Supply chain performance data blockchain storage method based on multi-party secure computation

Abstract

The application relates to the technical field of supply chain data storage, and discloses a supply chain performance data blockchain storage method based on multi-party secure calculation. The method comprises the following steps: monitoring the performance data state and relative interaction relationship of each participant in a supply chain in real time, and judging whether the data concentration degree in a target supply chain region exceeds a safety threshold; when the threshold is exceeded, the interaction influence of the data of each participant is analyzed through a multi-party secure calculation protocol, and the privacy leakage risk caused by data sharing is evaluated; the spatiotemporal variation of historical performance data is processed by adopting distributed feature learning combined with time series analysis, and the potential interference degree of historical data fluctuation on the integrity of current performance data is evaluated; and based on the potential interference degree of the privacy leakage risk and the historical data fluctuation, it is judged whether overall blockchain storage optimization of the target supply chain region is needed. The method can be used to respond to the problems of supply chain data concentration and privacy protection, and realize storage optimization in combination with the spatiotemporal characteristics of historical data.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of supply chain data notarization, in particular to a supply chain performance data blockchain notarization method based on multi-party secure computation. BACKGROUND

[0002] Supply chain operation involves multiple parties, including suppliers, manufacturers, distributors and retailers, etc. A large amount of data is generated in the performance process of each participant, such as order information, logistics status, inventory level, etc. These data are not only the basis for business decision-making of each participant, but also the basis for the overall coordination of the supply chain.

[0003] With the acceleration of digitalization, supply chain data presents the characteristics of large scale, frequent flow and complex interaction. The data format and storage method of different participants are different, and the data is prone to information asymmetry and inconsistency when transmitted and shared across subjects. At the same time, the data of each link of the supply chain has certain sensitivity, containing business secrets and core business information, and the data sharing process faces the risk of privacy leakage.

[0004] Traditional data notarization methods rely on centralized agencies, and a single subject is responsible for data storage and management. In this mode, the integrity of the data depends on the credibility of the centralized agency, and once the agency has security vulnerabilities or operational errors, it may lead to data tampering and loss. In addition, centralized notarization is difficult to achieve data rights division under the participation of multiple parties, and each participant lacks effective protection of data control and access rights, which may lead to data ownership disputes.

[0005] In the supply chain performance scenario, the timeliness and relevance of data are particularly important. The fluctuation of historical data may have an impact on the current performance status, for example, past inventory fluctuations may affect the delivery cycle of current orders, and traditional notarization methods are difficult to effectively capture this time-space correlation and provide reliable historical data reference for dynamic adjustment of the supply chain. At the same time, when the data intensity in a specific region exceeds a certain limit, the complexity of data interaction will increase significantly, and the traditional notarization mechanism is difficult to cope with the challenges of privacy protection and data management brought by this. SUMMARY

[0006] The purpose of the present application is to provide a supply chain performance data blockchain notarization method based on multi-party secure computation to solve the problems raised in the background.

[0007] To achieve the above purpose, the present application provides a supply chain performance data blockchain notarization method based on multi-party secure computation, which comprises:

[0008] Real-time monitoring of the performance data state and relative interaction relationship of each participant in the supply chain, and judging whether the data intensity in the target supply chain region exceeds the safety threshold.

[0009] When the data-intensive degree in the target supply chain region exceeds the safety threshold, the interaction influence of the data of each participant is analyzed by constructing a multi-party secure computation protocol, and the privacy leakage risk caused by data sharing in the target supply chain region is evaluated;

[0010] The spatio-temporal variation of historical performance data in the target supply chain region is processed by a distributed feature learning method combined with time series analysis, and the potential interference degree of historical data fluctuation on the integrity of current performance data in the target supply chain region is evaluated;

[0011] Based on the privacy leakage risk caused by data sharing in the target supply chain region and the potential interference degree of historical data fluctuation on the integrity of current performance data in the target supply chain region, it is judged whether to perform overall blockchain notarization optimization of the target supply chain region.

[0012] Preferably, the method further comprises:

[0013] When the overall blockchain notarization optimization of the target supply chain region is needed, the performance data state of each participant and other participants in the target supply chain region is analyzed, and the data path complexity of each participant is evaluated;

[0014] Based on the data path complexity of each participant, the data notarization priority of each participant is determined;

[0015] Generate a notarization data block and submit it to the blockchain network for tamper-proof notarization;

[0016] Optimize the broadcast timing of the notarization data block by monitoring the notarization response time and the state of the blockchain network in real time.

[0017] Preferably, the performance data state and relative interaction relationship of each participant in the supply chain are monitored in real time to determine whether the data-intensive degree in the target supply chain region exceeds the safety threshold, specifically:

[0018] Real-time collection of performance data state information of all participants, including location information, transaction frequency, cooperation duration and data volume;

[0019] Determine whether the participant is located in the target supply chain region according to the performance data state information:

[0020] Compare the number of participants located in the target supply chain region with the logical capacity of the target supply chain region to determine the data-intensive degree in the target supply chain region;

[0021] Based on the comparison of the data-intensive degree in the target supply chain region and the corresponding safety threshold, the safety risk level of data interaction in the target supply chain region is determined.

[0022] Preferably, the interaction of each participant's data is analyzed by constructing a multi-party secure computing protocol, and the privacy leakage risk caused by data sharing in the target supply chain region is evaluated, specifically:

[0023] Real-time acquisition of shared data input of each participant in the target supply chain region, including data sensitivity, access permission and encryption level;

[0024] According to the performance data state information and shared data input of each participant, a data processing model based on secure multi-party computation is constructed to describe the privacy protection mechanism in the data interaction process;

[0025] Using the constructed data processing model, the data dependency relationship between participants and the disturbance influence between shared data inputs are calculated;

[0026] Using parallel computing technology to iteratively solve the data processing model, the coupling effect between participants and shared data in a high-density data interaction environment is simulated, and the privacy leakage risk caused by data sharing in the target supply chain region is evaluated; The privacy leakage risk caused by data sharing in the target supply chain region includes low risk and high risk.

