Supply Chain Data Security Management Method and System

Abstract

The present invention provides a supply chain data security management method and system. The method includes: hierarchically dividing and distributing and storing supply chain data according to production links to generate process-related hash values; constructing a feature extractor containing process constraint relationships based on the supply chain data and the process-related hash values, and using domain adversarial feature extraction to obtain supply chain data features that meet physical constraints; performing multi-level differential perturbation according to the supply chain data features and the process parameter association map, and correcting the perturbed data based on the process constraint relationships; dynamically performing on-chain adversarial verification on the model access behavior according to the perturbed process parameter data and the process-related hash values, and adjusting the differential perturbation intensity and access restriction when detecting abnormal access. The present invention prevents competitors from obtaining trade secret information through model reverse inference and ensures the security and effectiveness of data sharing by introducing process constraints and multi-level differential perturbation in the transfer learning process.

Description

Technical Field

[0001] The present invention relates to the technical field of data security management, and in particular to a supply chain data security management method and system. Background Art

[0002] In the current supply chain collaborative manufacturing environment, upstream and downstream enterprises usually adopt an intelligent prediction system based on transfer learning for production planning and quality management. A typical system architecture includes a data acquisition layer, a feature extraction layer, a model training layer, and a prediction output layer. Among them, the data acquisition layer is responsible for collecting various production data, such as equipment operation parameters, product quality data, material consumption data, etc.; the feature extraction layer performs dimensionality reduction and feature engineering on the original data; the model training layer trains and optimizes the model based on the transfer learning algorithm; the prediction output layer is responsible for generating various prediction results and decision-making suggestions.

[0003] However, during the process of model migration, since the production data of different enterprises often contains trade secrets such as core process parameters and formula data, although the existing management methods can ensure the security of data transmission and storage, they cannot effectively prevent competitors from extracting sensitive information in the original training data from the optimized model through model reverse inference. This implicit data leakage may lead to the loss of the enterprise's core competitiveness. Summary of the Invention

[0004] The main purpose of the present invention is to solve the technical problem that in the existing supply chain collaborative manufacturing environment, in the process of model migration of an intelligent prediction system based on transfer learning, it is easy for competitors to obtain core process parameters through reverse inference;

[0005] The first aspect of the present invention provides a supply chain data security management method, and the supply chain data security management method includes:

[0006] Dividing and distributing the storage of supply chain data according to the supply chain production links, and generating process-related hash values based on the process constraint information and the distributed supply chain data;

[0007] Constructing a feature extractor containing process constraint relationships based on the supply chain data and the process-related hash values, and performing domain adversarial feature extraction on the supply chain data based on the feature extractor to obtain supply chain data features that meet physical constraints;

[0008] Performing multi-level differential perturbation on the supply chain data according to the supply chain data features and a preset process parameter association map, and correcting the perturbed data according to the process constraint relationship to obtain perturbed process parameter data;

[0009] Perform dynamic on-chain adversarial verification on the model access behavior based on the perturbed process parameter data and the process-related hash value, and perform differential perturbation intensity adjustment and access restriction on the supply chain data stored distributively when detecting abnormal model access.

[0010] Optionally, in the first implementation manner of the first aspect of the present invention, the hierarchical division and distributed storage processing of the supply chain data according to the supply chain production links, and the generation of the process-related hash value based on the process constraint information and the distributively stored supply chain data include:

[0011] Classify the supply chain data according to the production links to obtain raw material link data, manufacturing link data, and product link data;

[0012] Use IPFS distributed storage to store the classified supply chain data in chunks to obtain data chunks and data chunk indexes;

[0013] Construct a Merkle tree for the data chunks based on the process constraint information to obtain a Merkle tree with process constraints;

[0014] Generate a process-related hash value according to the Merkle tree with process constraints and the process constraint information.

[0015] Optionally, in the second implementation manner of the first aspect of the present invention, the construction of a feature extractor containing process constraint relationships based on the supply chain data and the process-related hash value, and the domain adversarial feature extraction of the supply chain data based on the feature extractor to obtain supply chain data features that meet physical constraints include:

[0016] Perform data domain marking processing on the process parameter data in the supply chain data to obtain data samples with source identifiers, and group the process-related hash values according to the source identifiers to obtain the feature constraint conditions for each data domain;

[0017] Perform source domain and target domain division processing on the data samples to obtain an initial domain division result, and perform consistency verification processing on the initial domain division result according to the feature constraint conditions to obtain a domain division strategy that meets process constraints;

[0018] Perform feature space mapping processing on the process parameter data according to the domain division strategy to obtain an initial feature representation, and perform gradient reversal layer embedding processing on the initial feature representation to obtain an adversarial feature extractor;

[0019] Perform process constraint injection processing on the adversarial feature extractor to obtain a feature extraction network with process constraints, and perform feature extraction processing on the supply chain data according to the feature extraction network to obtain supply chain data features that meet physical constraints.

[0020] Optionally, in the third implementation manner of the first aspect of the present invention, the process parameter data is subjected to feature space mapping processing according to the domain partitioning strategy to obtain an initial feature representation, and the initial feature representation is subjected to gradient reversal layer embedding processing to obtain an adversarial feature extractor, including:

[0021] Perform feature dimension analysis processing on the domain partitioning strategy to obtain a dimension weight vector, and perform weighted mapping processing on the process parameter data according to the dimension weight vector to obtain an initial feature space;

[0022] Perform non-linear transformation processing on the initial feature space to obtain a hidden layer feature representation, and perform principal component projection processing according to the hidden layer feature representation to obtain a dimensionality-reduced feature matrix;

[0023] Perform reverse gradient calculation processing on the dimensionality-reduced feature matrix to obtain a gradient reversal parameter, and perform backpropagation processing according to the gradient reversal parameter to obtain an adversarial gradient matrix;

[0024] Perform feature reconstruction processing on the adversarial gradient matrix to obtain an adversarial feature representation, and perform feature extraction network construction processing according to the adversarial feature representation to obtain an adversarial feature extractor.

