Blockchain-driven data link trusted evidence method and system

Abstract

The application relates to a blockchain-driven data link credible evidence storage method and system, and relates to the field of blockchains. The historical malicious coefficient of a target node is obtained when data link information is uploaded to the target node, a malicious behavior rate prediction coefficient is configured, a prediction branch is selected by using the data interaction degree of the target node and other nodes and branch training data similarity to predict the malicious behavior rate, the number of consensus nodes is dynamically configured according to the malicious coefficient and the prediction result, the data link information is uploaded to a blockchain, and the consensus nodes are used to identify and store evidence when the data link information is changed subsequently. The technical problems of a fixed blockchain consensus mechanism, difficulty in adapting to different node malicious behaviors and low accuracy of malicious behavior prediction are solved, the safety and credibility of data link information evidence storage are effectively improved, and dynamic identification and prevention of malicious behaviors are realized.

Description

Technical Field

[0001] This invention relates to the field of blockchain, and in particular to a blockchain-driven method and system for trusted data link storage. Background Technology

[0002] In today's digital age, data has become an indispensable asset for businesses and individuals. With the rapid development of information technology, data faces increasing security challenges during transmission, storage, and use. As a critical path for data flow, the integrity and reliability of the data link directly affect the authenticity and validity of the data. However, traditional data link management methods often suffer from problems such as centralization, susceptibility to tampering, and lack of transparency, making it difficult to meet the growing demands for data security.

[0003] Blockchain technology, as an emerging distributed ledger technology, offers a new solution for secure data link management due to its decentralized, immutable, transparent, and traceable characteristics. Blockchain technology enables distributed storage and consensus verification of data link information, ensuring the integrity and trustworthiness of data during transmission and storage. However, existing blockchain applications still have some shortcomings in trusted data link storage. On the one hand, blockchain consensus mechanisms often use a fixed number of consensus nodes, making it difficult to adapt to malicious behavior from different nodes. In practical applications, the security and trustworthiness of different nodes vary, and a fixed number of consensus nodes may lead to insufficient system security or wasted resources. On the other hand, existing blockchain applications have limitations in predicting malicious behavior in data link information. Existing prediction methods are often based on simple rules or statistical models, making it difficult to accurately identify complex and ever-changing malicious behavior patterns. Furthermore, due to the complexity and diversity of data link information, effectively integrating and utilizing multi-source data to improve the accuracy of malicious behavior prediction is also a significant challenge. Summary of the Invention

[0004] This invention addresses the technical problems of fixed consensus mechanisms in existing blockchain technologies, which make it difficult to adapt to malicious behavior from different nodes and have low accuracy in predicting malicious behavior. It provides a blockchain-driven trusted data link evidence storage method and system to solve these problems.

[0005] The technical solution of the present invention to solve the above-mentioned technical problems is as follows:

[0006] In a first aspect, the present invention provides a blockchain-driven method for trusted data link storage, the method comprising: when a target node uploads data link information, obtaining a malicious coefficient of the target node over a historical period; configuring a malicious behavior rate prediction coefficient based on the malicious coefficient; selecting a malicious behavior prediction branch of at least one node based on the data interaction degree between the target node and other nodes and the similarity of branch training data, and performing malicious behavior rate prediction on the data link information to obtain a malicious behavior rate; configuring consensus nodes for the data link information based on the malicious coefficient and the malicious behavior rate to obtain a number of consensus nodes; configuring at least one consensus node based on the number of consensus nodes, uploading the data link information to the data link blockchain, and performing consensus judgment through the at least one consensus node to perform trusted data link storage when the data link information is subsequently modified, wherein the consensus node configuration is based on the data interaction degree between the target node and other nodes and the relevance of the data link information to other nodes.