[0027] Preferably, the spatiotemporal variation of historical performance data in the target supply chain region is processed by a distributed feature learning method combined with time series analysis to evaluate the potential interference degree of historical data fluctuations on the integrity of current performance data in the target supply chain region, specifically:

[0028] Real-time acquisition of historical performance data set in the target supply chain region;

[0029] Using a distributed feature learning method, the historical performance data set is mapped to a feature space to reconstruct the dynamic characteristics and change patterns of the data;

[0030] Using a time series analysis method, the historical performance data set in the feature space is processed at different time scales to extract feature information of historical data fluctuations;

[0031] According to the feature information of historical data fluctuations, a historical data spatiotemporal variation model is established to reflect the dependence of data on time stamp and geographical location;

[0032] Using the established historical data spatiotemporal variation model, the potential interference degree of historical data fluctuations on the integrity of current performance data in the target supply chain region is evaluated; The potential interference degree of historical data fluctuations on the integrity of current performance data in the target supply chain region includes significant and negligible.

[0033] Preferably, based on the privacy leakage risk caused by data sharing in the target supply chain region and the potential interference degree of historical data fluctuation on the integrity of the current fulfillment data in the target supply chain region, it is judged whether to perform overall blockchain notarization optimization of the target supply chain region, specifically:

[0034] When the privacy leakage risk caused by data sharing in the target supply chain region is low risk, and the potential interference degree of historical data fluctuation on the integrity of the current fulfillment data in the target supply chain region is negligible, it is determined that the overall blockchain notarization optimization of the target supply chain region is not performed.

[0035] Otherwise, it is determined that the overall blockchain notarization optimization of the target supply chain region is performed, and a notarization priority evaluation mechanism is triggered.

[0036] Preferably, when the overall blockchain notarization optimization of the target supply chain region is needed, the fulfillment data state of each participant and other participants in the target supply chain region is analyzed, and the data path complexity of each participant is evaluated, specifically:

[0037] According to the fulfillment data state information, the data interaction distance and state difference between each participant and other participants are calculated;

[0038] By analyzing the data interaction distance and state difference between participants, possible data conflict and interference scenarios are identified: for two participants whose data interaction distance is less than a preset safety threshold, all participant pairs whose state difference is less than a preset state difference are marked out.

[0039] Based on the data conflict and interference scenarios, the data path complexity of each participant is evaluated.

[0040] Preferably, based on the data conflict and interference scenarios, the data path complexity of each participant is evaluated, specifically:

[0041] The path complexity index of each participant is calculated, which is obtained by aggregating the normalized ratio of the state difference and the data interaction distance of all related participant pairs;

[0042] The higher the path complexity index, the more complex the data path of the participant, and the priority of notarization optimization needs to be performed.

[0043] Preferably, based on the data path complexity of each participant, the data notarization priority of each participant is determined, specifically:

[0044] The notarization priority score is calculated, which is obtained by multiplying the path complexity index and the participant weight factor;

[0045] The greater the storage evidence priority score is, the higher the data storage priority of the participant is, the participants are sorted from high to low based on the storage evidence priority score, and the storage evidence data of the participants with high priority is processed preferentially.

[0046] Preferably, the storage evidence data block is generated and submitted to the blockchain network for tamper-proof storage, specifically:

[0047] According to the data of the participant with a high storage evidence priority score, a storage evidence data block containing a timestamp and a data digest is generated;

[0048] The storage evidence data block is encrypted using a hash function to generate a unique data identifier;

[0049] The encrypted storage evidence data block is broadcast to the blockchain network node, verified by the consensus mechanism and written into the blockchain.

[0050] Compared with the prior art, the beneficial effects of the present application are:

[0051] By monitoring the fulfillment data state and the relative interaction relationship of each participant in the supply chain in real time, it can be determined whether the data concentration degree in the target supply chain region exceeds the safety threshold, so that the subsequent processing process can be started in a targeted manner to avoid management confusion caused by excessive data concentration. When the data concentration degree exceeds the threshold, the interaction influence of the data of each participant can be analyzed by means of the multi-party secure computing protocol, the potential privacy leakage risk in the data sharing process can be deeply mined, and the occurrence of privacy risk can be reduced from the root level of data interaction.

[0052] The distributed feature learning method combined with time series analysis processes the spatio-temporal changes of historical fulfillment data, which can comprehensively capture the correlation between historical data fluctuations and current fulfillment data, clearly present the potential interference of historical data fluctuations on the integrity of current data, and make each participant more clearly understand the internal relationship between historical data and current state. Based on the potential interference degree of privacy leakage risk and historical data fluctuations, it is determined whether to perform overall blockchain storage optimization, which can realize dynamic adjustment of the storage strategy and make the blockchain storage more suitable for the actual data state of the supply chain.