[0025] Optionally, in the fourth implementation manner of the first aspect of the present invention, the supply chain data is subjected to multi-level differential perturbation according to the supply chain data characteristics and a preset process parameter association map, and the perturbed data is corrected according to the process constraint relationship to obtain perturbed process parameter data, including:

[0026] Perform process sensitivity quantification processing on the supply chain data characteristics to obtain a parameter sensitivity matrix, and perform correlation analysis processing according to the parameter sensitivity matrix and the process parameter association map to obtain a multi-level sensitive data set;

[0027] Perform differential privacy budget allocation processing on the multi-level sensitive data set to obtain perturbation budget values at each level, and perform parameter perturbation amount calculation processing on the perturbation budget values to obtain initial perturbation data;

[0028] Perform constraint verification processing on the initial perturbation data and the process constraint relationship to obtain a constraint violation data identifier, and perform constraint inspection processing on the initial perturbation data according to the constraint violation data identifier to obtain a constraint violation degree;

[0029] Perform process constraint correction processing on the initial perturbation data according to the constraint violation degree to obtain corrected perturbation data, and perform process parameter reconstruction processing on the corrected perturbation data to obtain perturbed process parameter data.

[0030] Optionally, in the fifth implementation manner of the first aspect of the present invention, the differential privacy budget allocation process for the multi-level sensitive data set to obtain the perturbation budget values at each level, and the parameter perturbation amount calculation process for the perturbation budget values to obtain the initial perturbation data includes:

[0031] Perform a privacy sensitivity calculation process on the multi-level sensitive data set to obtain a sensitivity score matrix, and perform a budget weight allocation process according to the sensitivity score matrix to obtain an initial budget allocation plan;

[0032] Perform a constraint optimization process on the initial budget allocation plan to obtain the optimized budget value, and perform a Laplace noise generation process according to the optimized budget value to obtain a noise parameter set;

[0033] Perform a step-by-step perturbation amount calculation process on the noise parameter set to obtain a perturbation amount matrix, and perform a parameter perturbation superposition process according to the perturbation amount matrix to obtain a perturbation superposition result;

[0034] Perform a numerical normalization process on the perturbation superposition result to obtain a normalized perturbation data, and perform a parameter reconstruction process according to the normalized perturbation data to obtain the initial perturbation data.

[0035] Optionally, in the sixth implementation manner of the first aspect of the present invention, the dynamic on-chain adversarial verification of the model access behavior according to the perturbed process parameter data and the process association hash value, and the differential perturbation intensity adjustment and access restriction of the distributed storage supply chain data respectively when an abnormal model access is detected include:

[0036] Perform an intelligent contract deployment process on the perturbed process parameter data to obtain an adversarial verification contract, and perform a verification rule generation process according to the adversarial verification contract and the process association hash value to obtain a dynamic on-chain verification rule;

[0037] Perform a feature extraction process on the model access behavior to obtain an access feature vector, and perform an adversarial verification process according to the access feature vector and the dynamic on-chain verification rule to obtain a verification score and an anomaly flag;

[0038] Perform a credibility quantification process on the model access behavior according to the verification score and the anomaly flag to obtain an access credibility, and perform a dynamic threshold determination process on the access credibility to obtain an access control decision;

[0039] Perform a differential perturbation intensity calculation process on the distributed storage supply chain data according to the access control decision to obtain a perturbation adjustment plan, and perform an on-chain deployment process on the perturbation adjustment plan to obtain an updated access restriction policy.

[0040] In a second aspect of the present invention, a supply chain data security management system is provided. The supply chain data security management system includes:

[0041] A hierarchical storage module for segmenting and distributing the storage of supply chain data according to the production links of the supply chain, and generating process-related hash values based on the process constraint information and the distributed supply chain data;

[0042] A feature adversarial module for constructing a feature extractor containing process constraint relationships based on the supply chain data and the process-related hash values, and performing domain adversarial feature extraction on the supply chain data based on the feature extractor to obtain supply chain data features that meet physical constraints;

[0043] A differential perturbation module for performing multi-level differential perturbation on the supply chain data according to the supply chain data features and a preset process parameter association map, and correcting the perturbed data according to the process constraint relationships to obtain perturbed process parameter data;

[0044] An access verification module for dynamically performing on-chain adversarial verification on the model access behavior according to the perturbed process parameter data and the process-related hash values, and separately adjusting the differential perturbation intensity and restricting access to the supply chain data stored distributively when detecting abnormal model access.

[0045] In the above supply chain data security management method and system, the supply chain data is segmented and distributed for storage according to the production links to generate process-related hash values; a feature extractor containing process constraint relationships is constructed based on the supply chain data and the process-related hash values, and domain adversarial feature extraction is used to obtain supply chain data features that meet physical constraints; multi-level differential perturbation is performed according to the supply chain data features and the process parameter association map, and the perturbed data is corrected based on the process constraint relationships; dynamic on-chain adversarial verification is performed on the model access behavior according to the perturbed process parameter data and the process-related hash values, and the differential perturbation intensity is adjusted and access is restricted when abnormal access is detected. By introducing process constraints and multi-level differential perturbation in the transfer learning process, the present invention prevents competitors from obtaining trade secret information through model reverse inference and ensures the security and effectiveness of data sharing.

[0046] Other features and advantages of the present invention will be described in the following specification, and, in part, will be obvious from the specification, or will be understood by implementing the present invention. The objectives and other advantages of the present invention are realized and obtained by the structures specifically pointed out in the specification, the claims, and the drawings.

[0047] To make the above objectives, features, and advantages of the present invention more obvious and understandable, the following specific preferred embodiments are given and described in detail in conjunction with the accompanying drawings. Brief Description of the Drawings

[0048] Figure 1 It is a schematic diagram of the first embodiment of the supply chain data security management method in the embodiments of the present invention;

[0049] Figure 2 It is a schematic diagram of an embodiment of the supply chain data security management system in the embodiments of the present invention. Detailed Embodiments

[0050] To make the objectives, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Apparently, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0051] The terms "including" and "having" and any variations thereof mentioned in the embodiments of the present invention are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally further includes other steps or units not listed, or optionally further includes other steps or units inherent to these processes, methods, products or devices.

[0052] For ease of understanding of this embodiment, first, a supply chain data security management method disclosed in the embodiments of the present invention will be introduced in detail. As Figure 1 shown, the method includes the following steps:

[0053] 101. Perform hierarchical partitioning and distributed storage processing on supply chain data according to the production links of the supply chain, and generate process-related hash values based on process constraint information and the distributedly stored supply chain data;

[0054] In an embodiment of the present invention, the performing hierarchical partitioning and distributed storage processing on supply chain data according to the production links of the supply chain, and generating process-related hash values based on process constraint information and the distributedly stored supply chain data includes: classifying the supply chain data according to the production links to obtain raw material link data, manufacturing link data, and product link data; using IPFS distributed storage to perform block storage on the classified supply chain data to obtain data blocks and data block indexes; constructing a Merkle tree for the data blocks based on the process constraint information to obtain a Merkle tree with process constraints; and generating process-related hash values according to the Merkle tree with process constraints and the process constraint information.

