Test spline identity tracing method and system based on block chain

Abstract

The invention discloses a test spline identity tracing method and system based on a block chain. The method comprises the following steps: collecting full-life-cycle multi-modal data of a test spline embedded with a PUF chip for quality evaluation and semantic mapping; generating an attack judgment mark and an abnormal conduction path so as to embed an encrypted watermark in the control instruction; carrying out block chain evidence storage on the standardized traceability data set; constructing a test spline dynamic knowledge graph to generate a knowledge enhancement traceability data set; analyzing a control instruction stream carrying a watermark, monitoring a deviation value between an execution result and an expected physical behavior, generating an execution deviation value and watermark feedback data, enhancing a traceability data set in combination with base knowledge, generating attack source coordinate information and a traceability path map, and pushing a traceability report which cannot be tampered. The space-time calibration of the multi-source heterogeneous data is realized, and the integrity and consistency of the traceability data are ensured; a PUF chip is adopted, and watermark information which cannot be tampered is embedded in a control instruction, so that double protection from a physical layer to a data layer is realized.

Description

Technical Field

[0001] This invention relates to the field of blockchain technology, and in particular to a blockchain-based test strip identity tracing method and system. Background Technology

[0002] In the fields of industrial manufacturing and testing, product traceability and quality control have always been core technical challenges. Traditional product traceability systems mainly rely on single-dimensional data recording, such as QR codes and RFID tags, which suffer from problems such as information silos, easy data tampering, and incomplete traceability chains. With the development of blockchain technology, blockchain-based traceability systems are gradually being applied in the industrial field, ensuring the immutability and traceability of data through distributed ledger technology. However, existing technologies still have the following technical problems: the multimodal data collected by traditional traceability systems, such as temperature, pressure, and images, are heterogeneous, with inconsistent timestamps, spatial coordinates, and data formats from different data sources, leading to difficulties in data fusion, a lack of quality assessment mechanisms, and difficulty in forming a complete traceability evidence chain; during data transmission and control command execution, there is a lack of effective protection mechanisms, making it easy for malicious attackers to tamper with or forge data; and there is a lack of deep fusion and knowledge reasoning capabilities for multimodal data, making it difficult to accurately identify attack sources and generate complete traceability path maps, resulting in low traceability efficiency. Summary of the Invention

[0003] To address at least one of the aforementioned technical problems, this invention provides a blockchain-based test strip identity tracing method and system.

[0004] In a first aspect, the present invention provides a blockchain-based test spline identity tracing method, the method comprising: Collect multimodal data of the entire life cycle of the test spline, extract multimodal features, and perform quality assessment and semantic mapping on the multimodal data to generate a standardized traceability dataset; the test spline is embedded with a PUF chip, which generates a unique identifier based on the physical randomness of the test spline material; Attack detection flags and abnormal propagation paths are generated based on standardized source tracing datasets to embed encrypted watermarks in control commands and generate control command streams carrying watermarks; the standardized source tracing datasets are stored on the blockchain, and the PUF identifier, watermark hash value, and data storage address are written into the blockchain as evidence metadata. A dynamic knowledge graph of test splines is constructed based on multimodal features. Anomaly detection and relation reasoning are performed on the knowledge graph to generate a knowledge-enhanced tracing dataset. The control command stream carrying watermarks is parsed, and the deviation value between the execution result and the expected physical behavior is monitored to generate execution deviation value and watermark feedback data. Based on the knowledge-enhanced source tracing dataset, execution deviation values, and watermark feedback data, attack source coordinate information and source tracing path map are generated, and an tamper-proof source tracing report is pushed out.

[0005] Preferably, the quality assessment and semantic mapping of multimodal data includes: Extract multi-source heterogeneous data from multimodal data, establish data quality admission rules, and perform spatiotemporal interpolation to repair missing data; Construct a semantic recognition mechanism to extract entity relationships from unstructured data and achieve semantic mapping; Establish a spatiotemporal calibration mechanism to unify multi-source timestamps into standard time and transform spatial coordinates to the standard coordinate system.

[0006] Preferably, the method for generating the encrypted watermark includes: An initial seed value for the watermark is generated based on the attack determination flag, and the watermark sequence is generated iteratively through chaotic mapping. The watermark sequence is divided into independent watermark bits, a perturbable field of the control command is embedded, and a watermark header identifier is added to the reserved bits in the protocol.

[0007] Preferably, the blockchain-based evidence storage includes: Calculate the cryptographic hash value of the standardized traceability dataset and automatically verify the evidence metadata through a smart contract; The notarization request is verified by PBFT consensus through consensus nodes. Once the verification is successful, a digital notarization certificate is generated and anchored to the blockchain network.

[0008] Preferably, the construction of the dynamic knowledge graph includes: Extracting the temporal dynamic features of numerical time-series data, the semantic features of textual data, and the visual spatial features of image data; Based on the test spline tracing domain ontology, entity types and relation types are defined, and relation reasoning and incremental updates are performed through graph neural networks.