[0007] Secondly, this invention provides a blockchain-driven trusted data link notarization system, comprising: a coefficient acquisition module, used to acquire the malicious coefficient of the target node over a historical period when the target node uploads data link information; a behavior prediction module, used to configure a malicious behavior rate prediction coefficient based on the malicious coefficient, select at least one malicious behavior prediction branch of the target node based on the data interaction degree between the target node and other nodes and the similarity of branch training data, and predict the malicious behavior rate of the data link information to obtain a malicious behavior rate; a consensus configuration module, used to configure consensus nodes for the data link information based on the malicious coefficient and the malicious behavior rate to obtain a number of consensus nodes; and a trusted notarization module, used to configure at least one consensus node based on the number of consensus nodes, upload the data link information to the data link blockchain, and perform consensus judgment through the at least one consensus node when the data link information is subsequently modified, thereby performing trusted notarization of the data link, wherein the consensus node configuration is based on the data interaction degree between the target node and other nodes and the correlation between the data link information and other nodes.

[0008] The beneficial effects of this invention are as follows: by obtaining the historical malicious coefficient of the target node when uploading data link information, the malicious behavior rate prediction coefficient is configured, and the prediction branch is selected by using the data interaction degree between the target node and other nodes and the similarity of the branch training data to predict the malicious behavior rate. Finally, the number of consensus nodes is dynamically configured according to the malicious coefficient and the prediction result, the data link information is uploaded to the blockchain, and the consensus nodes are used to judge and store evidence when changes are made in the future. This effectively improves the security and credibility of data link information storage and realizes the dynamic identification and prevention of malicious behavior. Attached Figure Description

[0009] Figure 1 This is a flowchart illustrating a blockchain-driven trusted data link evidence storage method provided by the present invention.

[0010] Figure 2 This is a schematic diagram of the structure of a blockchain-driven trusted data link evidence storage system provided by the present invention.

[0011] Figure labeling: Coefficient acquisition module 11, Behavior prediction module 12, Consensus configuration module 13, Trusted evidence storage module 14. Detailed Implementation

[0012] 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.

[0013] In the description of this invention, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the stated features. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0014] In the description of this invention, the term "for example" is used to mean "used as an example, illustration, or description." Any embodiment described as "for example" in this invention is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use the invention. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that the invention can be made without using these specific details. In other instances, well-known structures and processes will not be described in detail to avoid obscuring the description of the invention with unnecessary detail. Therefore, the invention is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed herein.

[0015] Example 1:

[0016] like Figure 1 As shown, this embodiment of the invention provides a blockchain-driven trusted data link notarization method, the method comprising:

[0017] S10: When the target node uploads data link information, obtain the malicious coefficient of the target node within a historical time period.

[0018] For example, in the critical step of the target node planning to upload data link information, the analysis process of the target node's historical behavior is first initiated to accurately obtain the malicious coefficient of the node in the historical time period. This coefficient, as a quantitative indicator, directly reflects the potential degree of malicious data tampering by the target node.

[0019] Specifically, this involves retrieving all data interaction records of the target node over a past period, including but not limited to data transmission, reception, and editing operations, with particular attention paid to the number of potential Byzantine actions (i.e., malicious tampering, forgery, or denial-of-service attacks). By comparing the number of Byzantine actions of the target node with the maximum number of Byzantine actions recorded in the entire blockchain network, a malice coefficient value between 0 and 1 can be calculated.

[0020] For example, if a target node has performed 5 Byzantine actions in the past month, while the maximum number of Byzantine actions recorded by the entire network in the same time period is 100, then the maliciousness coefficient of the target node is 0.05, indicating a relatively low risk of malicious tampering. Conversely, if the maliciousness coefficient is close to 1, it means that the node has a high probability of malicious tampering and requires high vigilance. Obtaining this maliciousness coefficient provides an important data foundation for subsequent malicious behavior rate prediction and consensus node configuration, helping the system to dynamically adjust security strategies and ensure the reliable storage of data link information.

[0021] S20: Based on the malicious coefficient, configure the malicious behavior rate prediction coefficient, select at least one malicious behavior prediction branch of the target node and other nodes and the similarity of the branch training data, and perform malicious behavior rate prediction on the data link information to obtain the malicious behavior rate.