[0055] Specifically, when classifying data in the production process, first establish a data attribute mapping table, which contains multiple dimensions of information such as data type identifiers, data source identifiers, data timestamps, and data business attributes. For data in the raw material process, the system extracts supplier qualification information (such as quality system certifications, environmental certifications, etc.), raw material basic information (such as physical properties, chemical compositions, specifications, etc.), raw material inspection data (such as incoming inspection reports, physical property test data, chemical composition analysis data, etc.), inventory information (such as batch numbers, storage locations, storage conditions, etc.), and upstream supply chain traceability information (such as raw material origins, transportation information, transfer warehouse information, etc.). For data in the manufacturing process, focus on production equipment operation parameters (such as equipment temperature, pressure, rotation speed, power, etc.), process control parameters (such as processing temperature, holding time, cooling rate, additive ratio, etc.), on-line detection data (such as real-time quality detection data, process parameter monitoring data, equipment status monitoring data, etc.), production environment parameters (such as workshop temperature and humidity, cleanliness, air quality, etc.), and production process traceability information (such as operators, production batches, material input records, etc.). For data in the product process, the system collects finished product inspection data (such as appearance inspection, performance testing, reliability verification, etc.), packaging information (such as packaging specifications, protection requirements, identification information, etc.), warehousing and logistics data (such as storage conditions, transportation requirements, distribution information, etc.), customer usage feedback (such as product usage data, performance evaluation, fault information, etc.), and after-sales service records (such as repair records, upgrade and transformation information, life prediction, etc.).

[0056] Specifically, when using the IPFS distributed storage system for data storage, first preprocess the data for each process, including operations such as data format standardization, data encoding conversion, and data compression. Then, according to the principles of data temporal correlation and spatial locality, determine the optimal data slicing strategy. The system uses a dynamic slicing algorithm and adopts different slice sizes for different types of data: for structured data (such as production parameters, detection data, etc.), use fixed-size slices (256KB); for unstructured data (such as images, videos, etc.), dynamically adjust the slice size according to the complexity of the data content (512KB - 1MB). Calculate the content-addressable hash value of each data slice using the SHA-256 algorithm, and at the same time record the association relationship between data slices. The system also establishes a multi-level index structure: the first-level index records the basic information of data blocks (such as hash values, sizes, creation times, etc.); the second-level index records the storage locations and access paths of data blocks; the third-level index records the association relationship and organizational structure between data blocks. At the same time, for frequently accessed data blocks, the system retains a cache on the local node and establishes a cache update mechanism to ensure data consistency and availability.

[0057] Specifically, when constructing a Merkle tree with process constraints, it is first necessary to formally describe and mathematically model the process constraint information. The process constraint information includes quantitative relationships between parameters (such as thermodynamic equations, material mechanics equations, etc.), qualitative relationships (such as trends in process parameter changes, correlations of quality characteristics, etc.), and normative requirements (such as safety limits, process regulations, etc.). The system converts these constraint relationships into constraint verification functions. The input of the function is the relevant process parameters, and the output is the constraint satisfaction score. During the construction of the Merkle tree, the data blocks are first organized into a tree structure according to the process flow sequence. The leaf nodes store the hash values of the data blocks. In addition to storing the hash values of the child nodes, the non-leaf nodes also embed the corresponding process constraint verification functions. The system constructs the tree structure in a bottom-up manner: first, calculate the hash values of the leaf nodes, then determine the connection relationships and weights between the nodes according to the process constraint relationships, and finally calculate the hash values and constraint verification functions of the non-leaf nodes layer by layer.

[0058] Specifically, the process of generating the process-related hash value integrates two dimensions: data integrity verification and process constraint verification. First, the system executes the process constraint verification function for each node in the Merkle tree to obtain a series of constraint satisfaction scores. These scores reflect whether the data conforms to process laws and physical laws. Then, the system performs bitwise operations (including exclusive OR operations and circular shift operations) on the constraint satisfaction scores and the original hash values of the nodes to obtain intermediate hash values that incorporate the process constraint information. Next, the system uses an improved SHA-256 algorithm to perform multiple rounds of calculations on these intermediate hash values. The algorithm adds an encoding process for the process constraint information on the basis of the standard SHA-256 to ensure that the finally generated hash value can reflect the process relevance of the data. The finally generated process-related hash value is represented in hexadecimal. The high 128 bits are used for data integrity verification, and the low 128 bits are used for process constraint verification. This hash value, together with the data block index, is recorded in the blockchain network as the basis for data access and verification.

[0059] 102. Construct a feature extractor containing process constraint relationships based on supply chain data and process-related hash values, and perform domain adversarial feature extraction on the supply chain data based on the feature extractor to obtain supply chain data features that meet physical constraints;

[0060] In an embodiment of the present invention, constructing a feature extractor containing process constraint relationships based on the supply chain data and process-related hash values, and performing domain adversarial feature extraction on the supply chain data based on the feature extractor to obtain supply chain data features that meet physical constraints, including: performing data domain marking processing on process parameter data in the supply chain data to obtain data samples with source identifiers, grouping the process-related hash values according to the source identifiers to obtain feature constraint conditions for each data domain; performing source domain and target domain partitioning processing on the data samples to obtain an initial domain partitioning result, and performing consistency verification processing on the initial domain partitioning result according to the feature constraint conditions to obtain a domain partitioning strategy that meets process constraints; performing feature space mapping processing on the process parameter data according to the domain partitioning strategy to obtain an initial feature representation, performing gradient reversal layer embedding processing on the initial feature representation to obtain an adversarial feature extractor; performing process constraint injection processing on the adversarial feature extractor to obtain a feature extraction network with process constraints, and performing feature extraction processing on the supply chain data according to the feature extraction network to obtain supply chain data features that meet physical constraints.