[0009] Secondly, the present invention also provides a blockchain-based test strip identity traceability system, the system comprising: The data acquisition unit is used to collect multimodal data of the test spline throughout its entire life cycle, extract multimodal features, and perform quality assessment and semantic mapping on the multimodal data to generate a standardized traceability dataset. The test spline is embedded with a PUF chip, which generates a unique identifier based on the physical randomness of the test spline material. The watermark embedding unit is used to generate attack judgment flags and abnormal propagation paths based on the standardized traceability dataset, so as to embed encrypted watermarks in control commands and generate control command streams carrying watermarks; the standardized traceability dataset is stored on the blockchain, and the PUF identifier, watermark hash value, and data storage address are written to the blockchain as evidence metadata. The deviation analysis unit is used to construct a dynamic knowledge graph of test splines based on multimodal features, perform anomaly detection and relation reasoning on the knowledge graph, generate a knowledge-enhanced tracing dataset, parse the control command stream carrying watermarks, monitor the deviation value between the execution result and the expected physical behavior, and generate execution deviation value and watermark feedback data. The identity tracing unit is used to generate attack source coordinate information and tracing path map based on knowledge-enhanced tracing datasets, execution deviation values ​​and watermark feedback data, and push tamper-proof tracing reports.

[0010] Preferably, the data acquisition unit is used to perform quality assessment and semantic mapping on multimodal data, including: Extract multi-source heterogeneous data from multimodal data, establish data quality admission rules, and perform spatiotemporal interpolation to repair missing data; Construct a semantic recognition mechanism to extract entity relationships from unstructured data and achieve semantic mapping; Establish a spatiotemporal calibration mechanism to unify multi-source timestamps into standard time and transform spatial coordinates to the standard coordinate system.

[0011] Preferably, the method for generating the encrypted watermark includes: An initial seed value for the watermark is generated based on the attack determination flag, and the watermark sequence is generated iteratively through chaotic mapping. The watermark sequence is divided into independent watermark bits, a perturbable field of the control command is embedded, and a watermark header identifier is added to the reserved bits in the protocol.

[0012] Thirdly, the present invention also provides an electronic device including a processor and a memory, the memory being used to store computer program code, the computer program code including computer instructions, wherein when the processor executes the computer instructions, the electronic device performs the method as described in the first aspect above and any possible implementation thereof.

[0013] Fourthly, the present invention also provides a computer-readable storage medium storing a computer program, the computer program including program instructions that, when executed by a processor of an electronic device, cause the processor to perform a method as described in the first aspect above and any possible implementation thereof.

[0014] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention establishes a multimodal data quality assessment and semantic mapping mechanism, achieving spatiotemporal calibration and standardization of multi-source heterogeneous data, ensuring the integrity and consistency of traceability data, and solving the data silo problem. It employs a PUF chip to generate unique identifiers, combined with encrypted watermarking technology, embedding tamper-proof watermark information into control commands, achieving dual protection from the physical layer to the data layer, effectively resisting malicious attacks. By constructing a dynamic knowledge graph and integrating multimodal features for anomaly detection and relationship reasoning, it can accurately locate attack sources and generate complete traceability path graphs, significantly improving traceability efficiency. Based on knowledge-enhanced traceability datasets and real-time monitoring data, the system can proactively warn of abnormal behavior, realizing a shift from passive traceability to proactive protection, and improving the full lifecycle management level of industrial test samples. Through blockchain notarization technology, key data is stored on the blockchain, ensuring the immutability and verifiability of traceability reports.

[0015] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the background art, the accompanying drawings used in the embodiments of the present invention or the background art will be described below.

[0017] Figure 1 A flowchart illustrating a blockchain-based test spline identity tracing method provided in an embodiment of the present invention; Figure 2 A schematic diagram of the structure of a blockchain-based test strip identity traceability system provided in an embodiment of the present invention; Detailed Implementation

[0018] To enable those skilled in the art to better understand the present invention, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Please see Figure 1 , Figure 1 This is a flowchart illustrating a blockchain-based test spline identity tracing method provided in an embodiment of the present invention. Figure 1 As shown, the method includes: S10. Collect multimodal data of the entire life cycle of the test spline, extract multimodal features, and perform quality assessment and semantic mapping on the multimodal data to generate a standardized traceability dataset; the test spline is embedded with a PUF chip, and the PUF chip generates a unique identifier based on the physical randomness of the test spline material.

[0020] The mechanical properties (e.g., tensile strength, fatigue life) and environmental parameters (temperature, humidity) of the test specimens are monitored in real time using embedded sensors (e.g., strain gauges, temperature sensors). The sampling frequency is dynamically adjusted according to the specimen type (≥1 kHz for high-frequency impact tests, ≥10 Hz for static tests). A unique identifier (ID) is generated using the physical randomness of the PUF chip (e.g., differences in material microstructure), and the digital fingerprint (e.g., 256-bit hash value) output by the chip is collected via a dedicated reader. PUF response time and stability indicators (e.g., bit error rate) are recorded to ensure that the ID cannot be cloned during its lifecycle. Operation commands, timestamps, and equipment status (e.g., load accuracy, displacement resolution) are collected from the test equipment control system, and the device window handle is obtained in real time through Windows low-level API listening technology, avoiding reliance on network connections. During multimodal data synchronization, a hardware-triggered synchronization mechanism is used, using the timestamp generated by the PUF chip as a reference to align sensor data, operation logs, and image data, with synchronization errors controlled within ±1 ms. For unstructured data (e.g., the specimen deformation process recorded by high-speed cameras), keyframe extraction and time encoding are used to correlate it with physical parameters.