[0053] Through blockchain storage optimization, the tamper-proof feature of the blockchain can be used to provide a stable storage carrier for supply chain fulfillment data, so that the data remains consistent and traceable in a multi-party environment. The introduction of the multi-party secure computing protocol ensures effective data interaction while avoiding direct exposure of sensitive information of each participant, so that data sharing can be carried out in an orderly manner under the premise of security. The combination of distributed feature learning and time series analysis makes the value of historical data fully utilized, and through the analysis of the influence of historical fluctuations on current data, each participant in the supply chain can more comprehensively understand the business logic behind the data. BRIEF DESCRIPTION OF DRAWINGS

[0054] Figure 1 The timing diagram of the supply chain performance data blockchain storage method based on multi-party secure calculation of the present application;

[0055] Figure 2 The flowchart for data-intensive degree and security risk level determination;

[0056] Figure 3 The flowchart for privacy leakage risk assessment;

[0057] Figure 4 The flowchart for historical data fluctuation interference assessment;

[0058] Figure 5 The flowchart for data path complexity assessment. DETAILED DESCRIPTION

[0059] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0060] Please refer to Figure 1 The present application provides a supply chain performance data blockchain storage method based on multi-party secure calculation, which comprises:

[0061] The real-time monitoring of the compliance data state and the relative interaction relationship of each participant in the supply chain. The compliance data state information includes location information, transaction frequency, cooperation time length and data volume. By analyzing these information, it is determined whether the participants are located within the target supply chain region. The number of participants in the target supply chain region is compared with the logical capacity of the region, and the data-intensive degree is calculated. The data-intensive degree is compared with the preset safety threshold to determine whether it exceeds the safety threshold. When the data-intensive degree exceeds the safety threshold, a multi-party secure computation protocol is constructed to analyze the interaction influence of the data of each participant. The protocol is based on shared data input, including data sensitivity, access permission and encryption level, to evaluate the privacy leakage risk caused by data sharing in the target supply chain region. The privacy leakage risk is divided into low risk and high risk. At the same time, the historical compliance data set in the target supply chain region is obtained, and a distributed feature learning method is used to map the data set to a feature space to reconstruct the dynamic characteristics and change patterns of the data. Combined with the time series analysis method, the data in the feature space is processed at different time scales to extract the feature information of the historical data fluctuation. Based on these feature information, a model is established to describe the spatio-temporal change of the historical data, reflecting the dependence of the data on time stamp and geographical location. The model is used to evaluate the potential interference degree of the historical data fluctuation on the integrity of the current compliance data, which is divided into significant and negligible states. Finally, based on the evaluation results of the privacy leakage risk and the potential interference degree, it is judged whether to perform the overall blockchain storage optimization of the target supply chain region. If the privacy leakage risk is low and the potential interference degree is negligible, no optimization is performed; otherwise, the optimization mechanism is triggered.

[0062] Embodiment 1: refer to Figure 2 The real-time monitoring of the compliance data state and the relative interaction relationship of each participant in the supply chain. The compliance data state information is continuously obtained through the data acquisition module deployed on the terminal of the participant, which is connected with the supply chain management system to capture parameters such as location coordinates, transaction frequency per unit time, cooperation cycle length and data throughput. The location information is converted into latitude and longitude coordinates by using the geocoding technology, the transaction frequency is counted by using the event log analysis engine, the cooperation time length is calculated by extracting the start and end time stamps from the contract database, and the data volume is recorded by using the transmission layer flow monitoring tool. The target supply chain region is determined by the geographic fence algorithm, and its boundary coordinates are stored in the region configuration library. According to the inclusion relationship between the location coordinates and the geographic fence, the participant set located in the target region is selected. The logical capacity of the target region is generated by the infrastructure resource evaluation model, which generates a capacity baseline value by integrating parameters such as the maximum number of concurrent connections of the regional server cluster, the redundant space of the storage hardware and the peak value of the network bandwidth.

[0063] The number of screened participants is compared with the logical capacity baseline value. The ratio of the number of participants to the capacity baseline is used as a quantitative indicator of data intensity, which is kept to two decimal places. The preset safety threshold is stored in the risk policy library, which is dynamically updated according to the frequency of abnormal events in the historical operation log, with an update period of twenty-four hours. When the intensity indicator exceeds the safety threshold, the system determines that there is an increase in the level of data interaction risk, triggering the subsequent analysis process. The level of data interaction risk is divided into three levels, and the level division standard is preset according to different threshold intervals.

[0064] When the risk level reaches the preset condition, a deep analysis process is started for each participant in the target area. The process calculates the data interaction distance between participants using a composite algorithm of spatial location and interaction frequency. The location distance is calculated based on the spherical distance formula, and the interaction frequency distance uses the Euclidean distance algorithm of transaction frequency, and the two are fused into a single distance value through a weighting coefficient. The state difference calculation module is also running, which analyzes the cooperation time difference coefficient and data volume difference rate between participants, where the difference coefficient uses the standard deviation algorithm, and the difference rate is calculated by the absolute difference percentage.

[0065] The preset safety distance threshold and state difference threshold are stored in the conflict rule library, and the threshold values are derived from industry standard data sets. The system automatically identifies all participant combinations with an interaction distance below the safety distance threshold, and further screens participant pairs with a state difference less than the difference threshold in the combination. The identified participant pairs are recorded in the conflict matrix table, and the row and column indices of the matrix correspond to the unique identifiers of the participants.

[0066] Based on the conflict matrix table, the path complexity indicator of each participant is generated. The indicator is obtained by traversing all records related to the target participant in the conflict matrix, and normalizing the state difference rate and interaction distance ratio of each associated participant pair. The normalization process uses the min-max scaling algorithm to convert the ratio to the interval of zero to one, and then takes the arithmetic mean of all associated values, with the calculation result kept to three decimal places. The path complexity indicator has a value range of zero to one, and the value size reflects the complexity level of the data path.

[0067] The evidence priority score calculation module generates a score value based on the product of the path complexity indicator and the weight factor. The weight factor is extracted from the participant attribute library, which maintains weight coefficients for different role types, and the coefficients are set with reference to the node centrality in the supply chain topology. The score calculation uses a linear multiplication formula, and the result is converted to a percentage and stored in the priority queue. The queue sorting engine sorts the participants in descending order of score, and the participants with high scores are marked for priority processing.