[0061] Specifically, when performing data domain marking processing on the supply chain data, first construct a data source feature vector, which contains multi-dimensional identifiers such as the source enterprise information, production line information, equipment identifier, and acquisition time of the data. At the same time, perform semantic analysis on the process parameter data to extract process features therein. For example, for the injection molding process, extract key parameters such as mold temperature, injection pressure, and holding time; for the heat treatment process, extract process parameters such as heating temperature, holding time, and cooling rate. Based on these features, the system generates a unique source identification code for each data sample, and this identification code adopts a hierarchical coding method: the first layer represents the enterprise code, the second layer represents the production line code, the third layer represents the process type code, and the fourth layer represents the parameter type code. Then group the process-related hash values according to the source identifier, and establish a constraint relationship matrix between parameters within each group. This matrix describes the physical constraints, process limitations, and quality requirements between process parameters, thereby obtaining the feature constraint conditions for each data domain.

[0062] Specifically, the source domain and the target domain are then partitioned. First, the system constructs a feature space based on the statistical features (such as mean, variance, distribution pattern, etc.) and process features (such as parameter range, variation law, correlation, etc.) of the data samples. In this feature space, the clustering centers of the data samples are determined by calculating the Mahalanobis distance and process similarity between the samples. The initial partitioning of the source domain and the target domain is based on these clustering centers, and at the same time, the time series characteristics of the data are considered to ensure a certain continuity in the data distribution of adjacent time periods. For the initial domain partitioning result, the system conducts a consistency verification according to the feature constraint conditions, and the verification process includes parameter range verification, physical constraint verification, and process rule verification. If a partitioning result that does not meet the constraint conditions is found, it is optimized by adjusting the clustering parameters and boundary conditions, and finally a domain partitioning strategy that meets the process constraints is obtained.

[0063] Specifically, according to the domain partitioning strategy, the system constructs a feature space mapping network. This network adopts a multi-layer perceptron structure. The number of nodes in the input layer is the same as the dimension of the original process parameters. The hidden layer uses a non-linear activation function for feature transformation, and the dimension of the output layer is determined according to the principal component analysis result of the process features. Process constraint checking units are embedded in each layer of the network, and these units monitor the satisfaction of physical constraints during the feature transformation process. When performing gradient reversal layer embedding on the initial feature representation, first calculate the feature distribution difference between the source domain and the target domain, and then design a gradient reversal function, which maximizes the domain difference on the premise of keeping the process features unchanged. The network parameters are optimized through the backpropagation algorithm, so that the feature extractor can not only extract shared features but also maintain domain adversariality.

[0064] Specifically, finally, process constraints are embedded into the adversarial feature extractor to form a complete feature extraction network. The constraint injection process first converts physical laws and process rules into loss functions. For example, for heat treatment processes, thermodynamics laws are converted into constraint functions of temperature-time curves; for machining processes, material mechanics principles are converted into constraint functions of cutting parameters. These constraint functions are combined with the original loss function of the network to form a multi-objective optimization problem. Through an alternating optimization strategy, the satisfaction of process constraints is ensured while performing domain adversarial learning. When the network training converges, this feature extraction network is used to process the supply chain data, and the generated features not only maintain the essential features of the data but also meet physical constraints and have strong domain adaptability.

[0065] Further, the process of performing feature space mapping processing on process parameter data according to the domain partitioning strategy to obtain an initial feature representation and then performing gradient reversal layer embedding processing on the initial feature representation to obtain an adversarial feature extractor includes: performing feature dimension analysis processing on the domain partitioning strategy to obtain a dimension weight vector, performing weighted mapping processing on the process parameter data according to the dimension weight vector to obtain an initial feature space; performing non-linear transformation processing on the initial feature space to obtain a hidden layer feature representation, performing principal component projection processing on the hidden layer feature representation to obtain a dimensionality-reduced feature matrix; performing reverse gradient calculation processing on the dimensionality-reduced feature matrix to obtain a gradient reversal parameter, performing backpropagation processing according to the gradient reversal parameter to obtain an adversarial gradient matrix; performing feature reconstruction processing on the adversarial gradient matrix to obtain an adversarial feature representation, and performing feature extraction network construction processing according to the adversarial feature representation to obtain an adversarial feature extractor.

[0066] Specifically, when performing feature dimension analysis on the domain partitioning strategy, first calculate the contribution degree of each dimension of the process parameters to the overall feature space. The physical meaning and process importance of the parameters are considered during the analysis. For example, in the injection molding process, parameters such as mold temperature and injection pressure have higher weights, while auxiliary parameters such as mold opening and closing time have lower weights. By calculating the information entropy and mutual information of each dimension feature, the independence and information content of the features are evaluated, and the weights are corrected in combination with the knowledge of process experts. The system uses an adaptive weight allocation algorithm, which considers the physical coupling relationships between parameters, such as the relationships between temperature and pressure, speed and force, etc., and finally generates a dimension weight vector. Based on this weight vector, weighted mapping is performed on the process parameter data to convert the original parameter space into an initial feature space, which retains the physical meaning and correlation of the process parameters.

[0067] Specifically, when performing non-linear transformation on the initial feature space, a deep neural network structure is used, which includes multiple hidden layers. Each hidden layer uses a different non-linear activation function. For example, the ReLU function is used in the first layer to capture the positive activation features of the parameters, the tanh function is used in the second layer to process the bidirectional change features of the parameters, and the sigmoid function is used in the third layer to normalize the features. During the transformation process, multiple residual connections are set to ensure that the original process feature information is not lost in the deep network. At the same time, batch normalization is performed after each layer transformation to reduce the offset of the feature distribution. After non-linear transformation, a hidden layer feature representation is obtained, and these features contain the high-order correlations of the process parameters. Then, principal component analysis is performed on the hidden layer features, the covariance matrix and eigenvalues of the features are calculated, and the principal components with a cumulative contribution rate exceeding 95% are selected to construct a projection matrix, and the high-dimensional features are projected into a low-dimensional space to obtain a dimensionality-reduced feature matrix.

[0068] Specifically, when performing reverse gradient calculation, first construct a discriminator network, which is used to distinguish whether the features come from the source domain or the target domain. The discriminator adopts a multi-layer feedforward network structure and is trained by minimizing the binary cross-entropy loss. During the training process, calculate the gradients of the discriminator for the features, and these gradients reflect the contribution degree of the features to domain classification. Through the gradient reversal layer, reverse the directions of these gradients to generate gradient reversal parameters. Use these parameters for backpropagation to update the parameters of the feature extraction network, so that the extracted features can confuse the discriminator. This process is achieved through an alternating optimization algorithm. In each iteration, first fix the feature extractor to update the discriminator, and then fix the discriminator to update the feature extractor, and finally obtain the adversarial gradient matrix.

