the hash value sets. Each hash value in the hash value set is treated as a leaf node of the Merkle tree. The hash values ​​of adjacent leaf nodes are concatenated and the hash value is recalculated to form a parent node, and so on until the root hash value of the Merkle tree is generated. Simultaneously, the root hash value of the Merkle tree generated from the previous batch of data processing is retrieved and compared with the root hash value of the previous batch to determine if the hash values ​​match between batches. If the two root hash values ​​are the same, it indicates that there is no difference between the local fused data corresponding to the current hash value set and the previous batch of hash value sets; if the root hash values ​​are different, the leaf nodes in the Merkle tree where the hash values ​​do not match are located, confirming that there is a difference in the corresponding local fused data. This algorithm processing logic achieves efficient comparison of massive hash values ​​through layer-by-layer aggregation of the tree structure, avoiding the inefficiency of calculating each one individually, solving the problem of low verification efficiency in traditional techniques, and accurately locating the discrepancies, providing a clear direction for subsequent corrections.

[0089] Step S103: Introduce a differential privacy framework to the corrected fusion state to generate perturbation parameters, and randomize the data points in the initial fusion result using the perturbation parameters to generate a privacy-enhanced dataset.

[0090] In one specific embodiment, step S103 may specifically include the following steps:

[0091] The modified fusion state is analyzed using a differential privacy framework to determine the sensitivity parameters for privacy protection.

[0092] The perturbation parameter is generated based on the sensitivity parameter, and the updated batch data is randomized using the perturbation parameter to generate a dataset with added noise.

[0093] Verify whether the dataset with added noise meets the privacy constraints, and generate the privacy-enhanced dataset.

[0094] Specifically, in distributed data processing scenarios, a differential privacy framework is used to analyze the corrected fusion state to determine the sensitivity parameters for privacy protection. The corrected fusion state includes batch data integrated from each node after consistency verification, such as user transaction amount data in the financial field or patient diagnostic indicator data in the medical field. The differential privacy framework determines the sensitivity parameters by analyzing the impact of modifications to individual data points on the overall data statistics. Specifically, it calculates the maximum output difference between any two adjacent datasets (datasets differing by only one data point) in the corrected fusion state under the same query function; this maximum value is the sensitivity parameter. Taking financial transaction data as an example, if the query function is "calculate the average transaction amount within a certain time period", and adjacent datasets respectively include and exclude a user's single transaction of 1000 yuan, and the difference between the two query results is 0.5 yuan (assuming a total data volume of 2000 records), then the sensitivity parameter... The value is set to 0.5. In the background, most solutions do not incorporate a differential privacy framework to dynamically generate perturbation parameters, resulting in a mismatch between the strength of privacy protection and the characteristics of the data. However, by analyzing the corrected fusion state through a differential privacy framework to determine the sensitivity parameters, the privacy protection strategy can be adapted to the actual characteristics of the data, solving the problem of a lack of specificity in privacy protection.

[0095] Perturbation parameters are generated based on sensitivity parameters. The generation of perturbation parameters employs a Laplace mechanism, according to the formula... Calculate the scale parameter b of the Laplace distribution (where b is the scale parameter of the Laplace distribution). For sensitivity parameters, Privacy budgets are typically set based on business needs, such as privacy budgets in financial scenarios. (It can be set to 1.0), and this scale parameter b is the core perturbation parameter. The sensitivity parameter... =0.5, Privacy Budget Taking a value of 1.0 as an example, the scale parameter b is calculated to be 0.5. After generating the perturbation parameter, the updated batch data is randomized using this parameter. For each data point in the updated batch data, a noise value is randomly selected from the Laplace distribution that conforms to the scale parameter b. This noise value is then superimposed on the original data point value to generate the dataset with added noise. For example, if a user's transaction amount in the updated batch data is 5000 yuan, a noise value of 0.3 is selected from the Laplace distribution with a scale parameter b=0.5, and after superposition, a noise data of 5000.3 yuan is obtained. All data points are processed in this way to form the dataset with added noise. In traditional technologies, some solutions have fixed noise distributions that cannot adapt to data processing needs. However, by dynamically generating perturbation parameters based on sensitivity parameters and performing randomization, the intensity of noise addition can be correlated with data privacy sensitivity. This avoids excessive noise leading to decreased data availability or insufficient noise leading to privacy leaks, thus solving the problem of imbalance between privacy protection and data availability.

[0096] To verify whether the dataset with added noise meets the privacy constraints, we employ the differential privacy definition. The privacy model is validated by calculating the ratio of the output probability of the noisy dataset to that of any adjacent original dataset, and determining whether this ratio does not exceed [a certain threshold]. At the same time, ensure that the δ value (δ represents the probability of an extremely low probability event occurring) is lower than a preset threshold (e.g., ).by ε For example, if the output probabilities of adjacent datasets after adding noise are P1 and P2 respectively, then the following condition must be met: and If the validation passes, it indicates that the dataset with added noise maintains the overall statistical properties of the data while protecting individual privacy; this dataset is thus considered privacy-enhanced. If the validation fails, the privacy budget needs to be readjusted. ε (such as) ε (Adjusted from 1.0 to 0.8), the perturbation parameter generation and randomization steps are repeated until the privacy constraints are met. Traditional techniques lack a verification step for privacy protection effectiveness, which can easily lead to substandard privacy protection. However, verification through privacy constraints can ensure that the generated privacy-enhanced dataset meets the preset privacy protection standards, solving the problem of uncontrollable privacy protection effectiveness.