[0022] Furthermore, after obtaining the maliciousness coefficient of the target node, the process of predicting the malicious behavior rate begins. First, based on the maliciousness coefficient, the system dynamically configures a malicious behavior rate prediction coefficient. This coefficient will serve as a key basis for subsequently selecting the number of malicious behavior prediction branches. In other words, the maliciousness coefficient can be directly used as the malicious behavior rate prediction coefficient, because the maliciousness coefficient itself is a quantifiable indicator that directly relates to the target node's past behavior patterns. Therefore, using it as a prediction coefficient can more intuitively reflect the target node's potential future malicious behavior tendencies. Subsequently, based on the data transmission and interaction between the target node and other nodes, as well as the similarity of the branch training data, at least one node's branch is selected from the pre-trained malicious behavior prediction branches for prediction.

[0023] Specifically, considering the frequency of data transmission between the target node and other nodes, the data transmission records between the target node and other nodes in the blockchain network over a period of time include transmission frequency, transmission volume, and transmission direction (sending or receiving). These interactions reflect the degree of correlation and dependence between the target node and other nodes. The system then assesses the closeness of the target node with other nodes based on these interactions, tending to select nodes with less interaction or more independent interaction patterns for prediction. This is because nodes with less interaction may have a more independent perspective and less data bias, thus providing more objective prediction results.

[0024] Simultaneously, the similarity between the training data of the branches and the data of the target nodes is evaluated; that is, the degree of similarity between the training data used by the pre-trained malicious behavior prediction branches and the current data link information of the target node. Pre-trained branches refer to malicious behavior prediction branches pre-trained on multiple nodes in the blockchain network. Each branch is based on a specific dataset and training algorithm for that node. These branches can learn malicious behavior patterns on their respective nodes and use them to predict new data. Similarity can be evaluated by comparing data features, data distribution, and data sources. Branches with high similarity are prioritized because high similarity means the prediction model is closer to the actual situation of the target node, thereby improving prediction accuracy.

[0025] Based on the aforementioned data transmission interactions and the similarity of branch training data, the system selects at least one (possibly multiple, but at least one) branch from multiple pre-trained branches to predict the malicious behavior rate of the target node's data link information. By comprehensively considering the sparsity of data transmission interactions and the similarity of branch training data, the system can select branches with high confidence from numerous node branches, and then integrate these branches to predict the malicious behavior rate of the target node's data link information, thus deriving the malicious behavior rate. For example, if the target node interacts with node A relatively few times, and the branch training data of node A is highly similar to the target node's data, the system may prioritize selecting the branch of node A for prediction to ensure the reliability and accuracy of the prediction results.

[0026] S30: Configure consensus nodes for the data link information based on the malicious coefficient and malicious behavior rate to obtain the number of consensus nodes.

[0027] In detail, after obtaining the malicious coefficient of the target node and the predicted malicious behavior rate, the consensus node configuration process for data link information is initiated. This process aims to determine an appropriate number of consensus nodes to make consensus judgments when the data link information is modified in the future, thereby effectively preventing the target node from tampering with the data.

[0028] First, based on the malice coefficient and malicious behavior rate, a comprehensive malice rate is calculated using specific algorithms (such as weighted summation, product operations, etc.). This comprehensive malice rate serves as a quantitative indicator of the overall malicious tendency of the target node. Then, the number of consensus nodes is dynamically determined based on the comprehensive malice rate and the total number of other nodes in the blockchain network. During this process, a balance must be struck between security and efficiency. Theoretically, a higher number of consensus nodes provides higher security because more nodes need to reach a consensus before data can be modified, thus reducing the risk of single-point tampering. However, too many consensus nodes can also lead to system redundancy, increasing communication and computational overhead and reducing overall operational efficiency. For example, if the comprehensive malice rate is low, indicating that the target node's historical behavior is relatively trustworthy, fewer consensus nodes may be configured, such as three, to maintain efficient system operation while ensuring a certain level of security. Conversely, if the comprehensive malice rate is high, the number of consensus nodes will be increased, such as to seven, to enhance the supervision and verification of data modifications and ensure the immutability of data link information. Through this process, the number of consensus nodes can be flexibly adjusted according to the actual malicious tendency of the target node, achieving a balance between security and efficiency.