[0021] To extract multimodal features, temporal features (mean, variance, peak value) and frequency domain features (major frequency component obtained through FFT transform) are extracted from sensor data to characterize the dynamic response behavior of splines. For fatigue test data, wavelet transform is used to decompose the signal and extract energy distribution features at different scales. Error correction encoding is applied to the original PUF identifier to reduce bit drift caused by environmental interference and generate a stable enhanced ID. The consistency hash value of the PUF response is calculated as the basis for verifying the integrity of traceability data. Convolutional neural networks (CNNs) are used to extract key regional features (such as crack propagation areas) from the test image data, and a comprehensive feature vector is generated through a multimodal weighted fusion algorithm (e.g., physical feature weight 0.6, image feature weight 0.4).

[0022] In one embodiment, the quality assessment and semantic mapping of multimodal data includes: Extract multi-source heterogeneous data from multimodal data, establish data quality admission rules, and perform spatiotemporal interpolation to repair missing data; Construct a semantic recognition mechanism to extract entity relationships from unstructured data and achieve semantic mapping; Establish a spatiotemporal calibration mechanism to unify multi-source timestamps into standard time and transform spatial coordinates to the standard coordinate system.

[0023] Based on the national standard GB / T 36344-2018 "Information Technology Data Quality Evaluation Indicators", a quality assessment system is constructed, encompassing six dimensions: standardization, completeness, accuracy, consistency, timeliness, and accessibility. Specifically, it includes 17 core indicators such as: data standard compliance (e.g., date format, encoding rules), metadata completeness (data dictionary, field descriptions), business rule validity (e.g., ID card verification, age range constraints), field null value rate, data duplication rate, and value range compliance. By setting thresholds (e.g., null value rate ≤ 5%, duplication rate ≤ 3%), quantifiable assessment of data quality is achieved.

[0024] A two-stage iterative imputation strategy is adopted: the first stage addresses short-term random missing data by comparing and evaluating methods such as linear interpolation, K-nearest neighbor interpolation, and random forest interpolation, and selecting the optimal method for imputation; the second stage addresses long-term continuous missing data by combining an autoregressive moving average model for time series prediction imputation. For spatially missing data, Kriging interpolation or inverse distance weighting is used, combined with terrain compensation parameters to compensate for atmospheric refraction errors in elevation coordinates. A spatiotemporal joint verification mechanism ensures the physical rationality and topological consistency of the imputed data.

[0025] A heterogeneous graph-based entity relation extraction method based on semantic fusion is adopted: First, a pre-trained model is used to perform context encoding on unstructured text to obtain word vector representations; then, a heterogeneous graph is constructed, and relation information is incorporated as prior knowledge into the word representations to enhance entity recognition capabilities; finally, a semantic fusion module aggregates multi-granularity features (entity referential features, fragment features, and sentence-level features) to achieve relation classification. This method can effectively handle single entity overlap and entity pair overlap problems, improving the F1 score on the NYT and WebNLG datasets. A knowledge graph semantic mapping framework is used to transform unstructured data into structured knowledge: a unified semantic model is defined to map raw features to standardized attributes; the R2RML mapping language is used to convert relational database table structures into RDF triple format; entity mapping rules are established to bind graph node types to semantic entities, and relation types are transformed into action paths that can be recognized by NLP engines. A multi-feature semantic enhancement mechanism is used to fuse lexical features, syntactic features, and semantic role features to improve the accuracy of semantic understanding.

[0026] Based on Coordinated Universal Time (UTC), the timestamp sequences of multi-source data are dynamically calibrated using Network Time Protocol (NTP) or Precise Time Protocol (PTPv2). Timestamp alignment algorithms (such as linear interpolation and spline interpolation) compensate for millisecond-level timing misalignments caused by clock drift between devices. A time quality identifier field is established to label the time source type (such as BeiDou or GPS) and synchronization accuracy, facilitating verification of credibility. For asynchronous data, an event-based timestamp calibration mechanism is employed to ensure the accuracy and consistency of multi-source data in the time dimension. A three-level spatial registration process is implemented: Level 1 transformation is based on geographic coordinates in device metadata, converting local relative coordinates to geodetic coordinates using an affine transformation matrix; Level 2 correction loads terrain compensation parameters based on device type identifiers to compensate for atmospheric refraction errors in elevation coordinates; Level 3 alignment uses a nonlinear model (such as a thin-plate model or spherical projection model), solving for transformation parameters using the least squares algorithm to achieve centimeter-level spatial coordinate unification. Through spatial error correction and dynamic monitoring and model adjustment, the geometric comparability and topological computability of cross-device spatial data are ensured. The system performs joint verification of the physical rationality and topological consistency of the execution time-space trajectory, generating spatiotemporal metadata with verification tags, including drift error values, coordinate transformation confidence scores, and indexes of abnormal trajectory segment locations. By constructing a spatiotemporal data cube, using the device identifier as the primary key and the timestamp sequence as the time-series index, and storing it in a columnar storage format according to a preset time window slice, efficient management and querying of spatiotemporal data are achieved.

[0027] S20. Based on the standardized traceability dataset, generate attack judgment flags and abnormal propagation paths to embed encrypted watermarks in control commands and generate a control command stream carrying the watermark; perform blockchain notarization on the standardized traceability dataset and write the PUF identifier, watermark hash value, and data storage address as notarization metadata into the blockchain.