[0097] Step S104: Adjust the noise distribution according to the type of set operation performed on the privacy-enhanced dataset, and adjust the privacy-enhanced dataset according to the noise distribution to generate a perturbation dataset adapted for the operation.

[0098] In one specific embodiment, step S104 may specifically include the following steps:

[0099] Determine the type of set operation to be performed on the privacy-enhanced dataset;

[0100] If the set operation is of the type of intersection operation, then the noise reduction algorithm is used to reduce the noise amplitude and generate the first noise distribution;

[0101] If the set operation is a non-intersection operation, then a noise enhancement algorithm is used to increase the noise amplitude and generate a second noise distribution;

[0102] The privacy-enhanced dataset is adjusted according to the first noise distribution or the second noise distribution to generate the perturbation dataset for adaptation computation.

[0103] Specifically, determining the type of set operation to perform on a privacy-enhanced dataset requires first analyzing the operational requirements of the dataset within the business scenario. Privacy-enhanced datasets are noisy datasets generated after processing with a differential privacy framework, commonly found in areas such as financial transaction data aggregation and medical patient information statistics. Taking the financial sector as an example, if the business requirement is "extract user data with transfer records from both Bank A and Bank B," the operation instruction corresponds to the intersection operation in set operations. If the business requirement is "aggregate all user data with transfer records from either Bank A or Bank B," the operation instruction corresponds to the union operation (such as union or difference) or non-intersection operation in set operations. During the analysis process, by identifying logical keywords in the operation instructions (such as "with both" corresponding to intersection, "or" corresponding to union), and combining this with the business attributes of the dataset, the specific type of set operation can be determined. In traditional technologies, most solutions have a fixed noise distribution, making it impossible to adapt and adjust according to the type of set operation, resulting in a difficulty in balancing privacy protection and operational requirements. Clearly defining the type of set operation can provide a targeted direction for subsequent noise distribution adjustments, solving the problem of the disconnect between noise distribution and the operational scenario.

[0104] If the set operation is an intersection operation, a noise reduction algorithm is used to reduce the noise amplitude and generate the first noise distribution. Intersection operations need to preserve common data features between different datasets; excessive noise can distort these common features, affecting the accuracy of the operation results. The noise reduction algorithm first extracts the current noise amplitude of each data point in the privacy-enhanced dataset. This amplitude is determined by the perturbation parameters in the previous differential privacy framework, such as the initial noise amplitude of financial user transfer amount data. Then, based on the overlap coefficient corresponding to the intersection operation... α ( α The value ranges from 0.4 to 0.6. Overlap refers to the proportion of common data points between different datasets; the higher the proportion, the better. α (The larger the value), the greater the noise amplitude will be when calculating the reduced noise level. Suppose that in a certain financial dataset, the overlap between user data from Bank A and Bank B is 70%, corresponding to... α=0.5, then the reduced noise amplitude Based on the reduced noise amplitude The first noise distribution is generated using a Laplace distribution, and the scale parameter of this distribution is set to... This ensures that the noise distribution meets the data accuracy requirements of intersection operations. Traditional techniques, in intersection operations, can easily lead to ambiguity of common data features due to fixed noise. Generating a first noise distribution through a noise reduction algorithm can reduce the interference of noise on the accuracy of intersection operation results while protecting privacy, thus resolving the contradiction between privacy protection and result accuracy in intersection operations.

[0105] If the set operation is a non-intersection operation, a noise enhancement algorithm can be used to increase the noise amplitude and generate a second noise distribution. Non-intersection operations (such as union and difference) easily expose unique data features in the dataset. If the privacy protection of these unique data points is insufficient, it can easily lead to information leakage risks. The noise enhancement algorithm first obtains the current noise amplitude of the data points in the privacy-enhancing dataset. Then, based on the risk coefficient corresponding to the non-intersection operation... β ( β The value ranges from 1.2 to 1.8. The risk coefficient refers to the probability that a unique data point will be leaked; the higher the probability, the greater the risk. β (The larger the value), the greater the noise amplitude will be. Taking medical patient treatment data as an example, in the union operation of patient data from a certain hospital, unique patients (patients who only exist in this hospital) account for 60%, corresponding to... =1.5, if the initial noise amplitude =0.4, then the enhanced noise amplitude Based on the enhanced noise amplitude A second noise distribution is generated using a Gaussian distribution, and the variance of this distribution is set to... This method enhances privacy protection for unique data points by increasing the noise amplitude. Traditional techniques, using fixed noise, struggle to address the risk of privacy leakage for unique data points in non-intersecting computation scenarios. By generating a second noise distribution through noise enhancement algorithms, the strength of privacy protection in non-intersecting computation scenarios can be improved, solving the problem of high privacy leakage risk for unique data points in non-intersecting computations.

[0106] The privacy-enhanced dataset is adjusted based on either the first or second noise distribution to generate a perturbed dataset adapted for the operation. For intersection operations, noise values ​​are re-extracted from the first noise distribution to replace the original noise values ​​in the privacy-enhanced dataset. For example, if a financial user's transfer amount in the privacy-enhanced dataset is 5000.4 yuan (original amount 5000 yuan + initial noise 0.4 yuan), a new noise value of 0.2 is extracted from the first noise distribution (scale parameter 0.3), resulting in an adjusted data of 5000.2 yuan. For non-intersection operations, noise values ​​are re-extracted from the second noise distribution to replace the original noise values. For example, if a medical patient's body temperature data in the privacy-enhanced dataset is 37.2℃ (original body temperature 37℃ + initial noise 0.2℃), a new noise value of 0.2 is extracted from the second noise distribution (variance). A new noise value of 0.5 is extracted from the data, and the adjusted data is 37.5℃. After noise replacement is completed for all data points, a perturbed dataset adapted to the current set operation type is formed. This adjustment process ensures that the noise distribution is deeply matched with the set operation type, solving the problem in the background technology where the noise distribution is fixed and cannot adapt to different operation requirements, and ensuring that subsequent set operations achieve a balance between privacy protection and result usability.