[0069] Specifically, finally, perform feature reconstruction based on the adversarial gradient matrix. The reconstruction process adopts an encoder-decoder structure. The encoder encodes the adversarial gradient information into the feature space, and the decoder restores the encoded features to the original parameter space. During the encoding process, use the attention mechanism to highlight important process features while maintaining the adversarial nature of the features. The reconstructed adversarial feature representation contains both the original process features and has strong domain transfer ability. Based on these feature representations, construct a multi-layer feature extraction network, which includes a feature extraction layer, an adversarial learning layer, and a feature fusion layer. The feature extraction layer uses a convolutional structure to capture local features, the adversarial learning layer realizes domain adaptation, and the feature fusion layer integrates features at different levels to finally form an adversarial feature extractor.

[0070] 103. Perform multi-level differential perturbation on the supply chain data according to the supply chain data characteristics and the preset process parameter association map, and correct the perturbed data according to the process constraint relationship to obtain the perturbed process parameter data;

[0071] In one embodiment of the present invention, the multi-level differential perturbation of the supply chain data is performed according to the supply chain data characteristics and the preset process parameter association map, and the perturbed data is corrected according to the process constraint relationship to obtain the perturbed process parameter data, including: quantitatively processing the process sensitivity of the supply chain data characteristics to obtain a parameter sensitivity matrix, performing correlation analysis processing according to the parameter sensitivity matrix and the process parameter association map to obtain a multi-level sensitive data set; performing differential privacy budget allocation processing on the multi-level sensitive data set to obtain perturbation budget values at all levels, performing parameter perturbation amount calculation processing on the perturbation budget values to obtain initial perturbation data; performing constraint verification processing on the initial perturbation data and the process constraint relationship to obtain a constraint violation data identifier, performing constraint inspection processing on the initial perturbation data according to the constraint violation data identifier to obtain the degree of constraint violation; performing process constraint correction processing on the initial perturbation data according to the degree of constraint violation to obtain corrected perturbation data, and performing process parameter reconstruction processing on the corrected perturbation data to obtain the perturbed process parameter data.

[0072] Specifically, when quantitatively processing the process sensitivity of the supply chain data characteristics, a parameter sensitivity evaluation system is first established, which includes three dimensions: process influence degree, commercial value degree, and leakage risk degree. For the process influence degree, scores are given by analyzing the direct influence degree of parameters on product quality. For example, in the steel smelting process, parameters such as temperature and chemical composition have a high process influence degree; for the commercial value degree, it is evaluated based on the correlation degree between parameters and the core competitiveness of the enterprise. For example, formula ratios, special process parameters, etc. have a high commercial value degree; for the leakage risk degree, it is quantified according to the possible loss degree caused by competitors obtaining the parameters. Through the comprehensive scoring of these three dimensions, a parameter sensitivity matrix is constructed. Then, cross-analysis is performed on this matrix and the process parameter association map to identify the association relationship and influence path between parameters, and finally the data is divided into three-level data sets of high sensitivity, medium sensitivity, and low sensitivity.

[0073] Specifically, when performing differential privacy budget allocation, an adaptive budget allocation strategy is adopted. First, a basic privacy budget is set for each sensitivity level. A larger privacy budget is allocated to the high-sensitive data set to provide stronger protection, while relatively smaller budgets are allocated to the medium-sensitive and low-sensitive data sets. Budget allocation also needs to consider the usage frequency and access scenario of the data. The budget value is appropriately increased for frequently accessed data, while it can be reduced for infrequently accessed data. Based on the allocated budget values, the Laplace mechanism is used to calculate the perturbation amount of each parameter. The size of the perturbation amount is inversely proportional to the budget value and directly proportional to the sensitivity of the parameter. By adding the calculated noise, the initial perturbation data is obtained.

[0074] Specifically, when performing constraint verification on the initial perturbation data, the quantitative constraints and qualitative constraints in the process constraints are first extracted. The quantitative constraints include the physical limits of parameters (such as temperature range, pressure range, etc.) and the numerical relationships between parameters (such as proportional relationship, summation relationship, etc.); the qualitative constraints include the change trends of parameters, the directions of correlation, etc. The perturbed data is verified through these constraint conditions. If a certain parameter value exceeds the physical limit or violates the relationship between parameters, it is marked as constraint-violating data. For the marked constraint-violating data, the degree of deviation from the constraint conditions is calculated, including the degree of exceeding the limit, the degree of violating the relationship, etc. These indicators together constitute the quantitative indicators of the constraint violation degree.

[0075] Specifically, finally, based on the constraint violation degree, the process constraints of the perturbation data are corrected. The correction process uses an iterative optimization method. In each iteration, the parameters with a larger violation degree are first corrected. When correcting, the original perturbation amount and the constraint conditions are comprehensively considered. If there is a strong correlation between parameters, the relevant parameters need to be corrected simultaneously to maintain the process constraint relationship. The corrected data needs to be re-verified for constraints until all parameters meet the process constraint conditions. Then, the process parameters of the corrected perturbation data are reconstructed. The reconstruction process includes parameter normalization, dimension restoration, and numerical modification to ensure that the reconstructed parameters not only maintain the perturbation effect but also meet the process requirements, and finally obtain the perturbed process parameter data that meets the actual process requirements.

[0076] Further, the differential privacy budget allocation process is performed on the multi-level sensitive data set to obtain the perturbation budget values at each level, and the parameter perturbation amount calculation process is performed on the perturbation budget values to obtain the initial perturbation data, including: performing a privacy sensitivity calculation process on the multi-level sensitive data set to obtain a sensitivity score matrix, performing a budget weight allocation process according to the sensitivity score matrix to obtain an initial budget allocation plan; performing a constraint optimization process on the initial budget allocation plan to obtain an optimized budget value, performing a Laplace noise generation process according to the optimized budget value to obtain a noise parameter set; performing a step-by-step perturbation amount calculation process on the noise parameter set to obtain a perturbation amount matrix, performing a parameter perturbation superposition process according to the perturbation amount matrix to obtain a perturbation superposition result; performing a numerical normalization process on the perturbation superposition result to obtain a normalized perturbation data, and performing a parameter reconstruction process according to the normalized perturbation data to obtain the initial perturbation data.