[0028] Based on a standardized source tracing dataset, an attack detection flag is generated using a multi-dimensional feature fusion algorithm. First, multimodal features are extracted from the standardized source tracing dataset, including physical features (such as strain, temperature, and load), PUF chip response features, and operation log features. A feature importance ranking algorithm (such as feature importance evaluation based on random forest) is used to select key feature dimensions sensitive to attack detection. Based on feature selection, a machine learning-based attack detection model is constructed. Support Vector Machines (SVM) or Deep Neural Networks (DNN) are used as classifiers, trained on historical attack samples and normal samples to learn the characteristic patterns of attack behavior. The model output is an attack detection flag, including attack type (such as malicious tampering, physical destruction, and forgery attacks), attack confidence score (a probability value between 0 and 1), and attack timestamp.

[0029] When analyzing anomaly propagation paths, a graph theory algorithm is used to construct an anomaly propagation network model. Multimodal data acquisition nodes, control command execution nodes, and physical response monitoring nodes of the test spline are considered as network nodes, and dependencies between nodes (such as data flow, control flow, and physical impact relationships) are considered as edges. Time series correlation analysis and causal inference algorithms are used to quantify the anomaly propagation strength between nodes. Anomaly propagation path identification employs a path search algorithm based on dynamic programming. Starting from the anomaly node corresponding to the attack determination flag, a breadth-first search is performed along the network edges to calculate the anomaly propagation probability of each path. By setting a propagation threshold, the most probable anomaly propagation path is selected. Finally, an anomaly propagation path graph containing information such as the attack source node, propagation path, propagation time, and impact range is generated.

[0030] Preferably, the method for generating the encrypted watermark includes: An initial seed value for the watermark is generated based on the attack determination flag, and the watermark sequence is generated iteratively through chaotic mapping. The watermark sequence is divided into independent watermark bits, a perturbable field of the control command is embedded, and a watermark header identifier is added to the reserved bits in the protocol.

[0031] The watermark generation and embedding process is as follows: Initial seed construction: Attack determination flags (such as attack type, confidence level, and hash value of timestamp) are used as the initial seed value for the chaotic system. This seed is highly sensitive; even small changes in the attack flags can lead to drastically different watermark sequences. Chaotic mapping iteration: Iteration is performed using chaotic models such as Logistic mapping or Lorenz systems. This process is extremely sensitive to initial conditions, generating near-random, unpredictable watermark sequences with very long periodicity, effectively enhancing the watermark's concealment and resistance to cracking. Field segmentation and embedding: The watermark sequence generated by chaos is segmented bit by bit. It is precisely embedded into perturbable fields in control instructions without affecting normal execution, such as the low bits of certain parameters or reserved fields in the protocol.

[0032] Add header identifier: Add a specific watermark header identifier to the reserved space in the communication protocol. This identifier does not participate in the business logic and is mainly used to quickly and accurately locate and extract the watermark in the subsequent traceability process.

[0033] Control commands embedded with encrypted watermarks are organized into a command stream according to their time sequence. The command stream employs a timestamp synchronization mechanism to ensure the consistency of command execution timing. Simultaneously, integrity check codes (such as CRC checks) are added to the command stream to prevent commands from being tampered with during transmission. The final generated control command stream carrying the watermark possesses the following characteristics: verifiable command integrity, extractable watermark, and traceable execution timing.

[0034] Because the watermark is seamlessly integrated into the instruction stream, it has virtually no impact on normal control operations (transparency). Furthermore, the watermark sequence based on chaotic mapping effectively resists various attacks such as tampering and duplication, ensuring the integrity and reliability of the traceability identifier. The embedded header identifier enables the watermark to be quickly identified and extracted from complex instruction streams. This lays the foundation for subsequent steps of parsing the instruction stream, monitoring execution deviations, and performing accurate traceability.

[0035] Preferably, the blockchain-based evidence storage includes: Calculate the cryptographic hash value of the standardized traceability dataset and automatically verify the evidence metadata through a smart contract; The notarization request is verified by PBFT consensus through consensus nodes. Once the verification is successful, a digital notarization certificate is generated and anchored to the blockchain network.

[0036] Specifically, the blockchain notarization and metadata writing process is as follows: A blockchain-based evidence storage platform is built using a consortium blockchain architecture, maintained jointly by multiple trusted nodes. Standardized traceability datasets generate data fingerprints (hash values) using a hash algorithm. These hash values, rather than the original data, are written to the blockchain, protecting data privacy while ensuring data integrity. Data upload to the blockchain is automatically executed via smart contracts; once the standardized traceability dataset is generated, the evidence storage process is automatically triggered.

[0037] The PUF identifier, watermark hash value, and data storage address are written into the blockchain as evidence-keeping metadata. The PUF identifier serves as a unique identifier for the test sample, ensuring the non-forgeability of the traceability data; the watermark hash value is used to verify the integrity and origin of control instructions; and the data storage address points to the actual storage location of the standardized traceability dataset (such as the IPFS distributed storage network). Metadata writing employs a transaction mechanism to ensure the atomicity of all metadata fields, avoiding data inconsistencies caused by partial writes.