[0107] Step S105: After processing the perturbation dataset of the adaptation operation with a noise adjustment algorithm, an equalized perturbation level is generated.

[0108] In one specific embodiment, step S105 may specifically include the following steps:

[0109] The noise distribution of the disturbance dataset is analyzed using a noise adjustment algorithm to determine the dynamic adjustment parameters of the noise amplitude.

[0110] The noise in the disturbance dataset is optimized based on the dynamic control parameters to generate a balanced noise distribution;

[0111] The disturbed dataset is resampled using the equalized noise distribution to generate an equalized dataset;

[0112] Verify whether the noise level of the equalized dataset meets the preset equalization conditions to obtain the equalized disturbance level.

[0113] Specifically, noise adjustment algorithms are used to analyze the noise distribution of a perturbation dataset used in adaptation operations to determine dynamic adjustment parameters for noise amplitude. This is commonly used in fields such as financial transaction data processing and medical diagnosis information statistics. Examples include user transfer amount perturbation datasets adapted for intersection operations in financial scenarios and patient vital sign perturbation datasets adapted for union operations in medical scenarios. The noise adjustment algorithm first extracts the noise values ​​of all data points in the perturbation dataset and then calculates the statistical characteristics of the noise distribution through statistical analysis, including the noise mean, noise variance, and extreme noise amplitudes. Taking a financial user transfer amount perturbation dataset as an example, assuming the dataset contains 1000 user transfer records with corresponding noise values ​​of 0.2, 0.3, 0.5, and 0.4, the calculated noise mean is 0.35, the noise variance is 0.012, the maximum noise amplitude is 0.6, and the minimum is 0.1. Based on these statistical characteristics and combined with preset noise equalization targets (such as noise mean fluctuation range ±0.05 and noise variance threshold 0.015), dynamic control parameters are determined using the proportional coefficient method. These dynamic control parameters include the noise amplitude amplification factor. With reduction factor ,in Used to adjust the noise amplitude below the mean towards the mean (when the minimum value < the noise mean). =Noise mean / minimum value), Used to adjust noise amplitudes above the mean towards the mean (when the maximum value > the noise mean). =Maximum value / Noise mean). For the above financial dataset, the calculation is... ; Traditional techniques lack dynamic optimization mechanisms for noise distribution, resulting in uneven noise distribution and excessive or insufficient privacy protection for some data points. By analyzing noise distribution and determining dynamic adjustment parameters through noise adjustment algorithms, noise amplitude can be optimized in a targeted manner, thus solving the problem of unbalanced noise distribution.

[0114] The noise in the perturbation dataset is optimized based on dynamic control parameters to generate a balanced noise distribution. For each data point in the perturbation dataset, if the noise value is less than the noise mean, the noise value is multiplied by an amplification factor. Adjustments are made; if the noise value is greater than the noise mean, divide the noise value by the reduction factor. Adjustments are made; if the noise value is within ±0.05 of the noise mean, it remains unchanged. Taking the financial user transfer amount perturbation dataset as an example, if the noise value of a data point is 0.1 (less than the noise mean of 0.35), the adjusted noise value is 0.1 × 3.5 = 0.35; if the noise value of another data point is 0.6 (greater than the noise mean of 0.35), the adjusted noise value is 0.6 / 1.71 ≈ 0.35; and if the noise value of another data point is 0.32 (within the noise mean of 0.35 ± 0.05), it remains unchanged at 0.32. All adjusted noise values ​​are collected, and a balanced noise distribution is generated by fitting the kernel density estimation method. The mean of this distribution is controlled within the preset balanced target range, and the variance is below the threshold of 0.015. Traditional techniques often have fixed and unoptimized noise distributions, failing to adapt to the actual noise requirements of data points. By dynamically adjusting parameters to optimize noise and generate a balanced noise distribution, the noise distribution can be made more uniform, solving the problem of uneven privacy protection caused by excessive noise amplitude differences.

[0115] A balanced dataset is generated by resampling the disturbed dataset to achieve a uniform noise distribution. Monte Carlo sampling is used to re-extract noise values ​​from the uniform noise distribution for each original data point in the disturbed dataset. These newly extracted noise values ​​are then superimposed on the original data point values, replacing the noise data in the original disturbed dataset. For example, in financial user transfer data, an original data point with an amount of 5000 yuan and an original noise value of 0.6 is re-extracted with a noise value of 0.35 from the uniform noise distribution, resulting in a new data point of 5000.35 yuan. Similarly, another original data point with an amount of 8000 yuan and an original noise value of 0.1 is re-extracted with a noise value of 0.35, resulting in a new data point of 8000.35 yuan. After resampling and noise superposition of all original data points, a balanced dataset is formed. This process ensures that the noise value of each data point comes from the uniform noise distribution, avoiding the problem of local noise anomalies in the original disturbed dataset, solving the problem of poor noise adaptability of data points, and improving the overall noise balance of the dataset.