[0077] Specifically, when calculating the privacy sensitivity of a multi-level sensitive data set, a multi-dimensional sensitivity evaluation index system is first established. This system includes a data value dimension (such as the commercial value, technical value, strategic value, etc. of the data), a security risk dimension (such as the degree of impact of data leakage, the risk of data abuse, etc.), and a usage scenario dimension (such as the data access frequency, data sharing scope, etc.). For each dimension, detailed scoring criteria are set. For example, in the data value dimension, core process parameters, key quality indicators, etc. have a high value score; in the security risk dimension, formula data, process control parameters, etc. have a high risk score; in the usage scenario dimension, real-time monitoring data, quality inspection data, etc. have a high usage score. Through the analytic hierarchy process method, weighted calculations are performed on each dimension, and finally a sensitivity score matrix containing all evaluation indicators is obtained. Then, based on this matrix, a budget weight allocation strategy is designed, and a non-linear mapping function is used to convert the sensitivity score into a budget weight. The higher the sensitivity of the data, the higher the budget weight obtained, forming an initial budget allocation plan.

[0078] Specifically, when performing constraint optimization on the initial budget allocation plan, three main constraint conditions need to be considered: the overall privacy budget constraint, the process quality constraint, and the data availability constraint. First, based on the total privacy budget limit set by the system, the budget allocation values of each data item are normalized to ensure that the total budget does not exceed the limit. Second, according to the process quality requirements, a minimum interference threshold for key process parameters is set to ensure that the data after adding noise can still meet the process requirements. Finally, based on the data availability requirements, a tolerance range for data perturbation is set to ensure that the data still has practical application value after perturbation. The Lagrange multiplier method is used to solve these constraint conditions to obtain the optimized budget value that satisfies multiple constraints. According to the optimized budget value, random noise is generated using the Laplace distribution, and the scale parameter of the noise is inversely proportional to the budget value, thereby obtaining a set of noise parameters.

[0079] Specifically, when calculating the perturbation amount level by level, a hierarchical perturbation strategy is adopted. First, the perturbation amount of high-sensitivity data is calculated, and the magnitude of the perturbation amount is proportional to the budget value and the sensitivity of the data. Then, the perturbation amount of medium-sensitivity data is calculated, and at the same time, the correlation with high-sensitivity data is considered to avoid data inconsistency caused by perturbation. Finally, the perturbation amount of low-sensitivity data is calculated to ensure that the data after perturbation still maintains basic statistical characteristics. The perturbation amounts of all levels are organized into a perturbation amount matrix, which reflects the perturbation degrees of different levels and different parameters. Then, based on this matrix, perturbation superposition is performed on the original data, and the additive perturbation mechanism is adopted during the superposition process while maintaining the relative relationship between parameters.

[0080] Specifically, finally, numerical normalization is performed on the perturbation superposition result to ensure the consistency of data distribution. In the normalization process, the statistical characteristics of the perturbation data (such as mean, variance, etc.) are first calculated, and then normalization is performed by methods such as min-max normalization or z-score normalization. Different normalization strategies are adopted for different types of parameters. For example, linear normalization is used for continuous parameters, and interval mapping normalization is used for discrete parameters. The normalized data needs to be parameter-reconstructed. The reconstruction process includes dimension restoration, physical meaning conversion, and numerical modification to ensure that the reconstructed data not only maintains the privacy protection effect but also conforms to the value characteristics of actual process parameters, and finally the initial perturbation data is obtained.

[0081] 104. Perform dynamic on-chain adversarial verification on the model access behavior based on the perturbed process parameter data and the process-related hash value, and when an abnormal model access is detected, differentially adjust the perturbation intensity and access restrictions on the supply chain data stored distributively.

[0082] In an embodiment of the present invention, the performing dynamic on-chain adversarial verification on the model access behavior based on the perturbed process parameter data and the process-related hash value, and differentially adjusting the perturbation intensity and access restrictions on the supply chain data stored distributively when an abnormal model access is detected includes: performing intelligent contract deployment processing on the perturbed process parameter data to obtain an adversarial verification contract, generating a verification rule processing according to the adversarial verification contract and the process-related hash value to obtain a dynamic on-chain verification rule; performing feature extraction processing on the model access behavior to obtain an access feature vector, performing adversarial verification processing according to the access feature vector and the dynamic on-chain verification rule to obtain a verification score and an anomaly flag; performing credibility quantification processing on the model access behavior according to the verification score and the anomaly flag to obtain an access credibility, performing dynamic threshold determination processing on the access credibility to obtain an access control decision; calculating a differential perturbation intensity on the supply chain data stored distributively according to the access control decision to obtain a perturbation adjustment plan, and performing on-chain deployment processing on the perturbation adjustment plan to obtain an updated access restriction policy.

[0083] Specifically, when deploying a smart contract, it is first necessary to encode the perturbed process parameter data and convert it into a data structure that can be processed by the smart contract. The encoding process adopts a hierarchical structure. The first layer is the basic data layer, which includes the numerical value, unit, and timestamp of the parameter. The second layer is the perturbation information layer, which records the perturbation type, perturbation intensity, and perturbation range of the parameter. The third layer is the constraint information layer, which includes the physical constraints, process constraints, and quality constraints of the parameter. Based on these encoded data, an adversarial verification contract is constructed. The contract structure includes a data verification module, an access control module, and a perturbation adjustment module. By deploying the smart contract, the system embeds the process-related hash value into the contract and establishes a data integrity verification mechanism. During the generation of verification rules, the process constraint information is converted into rule expressions, including parameter threshold rules, parameter relationship rules, and timing constraint rules. These rules together constitute a dynamic on-chain verification rule set.

[0084] Specifically, for the feature extraction of model access behavior, a multi-dimensional analysis method is adopted. First, the frequency feature, cycle feature, and duration feature of the access are extracted from the time dimension; the geographical location feature, network topology feature, and device identification feature of the access are extracted from the space dimension; and the operation type feature, data volume feature, and access mode feature of the access are extracted from the behavior dimension. These features are standardized and then composed into an access feature vector. The vector dimension is determined by the number of features. During the adversarial verification process, the access feature vector is matched with the dynamic on-chain verification rules to calculate the compliance degree of the features and rules, and at the same time, the abnormality degree of the access behavior is analyzed. The verification score reflects the compliance of the access behavior. The score range is from 0 to 100, where a score higher than 80 indicates normal access, a score between 60 and 80 indicates that key attention is needed, and a score lower than 60 is marked as abnormal access. The abnormality identification details the specific type of abnormality, such as frequent access abnormality, unauthorized access abnormality, or pattern abnormality, etc.