[0029] S40: Based on the number of consensus nodes, configure at least one consensus node to upload the data link information to the data link blockchain. When the data link information is subsequently changed, consensus judgment is performed through the at least one consensus node to perform trusted data link storage. The consensus node configuration is based on the data interaction degree between the target node and other nodes and the correlation between the data link information and other nodes.

[0030] Specifically, after determining the number of consensus nodes, a refined configuration process for consensus nodes is carried out based on the data interaction degree between the target node and other nodes, as well as the relevance of the data link information uploaded by the target node to other nodes, to ensure the reliable storage of data links. Data interaction degree includes dimensions such as interaction frequency, interaction data volume, and interaction direction. By analyzing these dimensions, the degree of connection between the target node and other nodes can be quantified. The relevance of data link information to other nodes is measured by indicators such as data feature comparison, business logic association, and the number of shared data fields. The higher the relevance, the closer the connection between the data link information and the data of other nodes.

[0031] Based on the analysis results from the above two dimensions, nodes that interact with the target node too frequently in terms of data transmission, have conflicts of interest, or have low relevance to the data link information (which may indicate a lack of data sensitivity or motivation for oversight) are excluded. Consensus nodes are prioritized from those with moderate data interaction, maintain independence, and have high relevance to the data link information. For example, if the target node interacts with node A dozens of times per day in the past week, and the data link information is not directly related to node A's business, then node A is excluded from the consensus node selection; while node B interacts with the target node at a reasonable frequency, and its business is highly correlated with the core fields in the target node's data link information, then node B is included in the consensus node candidate range.

[0032] After consensus node configuration is complete, the data link information uploaded by the target node is stored in a specific block of the blockchain to ensure that the data is immutable and traceable. When subsequent changes to the data link information are required, a consensus verification mechanism is triggered. The configured consensus nodes verify the change request according to a preset consensus algorithm (such as variants of PBFT and PoS, the specific algorithm depending on the blockchain network design). The consensus nodes ensure that the changes comply with the blockchain network rules by checking the legality of the change request, data consistency, and permission compliance. For example, if the change request involves modifying key business fields in the data link information, the consensus node will verify whether the party making the change has the corresponding permissions and compare whether the data logic before and after the change is consistent.

[0033] A change request is considered valid and the data link information is updated only when more than a certain percentage (such as two-thirds or one-half, the specific percentage depending on the consensus algorithm) of the consensus nodes agree on it. If no consensus is reached, the change request will be rejected, and the data link information will remain unchanged. This process effectively prevents nodes close to the target node from colluding to illegally tamper with data, ensures the credible preservation of data link information, and maintains the data security and credibility of the blockchain network.

[0034] In a preferred embodiment, when a target node uploads data link information, obtaining the maliciousness coefficient of the target node over a historical period includes: obtaining data link information when the target node uploads data link information, wherein the data link information includes data ontology, data transmission node sequence, and data editing content sequence; obtaining the number of Byzantine actions of the target node over a historical period; extracting and calculating the maximum number of Byzantine actions among multiple nodes in the data link blockchain; and calculating the ratio of the number of Byzantine actions to the maximum number of Byzantine actions to obtain the maliciousness coefficient.

[0035] Specifically, when the target node initiates a data link information upload request, the data link information uploaded by the target node is first parsed and extracted. This data link information includes the data ontology (i.e., the core data content to be stored, such as transaction records, business vouchers, etc.), the data transmission node sequence (recording the identifiers of each node the data passes through from the source to the target node; for example, in a supply chain finance scenario, the data may pass through core enterprise nodes, logistics enterprise nodes, financial institution nodes, etc. in sequence), and the data editing content sequence (including records of editing operations that the data may undergo during transmission, such as data format conversion, encryption of key fields, etc.).