[0068] The data block generation component processes the participant data according to the priority queue order. This component adds a coordinated universal time timestamp when data is packaged, and generates a data digest through a feature extraction algorithm. The digest algorithm selects a fixed-length encoding scheme for key fields, preserving the uniqueness features of the data. The hash encryption engine performs a one-way conversion on the complete data block, using a standard encryption hash function to generate a fixed-length hexadecimal string as a data identifier, which has anti-collision properties.

[0069] The encrypted data block is broadcast to the blockchain node cluster through the peer-to-peer network protocol. The node cluster adopts a permissioned architecture, divided into two roles of data verification nodes and data storage nodes. The verification nodes perform a distributed consensus process, which guarantees data consistency through a state machine replication algorithm. The write operation links the data block to the end position of the distributed ledger in chronological order after being verified by more than a preset proportion of nodes.

[0070] The system continuously collects evidence response data during operation, including the time interval from broadcast initiation to blockchain confirmation, as well as network-level node online state proportion, data transmission rate, and other parameters. The response data input time sequence optimization controller adjusts the broadcast time interval of the data packet to adapt to the network state. The time interval is dynamically adjusted using an exponential backoff algorithm, and the interval period is automatically extended when the network delay exceeds the warning value. The priority queue manager synchronously optimizes the data packet distribution order, allocates priority transmission channels for high-score data blocks, and divides the bandwidth resources of the transmission channels according to the priority weight proportion.

[0071] Embodiment 2: Referring to Figure 3 , real-time acquisition of shared data input of each participant in the target supply chain region. Shared data is transmitted through a standardized interface, and the interface uses an encryption communication protocol to ensure transmission layer security. The data input contains three structured fields: the data sensitivity field uses a three-level classification coding system, corresponding to public level marked as P1, internal level marked as P2, and confidential level marked as P3; the access permission field records a role-based permission matrix, stored as a binary permission code; the encryption level field identifies the type of security algorithm currently applied, such as symmetric encryption identifier AES, homomorphic encryption identifier HE. After format verification, the input data is stored in a distributed cache area.

[0072] The data processing model based on secure multi-party computation adopts a hierarchical architecture design. The input layer is equipped with a data preprocessing module that performs normalization on shared data: converts text-based permission codes into numerical vectors, and maps discrete sensitivity codes into continuous weight factors. The calculation layer deploys privacy protection operation protocols, mainly including secret sharing sub-protocol and garbled circuit sub-protocol. The secret sharing sub-protocol divides the input data into random number fragments, and the number of fragments corresponds to the total number of participant nodes. The garbled circuit sub-protocol converts business logic into Boolean circuit representation, and realizes logic hiding through circuit gate encryption. The output layer is configured with a risk quantification module that generates a privacy leakage possibility index.

[0073] Data dependency is modeled through dynamic dependency graph. Dependency graph nodes represent participant entities, and node attributes include current data state snapshots. The directed edges between nodes record data transmission paths, and the edge weight values are calculated through the combination of interaction frequency and data volume. A full graph traversal algorithm is executed every ten minutes to update node connectivity status. The correlation coefficient matrix is generated based on the interaction history between entities, and the matrix element values are dynamically refreshed through the Pearson correlation algorithm. The disturbance of shared data input affects the implementation of sensitivity test: a small disturbance variation is implemented on the input field, with the variation amplitude controlled within one percent, and the output layer indicator fluctuation range is observed.

[0074] Parallel computing tasks are executed through a distributed computing framework. The framework master node receives complete data processing model parameter configuration and divides the computation graph into independent sub-tasks. The task scheduler dispatches task packages based on node resource idle state, each task package containing input data fragments and computation instruction set. The multi-threaded execution environment configures a flexible thread pool, and the thread pool size is automatically adjusted according to task complexity. The iterative solution process sets an initial privacy parameter baseline and outputs intermediate results after executing the computation core logic. The parameter adjustment engine compares the difference between the to-be-optimized parameters and the actual output, and generates a parameter fine-tuning vector. The iteration termination condition is monitored by the convergence determinator, and the computation stops when the fluctuation of the results of three consecutive iterations is less than five thousandths.

[0075] High-density data environment simulation adopts stress testing mode. The node coupling simulator generates load peak scenarios by constructing stress data sets through isometric amplification of actual interaction data. The coupling effect analysis module runs the covariance calculation program to calculate the statistical correlation between participant behavior patterns and data characteristics. The Monte Carlo component implements random sampling experiments: a random number generator is established to simulate 1000 data exchange events, and two participant entities are randomly selected for each event to perform hypothetical data interaction. The probability of privacy leakage is calculated based on the simulation results, and the probability value is rounded to two decimal places.

[0076] The privacy leakage risk assessment process includes two dimensions of probability classification and type judgment. The probability classification compares the calculation results with the preset interval standard: the probability value below a certain percentage is classified as low risk, and the probability value above a certain percentage is classified as high risk. The type judgment module cross analyzes the data characteristics and the leakage path to extract a set of core risk characteristics. The risk report generator outputs a structured document, and the document chapters include risk level labels, main leakage scene descriptions, and key influence factor lists. The report data is synchronized and pushed to the blockchain storage audit interface.

[0077] Embodiment 3: refer to Figure 4 The historical fulfillment data set in the target supply chain region is continuously acquired by a distributed data acquisition system deployed at the data warehouse interface layer of each node in the supply chain. The data set includes three types of core data: time series transaction records, logistics track coordinate sequences, and inventory change logs. The transaction record fields include transaction timestamp, participant identifier, commodity code, and transaction amount; the logistics track data records the spatial position sampling points of the transportation tool, each sampling point containing latitude and longitude coordinates and collection time; the inventory change log records the time, item code, and quantity change of warehouse in-out events. The data storage adopts a sharded architecture, and each partition is configured with an independent replica management strategy. The data retrieval service provides a time range query interface, supporting millisecond-level time precision filtering.