[0085] Specifically, in the stage of quantifying access credibility, the system comprehensively considers the verification score and abnormality identification and constructs a credibility evaluation model based on multiple factors. The evaluation factors include historical access records, current access features, the degree of abnormal behavior, and the relevance of process knowledge. Each factor has a corresponding weight coefficient, and the access credibility score is obtained through weighted calculation. The dynamic threshold determination adopts an adaptive method, and the threshold parameters are dynamically adjusted according to the system operation status and security level. When the access credibility is lower than the threshold, the system generates corresponding control decisions. The decision types include temporarily restricting access, increasing the perturbation intensity, or completely rejecting access, etc. These decisions are then converted into specific access control policies.

[0086] Specifically, in the process of differential calculation of the perturbation intensity, first, the basic principles of perturbation adjustment are determined according to the access control decision. For access behaviors with low credibility, the system correspondingly increases the intensity of data perturbation, and the perturbation intensity is negatively correlated with the credibility. The perturbation adjustment scheme includes three key elements: perturbation parameter selection, perturbation intensity calculation, and perturbation implementation strategy. The specific calculation of the perturbation intensity takes into account the sensitivity of the parameters, the abnormality of the access, and the tolerance of the process. After the calculated perturbation adjustment scheme passes the security verification, it is deployed to the blockchain network through a smart contract. The execution of the contract will automatically trigger corresponding access restriction measures, including increasing data noise, reducing data precision, or restricting access frequency, etc. The transaction records generated during the entire deployment process are permanently stored in the blockchain, ensuring the transparency and immutability of the access control policy, and ultimately realizing the security protection of supply chain data.

[0087] In this embodiment, by collecting and fusing the multi-dimensional environmental data of the construction site, a real-time updated environmental digital twin model is constructed. According to this model and the individual physiological data of the workers, the safety status of each worker is adaptively evaluated to obtain an individualized dynamic safety threshold. Combining the environmental digital twin model, the individualized dynamic safety threshold, and the real-time status data, the graph neural network inference technology is used to comprehensively evaluate the safety situation and predict the potential risk trend. According to the safety situation assessment results and risk prediction trends, a context-aware hierarchical processing method is adopted to generate adaptive early warning and intervention plans for different risk levels, providing accurate safety tips. This method can monitor and predict the safety status of workers in real time according to the dynamic changes of the construction site, improve the intelligent level and accuracy of construction safety management, and reduce the risk of accidents.

[0088] The supply chain data security management method in the embodiments of the present invention has been described above. Next, the supply chain data security management system in the embodiments of the present invention will be described. Please refer to Figure 2 , an embodiment of the supply chain data security management system in the embodiments of the present invention includes:

[0089] The hierarchical storage module 201 is used to perform hierarchical division and distributed storage processing on the supply chain data according to the supply chain production links, and generate a process-related hash value based on the process constraint information and the distributed storage supply chain data;

[0090] The feature confrontation module 202 is used to construct a feature extractor containing process constraint relationships based on the supply chain data and the process-related hash value, and perform domain confrontation feature extraction on the supply chain data based on the feature extractor to obtain supply chain data features that meet physical constraints;

[0091] The differential perturbation module 203 is used to perform multi-level differential perturbation on the supply chain data according to the supply chain data characteristics and a preset process parameter association map, and correct the perturbed data according to the process constraint relationship to obtain the perturbed process parameter data;

[0092] The access verification module 204 is used to perform dynamic on-chain adversarial verification on the model access behavior according to the perturbed process parameter data and the process association hash value, and perform differential perturbation intensity adjustment and access restriction on the supply chain data stored distributively when detecting abnormal model access.

[0093] In the embodiments of the present invention, the supply chain data security management system runs the above supply chain data security management method. The supply chain data security management system generates a process association hash value by hierarchically dividing and distributively storing and processing the supply chain data according to production links; constructs a feature extractor containing process constraint relationships based on the supply chain data and the process association hash value, and uses domain adversarial feature extraction to obtain supply chain data characteristics that meet physical constraints; performs multi-level differential perturbation according to the supply chain data characteristics and the process parameter association map, and corrects the perturbed data based on the process constraint relationship; performs dynamic on-chain adversarial verification on the model access behavior according to the perturbed process parameter data and the process association hash value, and performs differential perturbation intensity adjustment and access restriction when detecting abnormal access. The present invention prevents competitors from obtaining trade secret information through model reverse inference and ensures the security and effectiveness of data sharing by introducing process constraints and multi-level differential perturbation in the transfer learning process.

[0094] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the above-described system, device, or unit can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0095] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The foregoing storage medium includes: various media such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disc that can store program codes.

[0096] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A supply chain data security management method, characterized in that: The supply chain data security management method comprises: The supply chain data is divided into layers and distributedly stored according to the production links of the supply chain, and the process-related hash value is generated based on the process constraint information and the distributedly stored supply chain data; Building a feature extractor containing a process constraint relationship based on the supply chain data and process-related hash values, and performing domain adversarial feature extraction on the supply chain data based on the feature extractor to obtain supply chain data features that meet physical constraints; Performing process sensitivity quantification processing on supply chain data features to obtain a parameter sensitivity matrix, performing correlation analysis processing based on the parameter sensitivity matrix and the process parameter correlation map to obtain a multi-level sensitive data set; performing differential privacy budget allocation processing on the multi-level sensitive data set to obtain disturbance budget values ​​at each level, performing parameter disturbance amount calculation processing on the disturbance budget values ​​to obtain initial disturbance data; performing constraint verification processing on the initial disturbance data and process constraint relationship to obtain constraint violation data identification, performing constraint verification processing on the initial disturbance data based on the constraint violation data identification to obtain the degree of constraint violation; performing process constraint correction processing on the initial disturbance data based on the degree of constraint violation to obtain corrected disturbance data, performing process parameter reconstruction processing on the corrected disturbance data to obtain disturbed process parameter data; The model access behavior is dynamically verified on the chain based on the disturbed process parameter data and the process-associated hash value. When abnormal model access is detected, differentiated disturbance intensity adjustments and access restrictions are performed on the distributed stored supply chain data.