[0116] Verify whether the noise level of the equalized dataset meets the preset equalization conditions to obtain the equalized disturbance level. The preset equalization conditions include the noise mean fluctuation range (e.g., ±0.05), the noise variance threshold (e.g., 0.015), and the noise amplitude extreme difference threshold (e.g., 0.5). Calculate the noise mean, noise variance, and noise amplitude extreme difference (maximum value minus minimum value) of the equalized dataset. If the noise mean of the equalized dataset is within the preset mean range, the noise variance is below the variance threshold, and the noise amplitude extreme difference is below the extreme difference threshold, then the preset equalization conditions are met, and the noise level of the equalized dataset is the equalized disturbance level. If not, readjust the dynamic control parameters (e.g., adjust the adjustment parameters). , The calculation coefficients are repeated, and the noise optimization and resampling steps are repeated until the preset equalization conditions are met. Taking a financial scenario as an example, the noise mean of the equalized dataset is 0.34 (within ±0.05), the noise variance is 0.01 (below the threshold of 0.015), and the extreme difference of noise amplitude is 0.4 (below the threshold of 0.5). Therefore, the preset equalization conditions are met, and this noise level is determined as the equalized disturbance level. Traditional techniques lack a verification step for noise levels, making it impossible to ensure the noise equalization effect. Verification through preset equalization conditions ensures that the equalized disturbance level meets business requirements, solves the problem of uncontrollable noise equalization effects, and provides a stable noise foundation for subsequent fusion processing in a trusted execution environment.

[0117] Step S106: Perform fusion processing on the equalized perturbation level in a trusted execution environment, and conduct privacy inference risk assessment on the fusion result to obtain secure computational output.

[0118] In one specific embodiment, step S106 may specifically include the following steps:

[0119] The equalized perturbation levels are fused using a trusted execution environment to generate a fusion calculation result.

[0120] The privacy inference risk value is determined by analyzing the fusion operation results using a privacy inference risk assessment algorithm.

[0121] If the privacy inference risk value is lower than the preset fusion threshold, the fusion operation result is output as a safe operation output.

[0122] Specifically, a trusted execution environment (TEA) is used to fuse the balanced perturbation levels to generate a fusion result. This process is common in datasets containing balanced noise, such as financial transaction data aggregation and medical patient information statistics. Examples include user transfer amount datasets with balanced noise in financial scenarios and patient vital sign index datasets with balanced noise in medical scenarios. The TEA uses hardware-level isolation technology (such as Intel SGX) to build an independent computing space. After importing the dataset corresponding to the balanced perturbation levels into this environment, a suitable fusion algorithm is selected based on business requirements. For numerical data (such as transaction amounts and body temperature indicators), a weighted average algorithm is used for fusion, with weights determined based on the data contribution of each distributed node (data contribution is calculated as the proportion of data processed by each node; for example, if node A processes 30% of the total data, its weight is 0.3). For categorical data (such as transaction types and disease diagnosis results), a voting algorithm is used for fusion, counting the frequency of each category and selecting the category with the highest frequency as the fusion result. Taking financial user transfer data as an example, assuming there are three distributed nodes, node 1 accounts for 30% of the data (weight 0.3), node 2 accounts for 40% (weight 0.4), and node 3 accounts for 30% (weight 0.3). The average transfer amounts corresponding to the balanced disturbance levels of each node are 5000.35 yuan, 5000.32 yuan, and 5000.36 yuan, respectively. The weighted average algorithm calculates the fusion result as: 5000.35 × 0.3 + 5000.32 × 0.4 + 5000.36 × 0.3 ≈ 5000.34 yuan. Traditional techniques...

[0123] The lack of a trusted execution environment for fusion processing makes it easy for data to be tampered with or leaked during the fusion process. By using a trusted execution environment for isolation protection and adapting the fusion algorithm, the security of the fusion process and the accuracy of the results can be ensured, thus solving the problem of data security not being guaranteed during the fusion process.

[0124] A privacy inference risk assessment algorithm is used to analyze the fusion operation results to determine the privacy inference risk value. Based on the differential privacy principle, the algorithm first extracts data features from the fusion operation results, including the data distribution value H (entropy is used to measure data uncertainty; the lower the entropy value, the easier the data is to infer) and the correlation degree R between the data and the original sensitive data (correlation degree is determined by calculating the cosine similarity between the fusion result and known sensitive data). Taking the fusion result of a medical patient's body temperature as an example, the fusion operation result is 37.3℃. The entropy value H of the dataset containing this data is extracted as 2.8 (entropy value ranges from 0-5, the lower the value, the higher the privacy risk). The correlation degree R between this value and the patient's original body temperature of 37℃ is calculated as 0.92 (correlation degree ranges from 0-1, the higher the value, the higher the privacy risk). Then, the formula is used... Calculate the privacy inference risk value, where The entropy value is used to convert the risk coefficient R into a correlation coefficient; the product of the two is the risk value. Substituting the data, we get: (1-2.8 / 5)×0.92=0.4048 (the risk value ranges from 0 to 1, with higher values ​​indicating a higher risk of privacy leakage). Traditional techniques do not perform privacy inference risk assessment on the fusion results, which can easily lead to privacy leakage. By quantifying the risk value through a privacy inference risk assessment algorithm, the privacy security level of the fusion results can be clearly defined, solving the problem of the inability to quantify privacy risks.