[0078] The distributed feature learning system adopts a double-layer network architecture. The bottom layer feature extraction network is composed of parallel working autoencoders. Each autoencoder instance is deployed on an independent computing node. The encoder part of the autoencoder includes five layers of fully connected neural networks, and the hidden layer activation function uses the rectified linear unit. The input layer receives the standardized historical data window, and the window length is configurable by default, which is twenty-four time units. The encoding process compresses the input data into a low-dimensional latent space, and the compression ratio is set to one-eighth of the input dimension. The decoder network is symmetric to the encoder structure, and the reconstruction error is minimized through the backpropagation algorithm. The feature space mapping process performs batch normalization operation to eliminate the dimension difference between different data sources.

[0079] The time series analysis module is configured with a multi-scale processing pipeline. The short-term analysis unit processes the hour-grained data and extracts local fluctuation patterns using the sliding window mechanism. The window sliding step is set to six time units, and the overlap ratio is fifty percent. The medium-term analysis unit operates on the day-grained data and applies the seasonal trend decomposition algorithm to separate the periodic components. The long-term analysis unit processes the month-grained data and fits the macro trend curve through the state space model. The multi-scale feature fusioner aggregates the analysis results of different granularities in a hierarchical manner, and the aggregation weights are dynamically calculated through the attention mechanism. The feature information extraction process applies spectral analysis method to identify the dominant frequency components in the data fluctuations.

[0080] The historical data spatio-temporal variation modeling adopts an extended state space representation. The model state variables contain time dimension components and space dimension components, the time components are processed by the temporal convolution network, and the space components are modeled by the graph neural network. The temporal convolution network is configured with dilated causal convolution kernels, and the number of convolution layers is set to four, with an exponential growth of the dilation factor in each layer. The node features of the graph neural network include the geographical location encoding and historical interaction features of the participants, and the edge weights reflect the spatial correlation strength. The model parameter training process adopts a phased optimization strategy, first fixing the spatial network parameters to train the temporal network, and then jointly fine-tuning all parameters. The training data is divided using the time sequence segmentation method, and the last three months of data are reserved as the validation set.

[0081] The spatio-temporal variation model evaluates the integrity of the current data by executing Monte Carlo sampling. The sampler generates simulated interference data that conforms to the historical statistical characteristics, and the interference strength is controlled within ten percent of the historical fluctuation range. The injection test mixes the simulated data into the real data stream in proportion, and the mixing ratio is gradually increased from five percent to twenty percent. The integrity detector compares the feature distribution difference between the original data and the disturbed data, and the difference measure adopts the improved Wasserstein distance D w :

[0082]

[0083] where K represents the total number of feature dimensions, F k represents the original cumulative distribution function of the kth feature, G k represents the cumulative distribution function of the disturbed data. The potential interference level classifier determines the interference level according to the distance threshold, and the threshold interval is calibrated by historical benchmark tests. The determination result is output as a discrete label, including negligible, slight, and significant.

[0084] The joint evaluation of privacy leakage risks and data interference levels implements matrix decision analysis. The row dimension of the decision matrix represents the privacy risk level, the column dimension represents the interference level, and the matrix unit stores the pre-defined optimization strategy code. The evaluation engine queries the matrix to obtain the decision rule, and triggers the optimization flag when any dimension reaches the pre-set alert level. The optimization trigger signal activates the evidence priority evaluation process and sends a resource pre-allocation request to the blockchain network. The system maintains a real-time state board that visualizes the risk indicators and optimization status of each participant. The board data update frequency is synchronized with the underlying monitoring module to ensure the timeliness of the decision basis.

[0085] Example 4: see Figure 5In the data evidence optimization process within the target supply chain region, the assessment of the complexity of the data path of the participants needs to be based on detailed analysis of actual interaction data. Suppose a certain electronic product supply chain contains six core participants: chip supplier (P1), screen manufacturer (P2), assembly plant (P3), logistics service provider (P4), wholesaler (P5), and retailer (P6). The interaction data of these participants in the past quarter is recorded in the interaction relationship table.

[0086] Table 1: Participant Interaction Data Record Table

[0087] Party pair Interaction distance (km) Daily interaction frequency Data volume difference rate (%) Cooperation duration difference (month) P1-P2 150 28 12.5 6 P1-P3 80 45 8.2 3 P2-P3 220 32 15.7 9 P3-P4 50 68 5.3 2 P4-P5 180 25 18.9 12 P5-P6 30 55 3.1 1

[0088] The system first sets the safety distance threshold to 100 kilometers, and the state difference threshold includes the data volume difference rate of 10% and the cooperation duration difference of 6 months. According to these thresholds, the participant pairs that need to be focused on are automatically identified. For example, the P1-P3 pair of participants, although the interaction distance of 80 kilometers is lower than the safety threshold, the data volume difference rate of 8.2% and the cooperation duration difference of 3 months do not exceed the corresponding thresholds, so it is marked as a potential conflict pair. Similarly, the P3-P4 and P5-P6 pairs are also marked because the interaction distance is lower than the threshold.

[0089] In the path complexity assessment stage, the system calculates the ratio of the state difference of the associated participant pair to the interaction distance for each participant. Taking the assembly plant P3 as an example, it has interactions with the chip supplier P1, the screen manufacturer P2, and the logistics service provider P4. For the P3-P1 pair, the state difference comprehensive value is calculated by the weighted average of the data volume difference rate and the cooperation duration difference, and then divided by the interaction distance of 80 kilometers to get the ratio. Similarly, the ratios of P3-P2 and P3-P4 are calculated, and finally the three ratios are normalized and averaged to get the path complexity index of P3.