[0036] Simultaneously, in-depth analysis is conducted on the target node's historical behavior records to count its Byzantine behavior. Byzantine behavior specifically refers to actions by a node in the blockchain network that violate network protocols and rules, such as maliciously tampering with data, forging identities, or refusing to respond to consensus requests. For example, during past data transmissions, the target node may have arbitrarily altered the transaction amount field in the data ontology or forged data transmission node sequences to conceal illegal operation paths. Such behaviors are all counted in the Byzantine behavior count.

[0037] To quantify the relative level of malice of a target node within the entire blockchain network, historical behavior data from multiple nodes within the data link blockchain is further extracted. Through comprehensive analysis and comparison, the maximum number of Byzantine actions among all nodes is determined. For example, in a blockchain network containing one hundred nodes, if statistics show that a certain node has accumulated twenty Byzantine actions in the past month, which is the highest value in the entire network, then these twenty actions are identified as the maximum number of Byzantine actions.

[0038] Finally, the ratio of the number of Byzantine actions of the target node to the maximum number of Byzantine actions within the data link blockchain is calculated. The result is the maliciousness coefficient of the target node. This maliciousness coefficient ranges from 0 to 1. The closer the coefficient is to 1, the more significant the malicious tendency of the target node in its historical behavior, and the higher the risk of its uploaded data link information being tampered with or forged. Conversely, if the coefficient is close to 0, it indicates that the target node's historical behavior is relatively normal, and the data credibility is relatively high. For example, if the target node has 3 historical Byzantine actions, and the maximum number of Byzantine actions in the network is 15, the calculated maliciousness coefficient is 0.2. The system can further take corresponding risk control measures based on this coefficient value, such as strengthening data verification and increasing the number of consensus nodes, to ensure the security and reliability of data link information storage.

[0039] In a preferred embodiment, based on the malicious coefficient, a malicious behavior rate prediction coefficient is configured. Based on the data interaction degree between the target node and other nodes and the similarity of branch training data, at least one malicious behavior prediction branch of a node is selected. Malicious behavior rate prediction is then performed on the data link information to obtain the malicious behavior rate. This includes: using the malicious coefficient as the malicious behavior rate prediction coefficient; calling multiple pre-trained malicious behavior prediction branches of other nodes; calculating multiple malicious prediction confidence levels between the target node and other nodes based on the data interaction degree between the target node and other nodes and the similarity of branch training data; calculating the number of malicious prediction branches based on the malicious behavior rate prediction coefficient and the total number of malicious behavior prediction branches, wherein the number of malicious prediction branches is greater than or equal to 1 and less than or equal to the total number; sorting the multiple malicious prediction confidence levels from high to low, selecting a malicious behavior prediction branch with the same number of malicious prediction branches based on the sorting result, inputting the data link information respectively, and predicting and outputting multiple predicted malicious behavior results, wherein each predicted malicious behavior result includes yes or no; calculating the proportion of yes predicted malicious behavior results among the multiple predicted malicious behavior results to obtain the malicious behavior rate.

[0040] Preferably, after obtaining the malicious coefficient of the target node, this malicious coefficient is directly used as the malicious behavior rate prediction coefficient. This coefficient provides a key quantitative basis for subsequent screening of malicious behavior prediction branches and calculation of the malicious behavior rate. Subsequently, multiple malicious behavior prediction branches corresponding to other nodes in the pre-trained blockchain network are invoked. These branches are trained based on the historical data and behavioral patterns of different nodes and have the ability to differentiate malicious behavior patterns.

[0041] To assess the correlation between the target node and other nodes, malicious prediction confidence is calculated using two dimensions: data interaction degree and branch training data similarity. Data interaction degree encompasses indicators such as interaction frequency, data transmission volume, and interaction direction. For example, if the target node and node A experience high-frequency, high-volume data transmission within a specific time period, their data interaction degree is high. Branch training data similarity is measured using methods such as feature comparison and distribution fitting. For instance, if the branch training data of node B is highly consistent with the current data link information of the target node in terms of field structure and numerical range, the similarity is high. Based on the above analysis, the malicious prediction confidence of each node and the target node is comprehensively calculated, quantitatively reflecting the predictive reliability of different node branches for the target node's data link information.