[0125] If the privacy inference risk value is lower than the preset fusion threshold, the fusion result is output as a safe output. The preset fusion threshold is set according to the privacy requirements of the business scenario. For financial scenarios, which have higher privacy security requirements, the threshold can be set to 0.3; for medical scenarios involving sensitive health information, the threshold can be set to 0.25; and for IoT device data scenarios, the threshold can be appropriately relaxed to 0.4. Taking a financial scenario as an example, with a preset fusion threshold of 0.3, if the privacy inference risk value of the aforementioned financial transfer amount fusion result is calculated to be 0.28 (lower than 0.3), then the fusion result is deemed to meet the privacy security requirements and is directly output as a safe output. If the risk value of a medical patient's heart rate fusion result is 0.26, while the preset threshold for medical scenarios is 0.25, the risk value is higher than the threshold. Therefore, it is necessary to return to step S103 to readjust the perturbation parameters of the differential privacy framework (e.g., reducing the privacy budget). ε (To enhance noise), the noise generation, distribution adjustment, equalization optimization, and fusion processing steps are repeated until the privacy inference risk value is lower than a preset threshold. Traditional technologies lack a privacy risk threshold judgment mechanism and cannot determine whether the fusion result is safe. By comparing the preset fusion threshold with the risk value, a fusion result that meets privacy and security requirements can be selected, solving the problem of no judgment standard for privacy leakage risk and ensuring that the output calculation result has both usability and security.

[0126] Step S107: Feed the secure computation output back to the blockchain distributed node, update the status of each node through the hash ring binding mechanism, and generate a fusion result for consistency maintenance after synchronization through the inter-node consistency protocol.

[0127] In one specific embodiment, step S107 may specifically include the following steps:

[0128] The secure computation output is fed back to the blockchain distributed node, and a new mapping relationship is determined through a hash ring binding mechanism;

[0129] The local state of each node is updated according to the new mapping relationship, and an updated set of node states is generated.

[0130] The updated set of node states is synchronized using an inter-node consistency protocol to determine the hash value of the global state.

[0131] The states of each node are integrated based on the hash value of the global state to generate a fusion result that maintains consistency.

[0132] Specifically, secure computational outputs are fed back to the distributed nodes of the blockchain, and a new mapping relationship is determined through a hash ring binding mechanism. This is commonly used in fields such as financial transaction data aggregation and medical patient information statistics. For example, in financial scenarios, secure fusion of user transfer amounts is used; in medical scenarios, secure fusion of patient vital signs is used. After receiving the secure computational output, the blockchain distributed nodes first use the hash ring binding mechanism to calculate the hash value of key fields in the secure computational output (such as user ID in financial scenarios and patient number in medical scenarios) using the SHA-256 algorithm. Simultaneously, each blockchain distributed node is mapped to multiple virtual nodes, and their hash values ​​are calculated. All hash values ​​are distributed across multiple virtual nodes. On a ring-shaped space (hash ring), traversing clockwise along the hash ring, the physical node associated with the first virtual node pointed to by the position corresponding to the hash value of the security operation output is determined as the belonging node of the security operation output, thus establishing a new mapping relationship. Taking a financial scenario as an example, the security operation output is "User ID: 1001, Transfer Amount: 5000.34 yuan". The hash value of User ID is calculated as 0x12345678. The first virtual node clockwise on the hash ring at this position corresponds to physical node A. Therefore, the new mapping relationship is "the belonging node of the security operation output of User 1001 is A". Traditional technologies lack a synchronization mechanism combining hash ring binding and consistency protocols after node state updates, making it difficult to achieve global state unification. By determining the new mapping relationship through the hash ring binding mechanism, the distribution rules of the security operation output among nodes can be clearly defined, solving the problem of node state chaos caused by disordered data distribution.

[0133] The local states of each node are updated according to the new mapping relationship, generating an updated set of node states. Each distributed blockchain node receives the security operation output according to the new mapping relationship and integrates it with its original local state data. If the data corresponding to the security operation output has no historical record in the local node, it is directly added to the local state; if there is a historical record, the original data is replaced or supplemented according to business rules (such as the accumulation of transaction amount in financial scenarios or the update of vital signs in medical scenarios). Taking financial node A as an example, its original local state includes "User 1001, total historical transfer amount: 20,000 yuan". After receiving the new secure operation output "User 1001, current transfer amount: 5,000.34 yuan", the local state is updated according to the rules to "User 1001, total historical transfer amount: 25,000.34 yuan". Similarly, medical node B's original local state includes "Patient 2001, historical temperature record: 37.0℃". After receiving the new secure operation output "Patient 2001, current temperature: 37.3℃", the local state is updated to "Patient 2001, historical temperature record: 37.0℃, 37.3℃". After all nodes complete their local state updates, the updated state data of each node is collected to form an updated node state set. Traditional technology lacks a unified basis for updating node local states, easily leading to inconsistencies in data between nodes. Updating local states according to the new mapping relationship ensures the relevance and accuracy of data updates for each node, solving the problem of disordered local state updates.