[0090] In the evidence priority determination link, the system assigns a weight factor to the participants based on their roles in the supply chain. For example, the assembly plant P3 as a core node gets a weight of 1.2, while the logistics service provider P4 as an auxiliary node gets a weight of 0.9. After multiplying the path complexity index by the weight factor, the evidence priority scores of the six participants are as follows: P3 (0.87), P1 (0.72), P5 (0.68), P6 (0.65), P2 (0.61), P4 (0.58). According to this ranking, the data of the assembly plant P3 will get the highest priority for evidence processing.

[0091] When a data block is generated, the system extracts key features for each participant and generates a data summary. For example, P3's summary includes the average transaction amount of the last ten transactions, a list of major partners, and the current inventory turnover rate. These summary information, along with a timestamp accurate to the millisecond, are packaged into a data block, which is encrypted to form a fixed-length hash value. The hash algorithm uses a collision-resistant design to ensure that different participants' data blocks, even if similar in content, will generate completely different identifiers.

[0092] After receiving these data blocks, the verification nodes in the blockchain network first check the validity of the hash values, and then confirm the writing order of the data blocks through a consensus algorithm. The network uses a practical Byzantine fault tolerance mechanism, which requires more than two-thirds of the verification nodes to agree on the order of the data blocks before performing a write operation. After a successful write, the data block is permanently recorded in the distributed ledger, forming an unalterable record of evidence.

[0093] Throughout the process, the system continuously monitors the execution efficiency of the record-keeping operation. For high-priority participants such as P3, the system dynamically adjusts the broadcast frequency of their data blocks, prioritizing their transmission bandwidth during network congestion. Meanwhile, the time delay of each data block from generation to successful chaining is recorded, and these delay data are used for subsequent network optimization analysis. When the record-keeping delay of a participant exceeds the preset alert value, the system automatically adjusts its data transmission path or retry strategy.

[0094] Through this analysis method based on actual interaction data, the system can accurately identify key nodes and potential conflict points in the supply chain network, thereby targetedly optimizing the blockchain record-keeping process. The entire process does not rely on any preset performance indicators or optimization targets, but dynamically adjusts the record-keeping strategy according to the actual interaction characteristics of participants. This data-driven implementation ensures that the record-keeping optimization scheme is highly consistent with the actual operation of the supply chain.

[0095] After the data record-keeping priority determination process of the target supply chain region is started, the system accesses the path complexity indicator database, which stores the numerical evaluation results generated by the distributed computing nodes in real time. The participant weight factor is stored in a separate attribute configuration library, which maintains a coefficient mapping table for different role types. Core manufacturing participants are assigned higher coefficients, while auxiliary service participants are assigned lower coefficients. The system iterates through all participant identifiers in the region, and for each identifier, performs a product operation: extracts the current path complexity indicator floating-point value from the indicator database, and obtains the corresponding weight coefficient value from the attribute library. The two values are processed by a floating-point multiplication processor to generate an original priority score. The original score is linearly mapped to the zero to one hundred interval by a percentage converter, and the two decimal precision is preserved during the conversion process. The calculation result is written to the priority sorting buffer area.

[0096] The participant priority sequence generation module runs a stable sorting algorithm. The sorting engine reads the percentile scores of all participants from the cache area, and uses a heap sorting data structure to build a maximum heap. Only the position references of participant identifiers are exchanged during the heap sorting process to avoid large data copy operations. The sorting result forms a priority queue data structure, and the head of the queue always points to the participant with the highest score. The queue manager maintains a dynamic index mapping table to ensure incremental updating of the queue when new participants join or their status changes. The data notarization scheduler sequentially extracts participant identifiers from the head of the queue, and automatically advances the queue pointer to the next element after each extraction.

[0097] The notarization data block generation process includes two parallel processes: timestamp embedding and data digest generation. The timestamp generator calls the Precision Time Protocol service to obtain the coordinated universal time millisecond-level timing signal. The timestamp value is formatted as a string structure that complies with the ISO8601 standard, including time zone offset identification. The data digest engine is started simultaneously, and according to the participant identifier currently being processed, extracts state information within a specific time window from the performance data warehouse. The digest algorithm selects the average transaction amount, the number of active partners, and the inventory turnover rate as the three core indicators, and compresses them into a fixed byte length binary sequence through a fixed length coding scheme. The timestamp string and the digest byte stream are encapsulated into a structured data container.

[0098] The hash encryption processor uses a multi-stage processing mechanism. The preprocessor serializes the structured container into a byte array and fills it to the standard block length. The initial hash value is generated by the hardware security module as a cryptographically secure random number. The encryption core executes the SHA-3512 algorithm for sixty-four rounds of permutation operations on the data block. Each round of operation includes θ, ρ, π, χ, ι five kinds of nonlinear transformation operations, and the χ function introduces confusion characteristics. The final output hash value is a five hundred and twelve bit binary sequence, encoded as a one hundred and twenty eight character hexadecimal string as a unique identifier. The encryption process is completely completed in a trusted execution environment, and the memory data is automatically erased after operation.

[0099] The blockchain network access layer establishes a persistent connection pool. The broadcast manager obtains the list of validation nodes from the connection pool and transmits data packets through the publish-subscribe mode. The data packet format complies with the standardized transaction structure: including a fifty-six byte identifier header, a variable length payload data area, and a sixty-four byte digital signature area. The signature generation uses the participant's registered elliptic curve private key to perform ECDSA signature operation based on the secp256k1 curve parameters. Network transmission uses message queue middleware to ensure transmission reliability, and each message is set with a time-to-live threshold.