2. The supply chain data security management method according to claim 1, characterized in that: The step of performing hierarchical division and distributed storage processing on the supply chain data according to the production links of the supply chain, and generating a process-related hash value based on the process constraint information and the distributedly stored supply chain data includes: Classify the supply chain data into production links to obtain raw material link data, manufacturing link data and product link data; Use IPFS distributed storage to store the classified supply chain data in blocks to obtain data blocks and data block indexes; Constructing a Merkle tree for the data block based on the process constraint information to obtain a Merkle tree with process constraints; A process-associated hash value is generated according to the Merkle tree with process constraints and process constraint information.

3. The supply chain data security management method according to claim 1, characterized in that: The step of constructing a feature extractor containing a process constraint relationship based on the supply chain data and the process-related hash value, and performing domain adversarial feature extraction on the supply chain data based on the feature extractor to obtain supply chain data features that satisfy physical constraints includes: Perform data domain labeling on process parameter data in the supply chain data to obtain data samples with source identifiers, group process process-related hash values ​​according to the source identifiers, and obtain feature constraint conditions for each data domain; Performing source domain and target domain division processing on the data sample to obtain an initial domain division result, performing consistency verification processing on the initial domain division result according to the feature constraint condition to obtain a domain division strategy that meets the process constraint; Performing feature space mapping processing on the process parameter data according to the domain partitioning strategy to obtain an initial feature representation, and performing gradient reversal layer embedding processing on the initial feature representation to obtain an adversarial feature extractor; The adversarial feature extractor is subjected to process constraint injection processing to obtain a feature extraction network with process constraints, and the supply chain data is subjected to feature extraction processing based on the feature extraction network to obtain supply chain data features that meet physical constraints.

4. The supply chain data security management method according to claim 3 is characterized in that: The process parameter data is subjected to feature space mapping processing according to the domain partitioning strategy to obtain an initial feature representation, and the initial feature representation is subjected to gradient reversal layer embedding processing to obtain an adversarial feature extractor, which includes: Performing feature dimension analysis processing on the domain partitioning strategy to obtain a dimension weight vector, and performing weighted mapping processing on the process parameter data according to the dimension weight vector to obtain an initial feature space; Performing nonlinear transformation processing on the initial feature space to obtain a hidden layer feature representation, and performing principal component projection processing on the hidden layer feature representation to obtain a reduced dimension feature matrix; Performing reverse gradient calculation processing on the dimension reduction feature matrix to obtain gradient reversal parameters, and performing back propagation processing according to the gradient reversal parameters to obtain an adversarial gradient matrix; The adversarial gradient matrix is ​​subjected to feature reconstruction processing to obtain an adversarial feature representation, and a feature extraction network is constructed based on the adversarial feature representation to obtain an adversarial feature extractor.

5. The supply chain data security management method according to claim 1, characterized in that: The performing differential privacy budget allocation processing on the multi-level sensitive data set to obtain disturbance budget values ​​at each level, and performing parameter disturbance amount calculation processing on the disturbance budget values ​​to obtain initial disturbance data includes: Performing privacy sensitivity calculation processing on the multi-level sensitive data set to obtain a sensitivity scoring matrix, and performing budget weight allocation processing according to the sensitivity scoring matrix to obtain an initial budget allocation plan; Performing constraint optimization processing on the initial budget allocation scheme to obtain an optimized budget value, and performing Laplace noise generation processing according to the optimized budget value to obtain a noise parameter set; Performing a step-by-step disturbance amount calculation process on the noise parameter set to obtain a disturbance amount matrix, and performing a parameter disturbance superposition process according to the disturbance amount matrix to obtain a disturbance superposition result; The disturbance superposition result is numerically normalized to obtain normalized disturbance data, and parameter reconstruction is performed according to the normalized disturbance data to obtain initial disturbance data.

6. The supply chain data security management method according to claim 1, characterized in that: The dynamic on-chain adversarial verification of the model access behavior is performed according to the disturbed process parameter data and the process-related hash value, and the differentiated disturbance intensity adjustment and access restriction are performed on the distributed storage supply chain data when abnormal model access is detected, including: Perform smart contract deployment processing on the disturbed process parameter data to obtain an adversarial verification contract, generate verification rules based on the adversarial verification contract and the process-related hash value, and obtain dynamic on-chain verification rules; Perform feature extraction on the model access behavior to obtain an access feature vector, perform adversarial verification based on the access feature vector and dynamic on-chain verification rules, and obtain a verification score and anomaly identification; Performing credibility quantification processing on the model access behavior according to the verification score and the abnormal identification to obtain access credibility, and performing dynamic threshold determination processing on the access credibility to obtain an access control decision; According to the access control decision, a differentiated disturbance intensity calculation process is performed on the distributedly stored supply chain data to obtain a disturbance adjustment plan, and the disturbance adjustment plan is deployed on-chain to obtain an updated access restriction strategy.

7. A supply chain data security management system, characterized in that: The supply chain data security management system includes: A hierarchical storage module is used to perform hierarchical division and distributed storage processing on the supply chain data according to the production links of the supply chain, and to generate a process-related hash value based on the process constraint information and the distributedly stored supply chain data; A feature adversarial module, used to construct a feature extractor containing a process constraint relationship based on the supply chain data and the process-related hash value, and to perform domain adversarial feature extraction on the supply chain data based on the feature extractor to obtain supply chain data features that meet physical constraints; A differential perturbation module is used to perform process sensitivity quantification processing on supply chain data features to obtain a parameter sensitivity matrix, perform correlation analysis processing based on the parameter sensitivity matrix and the process parameter correlation map to obtain a multi-level sensitive data set; perform differential privacy budget allocation processing on the multi-level sensitive data set to obtain perturbation budget values ​​at each level, perform parameter perturbation amount calculation processing on the perturbation budget values ​​to obtain initial perturbation data; perform constraint verification processing on the initial perturbation data and process constraint relationship to obtain a constraint violation data identifier, perform constraint verification processing on the initial perturbation data based on the constraint violation data identifier to obtain a constraint violation degree; perform process constraint correction processing on the initial perturbation data based on the constraint violation degree to obtain corrected perturbation data, perform process parameter reconstruction processing on the corrected perturbation data to obtain perturbed process parameter data; The access verification module is used to perform dynamic on-chain adversarial verification on the model access behavior based on the disturbed process parameter data and the process-associated hash value, and to perform differentiated disturbance intensity adjustment and access restriction on the distributedly stored supply chain data when abnormal model access is detected.

Images