[0134] The updated node state set is synchronized through an inter-node consensus protocol to determine the hash value of the global state. The inter-node consensus protocol adopts the Raft protocol. This protocol first determines a leader node through an election mechanism. The leader node extracts the state data of each node from the updated node state set, calculates the SHA-256 hash value of the state data of each node, and then concatenates the state hash values ​​of all nodes to calculate the root hash value. This root hash value is the hash value of the global state. Taking a blockchain network with 3 nodes as an example, the updated state hash value of node 1 is H1=0xabc123, that of node 2 is H2=0xdef456, and that of node 3 is H3=0x789ghi. The leader node concatenates H1, H2, and H3 to obtain the string "0xabc1230xdef4560x789ghi", calculates the SHA-256 hash value of this string as H=0x1a2b3c, and this is the hash value of the global state. The leader node broadcasts the hash value of the global state to all follower nodes. Each follower node verifies whether its locally calculated global state hash value matches the one broadcast by the leader node. If they match, synchronization is complete; otherwise, the follower node requests the complete updated set of node states from the leader node and recalculates the hash value until they match. Traditional technologies lack inter-node state synchronization mechanisms, making it difficult to achieve global state unification. By synchronizing the node state set and determining the global state hash value through the Raft protocol, it is ensured that all nodes have a consistent understanding of the global state, thus solving the problem of missing global state synchronization.

[0135] The system integrates the states of each node based on the hash value of the global state, generating a consistent fusion result. Using the hash value of the global state as a verification benchmark, it compares the hash values ​​of the local states of each node with the corresponding hash values ​​of the nodes in the global state hash value. If a match is found, the local state data of that node is extracted; otherwise, the node is triggered to resynchronize and update its state data. All matching local state data are aggregated according to business dimensions (e.g., user dimension in financial scenarios, patient dimension in medical scenarios). In financial scenarios, it aggregates the transfer records of the same user across nodes, generating fusion results such as total transfer amount and transfer frequency; in medical scenarios, it aggregates the vital signs indicators of the same patient across nodes, generating a complete record of the patient's medical data. Taking a financial scenario as an example, user 1001 has a transfer record of 25,000.34 yuan at node 1, no record at node 2, and a transfer record of 8,000 yuan at node 3. After aggregation, a fusion result of "User 1001, total transfer amount: 33,000.34 yuan, transfer nodes: node 1, node 3" is generated. In a medical scenario, patient 2001's body temperature record at node 2 is 37.3℃, and a blood pressure record of 120 / 80 mmHg is generated at node 1. After aggregation, a fusion result of "Patient 2001, body temperature 37.3℃, blood pressure: 120 / 80 mmHg" is generated. These aggregated results are the fusion results for consistency maintenance. Traditional technologies lack verification criteria for global state integration, which can easily lead to inconsistent fusion results. Integrating node states based on global state hash values ​​can ensure the data accuracy and consistency of the integration process, solving the problem of insufficient consistency in global fusion results.

[0136] Please see Figure 2 This specifically illustrates the distribution of virtual nodes and load distribution within the hash ring. Figure 2 (a) shows the distribution of virtual nodes in the hash ring, with different colors representing different numbers of virtual node clusters. Hash values ​​range from 0 to approximately The relatively dispersed distribution of virtual nodes indicates that the hash ring binding mechanism, through virtual node mapping, can distribute data more evenly across the hash space, preventing excessive concentration of data in certain hash intervals and improving the randomness and balance of data allocation. (See also...) Figure 2 (b) and Figure 2 (d) The node with node ID 6 is allocated a significantly larger amount of data than other nodes (nodes 1-5 have very little data, while node 6 has nearly 2500). This reflects a clear imbalance in data distribution under the current hash ring binding mechanism. Most of the data is concentrated on node 6, while other nodes receive very little data. This may be because the virtual node associated with node 6 on the hash ring has a larger hash interval, making it the main bearer of data distribution. Figure 2(c) shows the load distribution corresponding to the virtual nodes, with the load ranging from 0 to approximately 2500. Combined with the color scale, the darker areas correspond to the virtual nodes with higher loads, and the distribution of these high-load virtual nodes on the hash ring corresponds to the node cluster positions in the virtual node distribution subgraph. This indicates that the load borne by different virtual nodes on the hash ring is related to their distribution on the ring, and it also shows that some virtual nodes have higher loads, indicating a certain degree of load difference.

[0137] Please see Figure 3 This demonstrates the results of the distributed data trusted fusion processing process. Figure 3 (a) shows the changes in data values ​​for different node IDs before and after noise adaptation. As can be seen from the figure, the data values ​​of each node fluctuated significantly before adaptation, and the differences between different nodes were obvious. After adaptation, the data values ​​were more stable overall, and the fluctuation range of the data values ​​of each node decreased. This indicates that the noise adaptation process effectively adjusted the noise situation of the data, making the data more stable after adaptation. Figure 3 (b) shows a comparison before and after noise equalization. Before equalization, the data values ​​of each node deviated significantly from the global mean, and the differences between nodes were significant. After equalization, the data values ​​were closer to the global mean, and the consistency of data between nodes was improved. This indicates that noise equalization can make the data better fit the global average level. Figure 3 (c) shows the equalization front variance and equalization back variance. The height corresponding to the equalization front variance is significantly higher than that of the equalization back variance. The reduction in variance indicates that after processing, the dispersion of the data is reduced, and the data distribution is more concentrated, further demonstrating the optimization effect of noise equalization and other operations on data consistency. Figure 3 (d) shows the data fusion results under a trusted execution environment. The original node data had some fluctuations, while the fusion result was relatively stable, and the fusion result was correlated with the trend of the node data. It is an integration of the data from each node, which shows that the data from each node can be effectively fused under a trusted execution environment. Figure 3 (e) shows the results of the risk assessment, with a risk value of 0.0849, which is in the "safe" range. This indicates that after the preceding processes such as noise adaptation, equalization, and trusted execution environment integration, the risk of data privacy breaches is at a low level. Figure 3 (f) is a system performance radar chart, which shows the system performance from multiple dimensions such as "number of nodes", "privacy protection", "consistency" and "data volume". The system can meet the requirements well in multiple performance dimensions, which reflects the comprehensive performance of the distributed data fusion method in terms of node management, privacy protection, data consistency and data volume, and can comprehensively meet the needs of distributed data processing.