[0100] The consensus verification process is executed asynchronously within the group of validating nodes. A node receives a zone synchronization receives a broadcast packet, performs a format compliance check. The signature verification engine checks the validity of the signature using the sender's public certificate. The state consistency checker compares the latest state hash of the local ledger copy. Inter-node coordination employs a two-phase broadcast protocol: the proposal phase collects initial votes, and the commit phase performs final confirmation. Data blocks that pass verification are assigned an incremental block height and integrated into the global state tree by the Merkle tree builder. The ledger update module appends the new block to the chain structure and updates the world state database.

[0101] The notarization response monitoring system deploys distributed tracking probes. The timestamp collector sets five collection points on the critical path: the start of data block generation, the completion of encryption, the sending of broadcast, the first response reception, and the on-chain confirmation. The delay calculator automatically generates a report of the time difference between each stage. The network state sensor periodically collects twelve performance indicators of the validating node's central processing unit utilization, memory usage, network input and output throughput, etc. The monitoring data is written in real time to the time series analysis database.

[0102] The broadcast timing optimization controller implements a closed-loop feedback mechanism. The delay analyzer calculates the average end-to-end delay of different priority data blocks and establishes a delay variation trend model. The parameter adjuster dynamically adjusts the broadcast interval base value according to the current network performance indicators: shortens the interval during network idle period and extends the interval during congestion period. The priority weighter assigns a transmission acceleration coefficient to high-priority data blocks, which is exponentially related to the priority score. The concurrency controller limits the broadcast concurrency within a unit of time and triggers traffic shaping when the bandwidth utilization threshold is exceeded. The retry strategy manager implements a step backoff algorithm and sets a differentiated upper limit on the number of retries for failed data packets according to priority. All control parameters are visualized through the console interface, and administrators can manually adjust the baseline parameters.

[0103] It should be noted that, in the present document, the terms such as first and second, etc., are used only to distinguish one entity or operation from another, and do not necessarily require or imply these entities or operations to be in any such actual relationship or order. Moreover, the terms "comprising", "including" or any other variant thereof are intended to cover non-exclusive inclusion, so that processes, methods, articles or equipment including a series of elements not only include those elements, but also include other elements not explicitly listed or inherent to such processes, methods, articles or equipment.

[0104] While embodiments of the application have been shown and described, it is to be understood that the embodiments described are merely exemplary of the principles and application of the present application. Numerous modifications and adaptions can be effected without departing from the spirit and scope of the present application, which is not limited to the exact construction and arrangement described. It is intended, therefore, to cover all modifications and adaptions that fall within the scope of the claims and their equivalents.