[0038] After the notarization is completed, a notarization certificate containing information such as block height, transaction hash, and timestamp is generated. The certificate can then be used to verify the integrity of the data and the notarization time. During verification, the hash value of the standardized traceability dataset is recalculated and compared with the hash value stored on the blockchain. If they match, it proves that the data has not been tampered with. Simultaneously, the legitimacy of the PUF identifier and watermark hash value can be verified to ensure the data source is trustworthy.

[0039] By following the steps above, the accuracy of attack detection can be improved and the false positive rate reduced. Anomaly propagation path identification time is shortened to milliseconds, supporting real-time attack tracing. The encrypted watermark exhibits excellent robustness, with strong resistance to compression, pruning, and noise attacks. Blockchain-based evidence storage is tamper-proof, with a verification response time of less than 1 second. The storage efficiency of standardized traceability datasets is significantly improved, and it supports petabyte-level data storage, thus enabling trusted evidence storage and traceability of test sample data throughout its entire lifecycle.

[0040] S30. Construct a dynamic knowledge graph of test splines based on multimodal features, perform anomaly detection and relation reasoning on the knowledge graph, and generate a knowledge-enhanced tracing dataset; parse the control command stream carrying watermarks, monitor the deviation value between the execution result and the expected physical behavior, and generate execution deviation value and watermark feedback data.

[0041] A knowledge graph is constructed using a multi-source heterogeneous data fusion algorithm. First, physical features (strain, temperature, load, etc.), PUF chip response features, and operation log features are aligned, and a timestamp synchronization mechanism ensures consistency of multi-source data across time. Then, entity recognition algorithms (such as BERT-based named entity recognition) are used to extract key entities from unstructured data, including test spline ontology, experimental equipment, operators, and experimental parameters. A relation extraction model (such as a graph neural network-based relation classifier) ​​is used to identify semantic relationships between entities, including "test spline - located at - experimental equipment," "operator - performing - experimental operation," and "experimental parameter - affecting - test spline." A knowledge representation in RDF triple format is constructed, storing entities and relations in a graph database (such as Neo4j or JanusGraph). A dynamic update mechanism captures the state changes of the test spline throughout its entire lifecycle, forming a dynamically evolving knowledge graph. A domain ontology (such as a materials testing domain ontology) is introduced to semantically enhance the knowledge graph, mapping the original features to standardized semantic concepts. Knowledge representation learning (such as TransE and RotatE) is used to embed entities and relations into a low-dimensional vector space, supporting similarity calculation and relation reasoning. Rule-based reasoning is implemented through a rule-based reasoning engine, such as "If the test spline temperature exceeds the threshold, trigger an anomaly alarm".

[0042] A multi-scale anomaly detection framework is adopted, including: Graph Structure Anomaly Detection: An anomaly detection model based on Graph Neural Networks (GNNs) learns the structural patterns of knowledge graphs under normal conditions to detect anomalous nodes or edges. Temporal Anomaly Detection: Anomaly detection is performed on temporal attributes (such as load changes and temperature fluctuations) in the knowledge graph. Time series decomposition algorithms (such as STL decomposition) are used to separate trends, seasonality, and residuals. Residual anomalies are detected through isolated forests or autoencoders. Multimodal anomaly fusion: It fuses structural anomalies, temporal anomalies, and physical feature anomalies in the fusion graph, and generates a comprehensive anomaly score through a weighted fusion strategy.

[0043] Relationship reasoning based on knowledge graphs reveals implicit connections. Path ranking algorithms (such as PRA) are used to search the knowledge graph for paths connecting anomalous nodes and potential attack sources. Graph traversal algorithms (such as breadth-first search and depth-first search) are employed to discover anomalous propagation paths and identify attack transmission chains. Causal inference algorithms (such as Granger causality tests) are used to verify the causal relationships between anomalous nodes, enhancing the credibility of the reasoning results.

[0044] The generated control command stream carrying the watermark is parsed. First, the watermark location is identified by the watermark header, and the encrypted watermark information is extracted. An asymmetric decryption algorithm (such as RSA) is used to decrypt the watermark, obtaining the original watermark information (including attack detection flags, timestamps, etc.). The integrity and authenticity of the watermark are verified to ensure the command stream has not been tampered with. The execution results of the control commands are monitored in real time, including collecting physical response data (such as displacement, strain, and temperature) of the test sample through sensors; the expected physical response is calculated based on the control commands and the physical model of the test sample; the deviation between the actual response and the expected response is calculated using Euclidean distance or cosine similarity, generating a deviation time series. The watermark parsing results are correlated with the execution deviation values ​​to generate watermark feedback data. This includes: watermark integrity and authenticity verification results; analysis of the correlation between watermark information and execution deviations to determine if an attack caused the deviation; and encapsulation of the watermark information, deviation values, timestamps, etc., into structured feedback data.

[0045] This dataset integrates the original standardized source tracing dataset, knowledge graph inference results, execution deviation values, and watermark feedback data. A multi-source data alignment algorithm is employed, linking data based on timestamps and entity identifiers. New features are extracted through feature engineering, such as anomaly propagation path features, causal relationship strength features, and deviation change trend features. The integrated data is then converted into a unified knowledge-enhanced source tracing dataset format and stored in JSON-LD or RDF format, containing the following core fields: Entity information: Standardized descriptions of entities such as test strips, equipment, and personnel; Relational information: semantic and inferential relationships between entities; Anomaly information: anomaly detection results, anomaly score, anomaly type; Deviation information: execution deviation value, deviation time series, and deviation correlation analysis; Watermark information: watermark verification results, watermark correlation analysis; Metadata: data source, processing time, data quality score.