[0138] The above describes the data trust processing method integrating trusted computing and blockchain in the embodiments of this application. Please refer to [link / reference]. Figure 4 The following describes the data trust processing system 400 integrating trusted computing and blockchain in the embodiments of this application, including:

[0139] The data acquisition and fusion module 401 is used to acquire batch data through blockchain distributed nodes, update the batch data using a hash ring binding mechanism, and generate an initial fusion result.

[0140] The verification and correction module 402 is used to perform consistency verification on the initial fusion result and obtain the corrected fusion state.

[0141] The privacy enhancement module 403 is used to introduce a differential privacy framework into the corrected fusion state, generate perturbation parameters, and randomize the updated batch data using the perturbation parameters to generate a privacy-enhanced dataset.

[0142] The noise adaptation module 404 is used to adjust the noise distribution according to the privacy-enhanced dataset, adjust the privacy-enhanced dataset according to the noise distribution, and generate a perturbation dataset for adaptation operation.

[0143] The perturbation equalization module 405 is used to process the perturbation dataset of the adaptation operation through a noise adjustment algorithm to generate an equalized perturbation level.

[0144] The trusted output module 406 is used to perform fusion processing on the equalized perturbation level in a trusted execution environment, and to perform privacy inference risk assessment on the fusion result to obtain secure computational output.

[0145] The synchronization maintenance module 407 is used to feed back the secure computation output to the blockchain distributed nodes, update the status of each node through the hash ring binding mechanism, and generate a fusion result for consistency maintenance after synchronization through the inter-node consistency protocol.

[0146] Through the collaborative efforts of the aforementioned components, the system constructs a comprehensive blockchain data trust processing system encompassing "data acquisition, fusion verification, privacy enhancement, noise adaptation, disturbance equalization, trusted output, and state synchronization." This system achieves end-to-end secure processing from batch data acquisition across distributed nodes to the generation of globally consistent fusion results, ensuring trustworthiness throughout the entire data processing lifecycle.

[0147] The data acquisition and fusion module 401 collects multi-source batch data through distributed blockchain nodes. Relying on a hash ring binding mechanism, it achieves precise data allocation and gradual updates. The generated local fusion data is synchronized via inter-node communication protocols and integrated into an initial fusion result. The verification and correction module 402 receives the initial fusion result, extracts the hash value set, and compares the consistency between batches using a consistency verification algorithm. If a match is found, the initial result is retained; otherwise, the hash ring parameters are dynamically adjusted to redistribute data and update the local fusion data. The privacy enhancement module 403 takes the corrected fusion state as input, introduces a differential privacy framework to analyze data privacy sensitivity, generates suitable perturbation parameters to randomize data points, and generates a privacy-enhanced dataset after privacy constraint verification. The noise adaptation module 404 dynamically adjusts the noise distribution according to the type of set operation. Intersection operations reduce noise amplitude, while non-intersection operations amplify noise amplitude, generating a perturbation dataset adapted for the operation. The perturbation equalization module 405 optimizes the noise distribution in the perturbation dataset using a noise adjustment algorithm to generate an equal perturbation level. The trusted output module 406 performs fusion processing on the level of equilibrium perturbation in the trusted execution environment, and selects fusion results with risk values ​​below a threshold as secure computation outputs using a privacy inference risk assessment algorithm. The synchronization maintenance module 407 feeds the secure output back to the blockchain nodes, updates the local state of the nodes through a hash ring binding mechanism, determines the global state hash value after synchronization via the inter-node consistency protocol, integrates and generates a fusion result for consistency maintenance, achieves global node state unification, and ultimately completes the trusted processing of distributed data from collection to application.

[0148] Reference Figure 5 This application also provides a data trusted processing device 500 that integrates trusted computing and blockchain. This data trusted processing device can be a server, and its internal structure can be as follows: Figure 5 As shown. This data trusted processing device integrating trusted computing and blockchain includes a processor 502, a memory 503, a display screen 504, an input device 505, a network interface 506, and a database 507 connected via a system bus 501. The processor 502 provides computing and control capabilities. The memory 503 includes a non-volatile storage medium 5031 and internal memory 5032. The non-volatile storage medium 5031 stores the operating system and computer programs. The internal memory 5032 provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database 507 stores the data corresponding to this embodiment. The network interface 506 communicates with external terminals via a network connection. The computer program, executed by the processor, can implement the above-described method.