Claims

1. A supply chain performance data blockchain storage method based on multi-party secure calculation, characterized in that, The method comprises the following steps: Real-time monitoring of the performance data status and relative interaction of each participant in the supply chain, judging whether the data-intensive degree in the target supply chain region exceeds the safety threshold, specifically including: Real-time collection of performance data status information of all participants, including location information, transaction frequency, cooperation duration and data volume; Determine whether the participants are located in the target supply chain region according to the performance data status information: Compare the number of participants located in the target supply chain region with the logical capacity of the target supply chain region to determine the data-intensive degree in the target supply chain region; Based on the comparison of the data-intensive degree in the target supply chain region and the corresponding safety threshold, the safety risk level of data interaction in the target supply chain region is judged; When the data-intensive degree in the target supply chain region exceeds the safety threshold, the interaction influence of each participant data is analyzed by constructing a multi-party secure computing protocol to evaluate the privacy leakage risk caused by data sharing in the target supply chain region, specifically including: Real-time acquisition of shared data input of each participant in the target supply chain region, including data sensitivity, access permission and encryption level; According to the performance data status information and shared data input of each participant, a data processing model based on secure multi-party computation is constructed to describe the privacy protection mechanism in the data interaction process; Using the constructed data processing model, the data dependency relationship between participants and the disturbance influence between shared data inputs are calculated; Using parallel computing technology to iteratively solve the data processing model, simulating the coupling effect between participants and shared data in a high-density data interaction environment, and evaluating the privacy leakage risk caused by data sharing in the target supply chain region; The privacy leakage risk caused by data sharing in the target supply chain region includes low risk and high risk; The data processing model based on secure multi-party computation adopts a hierarchical architecture design, the input layer is equipped with a data preprocessing module to perform normalization processing on shared data, convert text type permission code to numerical vector, and map discrete type sensitivity code to continuous type weight factor; The calculation layer deploys a privacy protection operation protocol, including a secret sharing sub-protocol and a garbled circuit sub-protocol; The output layer is configured with a risk quantification module to generate a privacy leakage possibility index; The secret sharing sub-protocol divides the input data into random number fragments, the number of fragments corresponds to the total number of participant nodes, and the garbled circuit sub-protocol converts business logic into Boolean circuit representation, and realizes logic hiding through circuit gate encryption; The data dependency relationship is modeled by a dynamic dependency graph, the nodes of the dependency graph represent participant entities, the node attributes include current data state snapshot, the directed edges between nodes record data transmission paths, and the weight values of the edges are calculated by combining interaction frequency and data volume; The full graph traversal algorithm is executed every ten minutes to update the node connectivity state, and the correlation coefficient matrix is generated based on the interaction history between entities, and the matrix element value is dynamically refreshed by Pearson correlation algorithm; The disturbance of the shared data input affects the implementation sensitivity test, including implementing a slight disturbance variation to the input field, and the variation amplitude is controlled within one percent, and observing the fluctuation range of the output layer index; Through the distributed feature learning method combined with time series analysis, the spatio-temporal variation of the historical performance data in the target supply chain region is processed, and the potential interference degree of the historical data fluctuation on the integrity of the current performance data in the target supply chain region is evaluated, specifically including: Real-time acquisition of historical performance data set in the target supply chain region; Using the distributed feature learning method, the historical performance data set is mapped to the feature space to reconstruct the dynamic characteristics and change patterns of the data; Using the time series analysis method, the historical performance data set in the feature space is processed at different time scales to extract the feature information of the historical data fluctuation; According to the feature information of the historical data fluctuation, a spatio-temporal variation model of the historical data is established to reflect the dependence of the data on the timestamp and geographical location; Using the established spatio-temporal variation model of the historical data, the potential interference degree of the historical data fluctuation on the integrity of the current performance data in the target supply chain region is evaluated; the potential interference degree of the historical data fluctuation on the integrity of the current performance data in the target supply chain region includes significant and negligible; The spatio-temporal variation modeling of the historical data uses an extended state space representation method, the model state variables include time dimension components and space dimension components, the time component is processed by a temporal convolution network, and the space component is modeled by a graph neural network; the temporal convolution network is configured with a dilated causal convolution kernel, and the convolution layer number is set to four, and the expansion factor of each layer increases exponentially; the node features of the graph neural network include the geographical location encoding and historical interaction features of the participants, and the edge weight reflects the spatial correlation strength; the model parameter training process adopts a phased optimization strategy, first fix the spatial network parameters to train the temporal network, and then jointly fine-tune all parameters; the training data is divided using the time sequence segmentation method, and the last three months of data are reserved as the validation set; The spatiotemporal change model evaluates the current data integrity by performing Monte Carlo sampling, the sampler generates simulated interference data conforming to historical statistical characteristics, the interference intensity is controlled within ten percent of the historical fluctuation range, the injection test mixes the simulated data into the real data stream in proportion, the mixing proportion gradually increases from five percent to twenty percent, and the integrity detector compares the feature distribution difference between the original data and the disturbed data, and the difference measure adopts the improved Wasserstein distance : wherein: denotes the total number of feature dimensions, denotes the original cumulative distribution function of the th feature, denotes the cumulative distribution function of the disturbed data; the potential interference degree classifier determines the interference level according to the distance threshold, the threshold interval is calibrated by the historical benchmark test, and the determination result is output as a discrete label, including negligible, slight and significant three levels; Based on the potential interference degree of the historical data fluctuation on the integrity of the current performance data in the target supply chain region and the potential privacy leakage risk caused by data sharing in the target supply chain region, it is judged whether to perform overall blockchain storage optimization of the target supply chain region, specifically including: When the potential privacy leakage risk caused by data sharing in the target supply chain region is low risk, and the potential interference degree of the historical data fluctuation on the integrity of the current performance data in the target supply chain region is negligible, it is determined that the overall blockchain storage optimization of the target supply chain region is not performed; Otherwise, it is determined to perform overall blockchain storage optimization of the target supply chain region, and trigger the storage priority evaluation mechanism.

2. The supply chain performance data blockchain notarization method based on multi-party secure computation according to claim 1, characterized in that, Also including: When the overall blockchain storage optimization of the target supply chain region is needed, the performance data state of each participant and other participants in the target supply chain region is analyzed, and the data path complexity of each participant is evaluated; Based on the data path complexity of each participant, the data storage priority of each participant is determined; Generate storage data blocks and submit them to the blockchain network for tamper-proof storage; By monitoring the storage certificate response time and the blockchain network state in real time, the broadcast timing of the storage certificate data block is optimized.

3. The supply chain performance data blockchain notarization method based on multi-party secure computation according to claim 2, characterized in that, When the overall blockchain storage optimization of the target supply chain area is needed, the performance data state of each participant in the target supply chain area and other participants is analyzed, and the data path complexity of each participant is evaluated, specifically as follows: According to the performance data state information, the data interaction distance and state difference between each participant and other participants are calculated. By analyzing the data interaction distance and state difference between participants, possible data conflict and interference scenarios are identified: for two participants whose data interaction distance is less than a preset safety threshold, all participant pairs whose state difference is less than a preset state difference are marked out. Based on the data conflict and interference scenarios, the data path complexity of each participant is evaluated.

4. The supply chain performance data blockchain notarization method based on multi-party secure computation according to claim 3, characterized in that, Based on the data conflict and interference scenarios, the data path complexity of each participant is evaluated, specifically as follows: The path complexity index of each participant is calculated, which is obtained by aggregating the normalized ratio of the state difference and the data interaction distance of all relevant participant pairs. The higher the path complexity index, the more complex the data path of the participant, and the priority of storage optimization needs to be given.

5. The supply chain performance data blockchain notarization method based on multi-party secure computation according to claim 2, characterized in that, Based on the data path complexity of each participant, the data storage priority of each participant is determined, specifically as follows: The storage priority score is calculated, which is obtained by multiplying the path complexity index and the participant weight factor. The larger the storage priority score, the higher the data storage priority of the participant. Based on the storage priority score from high to low, the participants are sorted, and the storage data of the participants with high ranking is processed preferentially.

6. The supply chain performance data blockchain notarization method based on multi-party secure computation according to claim 2, characterized in that, The storage data block is generated, and the storage data block is submitted to the blockchain network for tamper-proof storage, specifically as follows: According to the data of the participant with high storage priority score, the storage data block containing the timestamp and the data digest is generated. The storage data block is encrypted using a hash function to generate a unique data identifier. The encrypted storage data block is broadcast to the blockchain network node, and the consensus mechanism is used to verify and write to the blockchain.

Images