[0046] Finally, the quality of the generated knowledge-enhanced tracing dataset was validated, including: Integrity verification: Check if key fields are missing; Consistency verification: Verifying the logical consistency between data; Timeliness verification: Ensure that the timestamps on the data are reasonable; Accuracy verification: Verify the accuracy of the data through sampling; In one embodiment, the construction of the dynamic knowledge graph includes: Extracting the temporal dynamic features of numerical time-series data, the semantic features of textual data, and the visual spatial features of image data; Based on the test spline tracing domain ontology, entity types and relation types are defined, and relation reasoning and incremental updates are performed through graph neural networks.

[0047] A multi-scale temporal feature extraction framework is employed to perform feature engineering on the physical response data (such as strain, temperature, and load) of test splines. First, the temporal data is segmented into fixed-length segments using a sliding window mechanism, and then the following features are extracted: Time-domain characteristics: mean, variance, peak value, zero-crossing rate, kurtosis, skewness Frequency domain characteristics: The dominant frequency component and frequency band energy distribution are extracted using Fast Fourier Transform (FFT). Time-frequency characteristics: Wavelet transform (such as Morlet wavelet) is used to obtain the time-frequency energy distribution at different scales. Nonlinear characteristics: approximate entropy, sample entropy, and Lyapunov exponent are used to characterize the chaotic properties of a system. Semantic feature extraction from textual data: Entity recognition: Using the BERT-BiLSTM-CRF model to identify named entities such as test splines, equipment, and personnel; Relation extraction: Extract semantic relations between entities based on SpanBERT or T5 models (e.g., "test spline - located at - test equipment"). Semantic representation: Obtain the semantic vector representation of the sentence through Sentence-BERT, and support similarity calculation.

[0048] Visual spatial feature extraction from image-based data: Basic feature extraction: High-level semantic features of images are extracted using pre-trained models such as ResNet; Spatial feature enhancement: Enhance the feature weights of key regions through an attention mechanism; Temporal modeling: Use 3D-CNN or ConvLSTM to model the spatiotemporal dynamic features of image sequences.

[0049] Furthermore, a semantic model for the test spline tracing domain is defined, including core entity class definitions, object attribute definitions, and data attribute definitions. Then, based on the domain ontology, entity types and relation types in the knowledge graph are defined. Graph Neural Networks (GNNs) are used for relation reasoning and knowledge completion. Entities and relations are embedded into a low-dimensional vector space using GraphSAGE, GAT (Graph Attention Network), or RGCN (Relational Graph Convolutional Network). Missing relation triples are predicted based on knowledge graph embedding models such as TransE, RotatE, or ComplEx. The graph structure is reconstructed using Graph Autoencoders (GAEs) or Graph Variational Autoencoders (VGAEs) to detect anomalous edges (i.e., anomalous relations). Incremental Graph Neural Networks (such as Incremental GNNs) are used to support the incremental addition of new nodes and edges, avoiding full retraining. Dynamic Graph Neural Networks (such as EvolveGCN and TGN) are used to model the temporal evolution of the knowledge graph, capturing the dynamic changes of entities and relations. When new data conflicts with existing knowledge, confidence weighting or time decay mechanisms are used to resolve the conflict, prioritizing the trust of the latest data. It searches for paths connecting entities in the knowledge graph and predicts missing relationships; it learns logical rules from the knowledge graph through rule mining algorithms such as AMIE+; and it supports multi-hop relationship reasoning to discover implicit associations.

[0050] S40. Based on the knowledge-enhanced tracing dataset, execution deviation value and watermark feedback data, generate attack source coordinate information and tracing path map, and push an tamper-proof tracing report.

[0051] First, the three types of data are spatiotemporally aligned, and a timestamp synchronization mechanism is used to ensure data consistency across the time dimension. Then, the following key features are extracted: Knowledge graph features: Extracting vector representations of entity nodes (such as attackers, vulnerabilities, and assets) and semantic features of relationships between entities from knowledge-enhanced tracing datasets; Execution deviation characteristics: Extract statistical features such as timing anomaly patterns, deviation magnitude, and deviation duration from execution deviation values; Watermark feedback features: Extract features such as watermark integrity verification results, watermark correlation, and watermark propagation path from watermark feedback data.

[0052] A graph neural network-based attack source localization model is employed, fusing multi-source features into a unified feature vector. A graph attention network (GAT) is used to weighted aggregate nodes in the knowledge graph, capturing the propagation path features between the attack source and the target. Simultaneously, a temporal convolutional network (TCN) is introduced to model the execution deviation time series, identifying the starting point of abnormal propagation. Finally, a classifier outputs the confidence score of the attack source node, locating the coordinates of the attack source.