[0149] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A data trust processing method integrating trusted computing and blockchain, characterized in that, The method includes: Step S101: Collect batch data through blockchain distributed nodes, update the batch data using a hash ring binding mechanism, and generate initial fusion results; Step S102: Perform a consistency check on the initial fusion result to obtain the corrected fusion state; Step S103: Introduce a differential privacy framework to the corrected fusion state to generate perturbation parameters, and randomize the data points in the initial fusion result using the perturbation parameters to generate a privacy-enhanced dataset; Step S104: Adjust the noise distribution according to the type of set operation performed on the privacy-enhanced dataset, and adjust the privacy-enhanced dataset according to the noise distribution to generate a perturbation dataset adapted for the operation; Step S105: After processing the perturbation dataset of the adaptation operation with a noise adjustment algorithm, an equalized perturbation level is generated; Step S106: Perform fusion processing on the equalized perturbation level in a trusted execution environment, and conduct privacy inference risk assessment on the fusion result to obtain secure computational output; Step S107: Feed the secure computation output back to the blockchain distributed node, update the state of each node through the hash ring binding mechanism, and generate a fusion result for consistency maintenance after synchronization through the inter-node consistency protocol; Step S104 includes: Determine the type of set operation to be performed on the privacy-enhanced dataset; If the set operation is of the type of intersection operation, then the noise reduction algorithm is used to reduce the noise amplitude and generate the first noise distribution; If the set operation is a non-intersection operation, then a noise enhancement algorithm is used to increase the noise amplitude and generate a second noise distribution; The privacy-enhanced dataset is adjusted according to the first noise distribution or the second noise distribution to generate the perturbation dataset for adaptation computation.

2. The data trust processing method integrating trusted computing and blockchain according to claim 1, characterized in that, Step S101 includes: Batch data is collected through blockchain distributed nodes; A hash ring binding mechanism is used to allocate the batch data and determine the data mapping relationship of each node; The batch data is progressively updated according to the data mapping relationship to generate local fused data for each node; The locally fused data is synchronized using an inter-node communication protocol to obtain a set of hash values; The initial fusion result is generated by integrating the local fusion data of each node based on the hash value set.

3. The data trust processing method integrating trusted computing and blockchain according to claim 2, characterized in that, Step S102 includes: Obtain the set of hash values ​​from the initial fusion result; The hash value set is compared using a consistency check algorithm to determine whether the hash values ​​between batches match. If a match is found, the initial fusion result is retained as the corrected fusion state. If there is a mismatch, the hash ring binding parameters are adjusted to redetermine the data mapping relationship, and the local fusion data is updated to generate the corrected fusion state.

4. The data trust processing method integrating trusted computing and blockchain according to claim 1, characterized in that, Step S103 includes: The modified fusion state is analyzed using a differential privacy framework to determine the sensitivity parameters for privacy protection. The perturbation parameter is generated based on the sensitivity parameter, and the updated batch data is randomized using the perturbation parameter to generate a dataset with added noise. Verify whether the dataset with added noise meets the privacy constraints, and generate the privacy-enhanced dataset.

5. The data trust processing method integrating trusted computing and blockchain according to claim 1, characterized in that, Step S105 includes: The noise distribution of the disturbance dataset is analyzed using a noise adjustment algorithm to determine the dynamic adjustment parameters of the noise amplitude. The noise in the disturbance dataset is optimized based on the dynamic control parameters to generate a balanced noise distribution; The disturbed dataset is resampled using the equalized noise distribution to generate an equalized dataset; Verify whether the noise level of the equalized dataset meets the preset equalization conditions to obtain the equalized perturbation level.

6. The data trust processing method integrating trusted computing and blockchain according to claim 1, characterized in that, Step S106 includes: The equalized perturbation levels are fused using a trusted execution environment to generate a fusion calculation result. The privacy inference risk value is determined by analyzing the fusion operation results using a privacy inference risk assessment algorithm. If the privacy inference risk value is lower than the preset fusion threshold, the fusion operation result is output as a safe operation output.

7. A data trust processing method integrating trusted computing and blockchain according to claim 1, characterized in that, Step S107 includes: The secure computation output is fed back to the blockchain distributed node, and a new mapping relationship is determined through a hash ring binding mechanism. The local state of each node is updated according to the new mapping relationship, and an updated set of node states is generated. The updated set of node states is synchronized using an inter-node consistency protocol to determine the hash value of the global state. The states of each node are integrated based on the hash value of the global state to generate a fusion result that maintains consistency.

8. A data trust processing system integrating trusted computing and blockchain, characterized in that, For implementing the data trust processing method integrating trusted computing and blockchain as described in any one of claims 1 to 7, the data trust processing system integrating trusted computing and blockchain comprises: The data acquisition and fusion module is used to collect batch data through blockchain distributed nodes, update the batch data using a hash ring binding mechanism, and generate an initial fusion result. The verification and correction module is used to perform consistency verification on the initial fusion result and obtain the corrected fusion state. The privacy enhancement module is used to introduce a differential privacy framework into the corrected fusion state, generate perturbation parameters, and randomize the updated batch data using the perturbation parameters to generate a privacy-enhanced dataset. The noise adaptation module is used to adjust the noise distribution according to the privacy-enhanced dataset, and adjust the privacy-enhanced dataset according to the noise distribution to generate a perturbation dataset for adaptation operation. The perturbation equalization module is used to process the perturbation dataset of the adaptation operation through a noise adjustment algorithm to generate an equalized perturbation level. The trusted output module is used to perform fusion processing on the equalized perturbation level in a trusted execution environment, and to perform privacy inference risk assessment on the fusion result to obtain secure computational output. The synchronization maintenance module is used to feed back the secure computation output to the blockchain distributed nodes, update the status of each node through the hash ring binding mechanism, and generate a fusion result for consistency maintenance after synchronization through the inter-node consistency protocol.

9. A data trust processing device integrating trusted computing and blockchain, characterized in that, The method includes a memory and a processor, wherein the memory stores a computer program that can run on the processor, and the processor executes the computer program to implement the data trust processing method integrating trusted computing and blockchain as described in any one of claims 1 to 7.

Images