[0053] Based on entity relationships and temporal information in the knowledge graph, a Path Ranking Algorithm (PRA) is used to reconstruct attack paths. Starting from the attack source node, a breadth-first search is performed along the relationship edges in the knowledge graph to identify all possible attack propagation paths. The Dynamic Time Warping (DTW) algorithm is used to calculate path similarity, merging similar paths to form a set of attack paths. The attack paths are then converted into a graph structure, where nodes represent attack steps (such as exploitation, privilege escalation, and lateral movement), and edges represent dependencies between attack steps. A force-directed layout algorithm is used to visualize the graph layout, using color coding to distinguish different attack stages (such as initial access, persistence, lateral movement, and data theft). Simultaneously, key timestamps, attack tools, and exploit vulnerabilities are labeled on the paths to form a complete source tracing path graph.

[0054] The source tracing report uses the standardized JSON-LD format and includes the following core modules: Attack event summary: attack time, attack type, scope of impact, attacker information Attack path details: A complete sequence of attack steps, including timestamps, exploits, attack tools, and target assets for each step. Chain of evidence: Hash values ​​of multi-source data (knowledge graph, execution bias, watermark feedback) used to verify report integrity. Impact assessment: List of affected assets, amount of data breach, duration of business disruption Recommendations: Vulnerability remediation suggestions, security hardening measures, and incident response suggestions The source tracing report generates a unique hash value using a hash algorithm (SHA-256) and writes the hash value to the blockchain. A consortium blockchain architecture is adopted, with multiple trusted nodes jointly maintaining the evidence storage record. Evidence storage metadata includes a PUF identifier, report hash value, timestamp, and evidence storage address. Through the blockchain's distributed ledger technology and consensus mechanism, it is ensured that the source tracing report is immutable and non-repudiable once it is on the chain. The source tracing report is pushed to relevant parties (such as the security operations center, emergency response team, and management), and a verification interface is provided. Verifiers can verify the report's integrity in the following ways: recalculate the report hash value and compare it with the hash value stored on the chain; query the evidence storage record through a blockchain explorer to verify the timestamp and evidence storage address; and call a smart contract to automatically verify the report's authenticity and integrity.

[0055] This significantly improves the accuracy of attack source location and greatly shortens the time required to reconstruct the tracing path. It enables rapid location and end-to-end tracing of attack events, providing a complete chain of attack evidence to support judicial evidence collection and accountability; and through tamper-proof tracing reports, it enhances the credibility and traceability of security incidents, supporting security situational awareness and threat intelligence sharing.

[0056] join Figure 2 In one embodiment, the present invention also provides a blockchain-based test spline identity traceability system, the system comprising: The data acquisition unit 100 is used to collect multimodal data of the entire life cycle of the test spline, extract multimodal features, and perform quality assessment and semantic mapping on the multimodal data to generate a standardized traceability dataset; the test spline is embedded with a PUF chip, which generates a unique identifier based on the physical randomness of the test spline material; The watermark embedding unit 200 is used to generate attack judgment flags and abnormal propagation paths based on the standardized traceability dataset, so as to embed encrypted watermarks in control commands and generate control command streams carrying watermarks; and to perform blockchain notarization on the standardized traceability dataset, writing the PUF identifier, watermark hash value, and data storage address as notarization metadata into the blockchain. The deviation analysis unit 300 is used to construct a dynamic knowledge graph of test splines based on multimodal features, perform anomaly detection and relation reasoning on the knowledge graph, generate a knowledge-enhanced tracing dataset, parse the control command stream carrying watermarks, monitor the deviation value between the execution result and the expected physical behavior, and generate execution deviation value and watermark feedback data. The identity tracing unit 400 is used to generate attack source coordinate information and tracing path map based on knowledge-enhanced tracing dataset, execution deviation value and watermark feedback data, and push an tamper-proof tracing report.

[0057] Preferably, the data acquisition unit 100 is used to perform quality assessment and semantic mapping on multimodal data, including: Extract multi-source heterogeneous data from multimodal data, establish data quality admission rules, and perform spatiotemporal interpolation to repair missing data; Construct a semantic recognition mechanism to extract entity relationships from unstructured data and achieve semantic mapping; Establish a spatiotemporal calibration mechanism to unify multi-source timestamps into standard time and transform spatial coordinates to the standard coordinate system.

[0058] Preferably, the method for generating the encrypted watermark includes: An initial seed value for the watermark is generated based on the attack determination flag, and the watermark sequence is generated iteratively through chaotic mapping. The watermark sequence is divided into independent watermark bits, a perturbable field of the control command is embedded, and a watermark header identifier is added to the reserved bits in the protocol.

[0059] It is understood that the system provided in this embodiment has functions or includes modules that can be used to execute the methods described in the above method embodiments. The specific implementation can be referred to the description of the above method embodiments, and for the sake of brevity, it will not be repeated here.

[0060] The present invention also provides an electronic device including a processor and a memory, the memory being used to store computer program code, the computer program code including computer instructions, wherein when the processor executes the computer instructions, the electronic device performs a method as described in any of the above possible implementations.

[0061] The present invention also provides a computer-readable storage medium storing a computer program, the computer program including program instructions that, when executed by a processor of an electronic device, cause the processor to perform a method as described in any of the above possible implementations.

[0062] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

Claims

1. A blockchain-based test swatch identity traceability method, characterized in that, The method comprises: Collecting full-life-cycle multi-modal data of a test sample, extracting multi-modal features, and performing quality assessment and semantic mapping on the multi-modal data to generate a standardized traceability dataset; the test sample internally embeds a PUF chip, and the PUF chip generates a unique identity based on the physical randomness of the material of the test sample; Based on the standardized traceability dataset, an attack judgment mark and an abnormal conduction path are generated to embed an encrypted watermark in a control instruction, and a control instruction stream carrying the watermark is generated; the standardized traceability dataset is stored in a blockchain, and a PUF identifier, a watermark hash value, and a data storage address are written into the blockchain as storage metadata; According to the multi-modal features, a dynamic knowledge graph of the test sample is constructed, and abnormal detection and relationship reasoning are performed on the knowledge graph to generate a knowledge-enhanced traceability dataset; the control instruction stream carrying the watermark is parsed, the deviation value of the execution result from the expected physical behavior is monitored, and an execution deviation value and watermark feedback data are generated; Based on the knowledge-enhanced traceability dataset, the execution deviation value, and the watermark feedback data, attack source coordinate information and a traceability path graph are generated, and an unalterable traceability report is pushed.

2. The blockchain-based test swathe identity traceability method of claim 1, wherein, The quality assessment and semantic mapping of the multi-modal data comprise: Extracting multi-source heterogeneous data in the multi-modal data, establishing data quality access rules, and performing temporal and spatial interpolation repair on missing data; Constructing a semantic recognition mechanism, performing entity relationship extraction on unstructured data, and realizing semantic mapping; Establishing a time and space calibration mechanism, unifying multi-source timestamps to a standard time, and converting spatial coordinates to a standard coordinate system.

3. The blockchain-based test swathe identity traceability method of claim 1, wherein, The generation method of the encrypted watermark comprises: Generating a watermark initial seed value based on the attack judgment mark, and generating a watermark sequence through chaotic mapping iteration; Dividing the watermark sequence into independent watermark bits by bit, embedding the watermark bits into the controllable field of the control instruction, and adding a watermark header identifier in the protocol reserved bit.

4. The blockchain-based test swathe identity traceability method of claim 1, wherein, The blockchain storage comprises: Calculating the cryptographic hash value of the standardized traceability dataset, and automatically verifying the storage metadata through a smart contract; Through a consensus node, the storage request is verified by PBFT consensus, and after verification, a digital storage certificate is generated and anchored to the blockchain network.

5. The blockchain-based test swathe identity traceability method of claim 1, wherein, The construction of the dynamic knowledge graph comprises: Extracting the time sequence dynamic features of numerical time series data, the semantic features of text data, and the visual spatial features of image data; Based on the entity type and relationship type defined by the test sample traceability domain ontology, relationship reasoning and incremental updating are performed through a graph neural network.

6. A blockchain-based test swatch identity traceability system, characterized in that, The system comprises: A data acquisition unit for collecting full-life-cycle multi-modal data of a test sample, extracting multi-modal features, and performing quality assessment and semantic mapping on the multi-modal data to generate a standardized traceability dataset; the test sample internally embeds a PUF chip, and the PUF chip generates a unique identity based on the physical randomness of the material of the test sample; The watermark embedding unit is configured to generate an attack judgment mark and an abnormal conduction path based on the standardized provenance dataset, embed an encrypted watermark in a control instruction to generate a control instruction stream carrying the watermark, and store the standardized provenance dataset in a blockchain, and write a PUF identifier, a watermark hash value, and a data storage address as storage metadata into the blockchain. The bias analysis unit is configured to construct a test spline dynamic knowledge graph according to the multi-modal features, perform abnormal detection and relationship reasoning on the knowledge graph to generate a knowledge-enhanced provenance dataset, analyze the control instruction stream carrying the watermark, monitor a deviation value of an execution result from an expected physical behavior, and generate an execution deviation value and watermark feedback data. The identity provenance unit is configured to generate attack source coordinate information and a provenance path graph based on the knowledge-enhanced provenance dataset, the execution deviation value, and the watermark feedback data, and push an unalterable provenance report.

7. The blockchain-based test strip identity traceability system of claim 6, wherein, The data acquisition unit is configured to perform quality evaluation and semantic mapping on the multi-modal data, including: extracting multi-source heterogeneous data in the multi-modal data, establishing data quality access rules, and performing temporal and spatial interpolation repair on missing data; constructing a semantic recognition mechanism to extract entity relationships from unstructured data and realize semantic mapping; establishing a time-space calibration mechanism to unify multi-source timestamps into a standard time and convert spatial coordinates to a standard coordinate system.

8. The blockchain-based test swathe identity traceability method of claim 6, wherein, The generation method of the encrypted watermark includes: generating a watermark initial seed value based on the attack judgment mark, generating a watermark sequence through chaotic mapping iteration, and dividing the watermark sequence into independent watermark bits according to a bit, embedding the watermark bits into a controllable field of the control instruction, and adding a watermark header identifier in a protocol reserved bit.

9. An electronic device, comprising: The electronic device includes a processor and a memory, and the memory is configured to store computer program code including computer instructions, and when the processor executes the computer instructions, the electronic device executes the blockchain-based test spline identity tracing method according to any one of claims 1 to 5. The computer readable storage medium stores a computer program including program instructions, and the program instructions, when executed by a processor of an electronic device, cause the processor to execute the blockchain-based test spline identity tracing method according to any one of claims 1 to 5.

10. A computer-readable storage medium, characterized in that, ​

Images