[0042] When determining the number of malicious prediction branches, a dynamic adjustment strategy is adopted. This involves multiplying the malicious behavior rate prediction coefficient by the total number of pre-trained malicious behavior prediction branches, and rounding the product to the nearest integer. The resulting value is the actual number of malicious prediction branches used for prediction, strictly limited to a range from 1 to the total number. For example, if the malicious behavior rate prediction coefficient is 0.3 and the total number of pre-trained branches is 10, then the calculated number of malicious prediction branches is 3 (0.3 × 10 = 3).

[0043] Furthermore, the system sorts multiple malicious prediction confidence levels from high to low. Based on this ranking, it prioritizes malicious behavior prediction branches with the same number of branches as the number of malicious prediction branches. For example, if calculations show that three branches are needed, then all three branches are selected as the prediction subjects. The system inputs the target node's data link information into the selected branches, and each branch independently outputs a prediction result based on its own training logic. Each result is presented in a binary classification format (i.e., "yes" or "no," representing the presence or absence of malicious behavior tendencies, respectively). Finally, the system counts the number of "yes" results in all predictions and calculates the proportion of these results to the total number of predictions. This proportion represents the malicious behavior rate of the target node's data link information. For example, if two out of three branches output "yes," the malicious behavior rate is 66.7% (2 / 3). Through this process, combining the malicious coefficient, branch confidence, and dynamic branch selection mechanism, a quantitative assessment of the malicious behavior risk of the target node's data link information is achieved, providing decision support for subsequent consensus node configuration and data notarization.

[0044] In a preferred embodiment, the training steps of the multiple malicious behavior prediction branches include: collecting multiple sets of sample data link information in multiple nodes, and labeling each sample data link information according to whether subsequent malicious behavior occurs, to obtain multiple sets of sample malicious behavior results; constructing multiple malicious behavior prediction branches using machine learning; and conducting supervised training and testing on the multiple sets of sample data link information and multiple sets of sample malicious behavior results, with training completed after the test accuracy converges.

[0045] Furthermore, in constructing the malicious behavior prediction mechanism, sample data collection is first carried out on multiple nodes within the blockchain network. Each node, based on its own business scenario and data storage characteristics, collects multiple sets of sample data link information, covering dimensions such as data ontology, transmission path, and operation records. For example, in a supply chain finance scenario, sample data link information may include cargo transportation trajectory, fund transfer records, and corresponding business operation logs. Subsequently, each sample data link information is manually labeled or automatically labeled based on historical event backtracking to determine whether it will subsequently trigger malicious behavior. For example, if a transaction associated with a certain data link information is ultimately proven to be fraudulent, it is labeled as a malicious behavior sample; otherwise, it is labeled as a normal behavior sample. This forms multiple sets of sample malicious behavior results, each corresponding one-to-one with a set of sample data link information.

[0046] Based on the labeled sample data, machine learning techniques are used to construct multiple malicious behavior prediction branches. Specifically, each node selects an appropriate machine learning algorithm (such as decision tree, random forest, neural network, etc.) to build its prediction model architecture according to its own data characteristics and business needs. For example, nodes with complex data features and non-linear relationships may choose a deep neural network model, while nodes with lower data dimensionality and clear logical relationships may use a decision tree model, thus forming multiple malicious behavior prediction branch frameworks with different characteristics.

[0047] During the model training phase, the set of sample data link information is used as input features, and the set of sample malicious behavior results is used as supervision labels to conduct supervised training on multiple malicious behavior prediction branches. During training, each branch iteratively optimizes its own parameters (such as the weights and biases of the neural network, the splitting threshold of the decision tree, etc.) to gradually learn the potential correlation patterns between data link information and malicious behavior. For example, in the neural network model, the backpropagation algorithm is used to adjust the connection weights between network layers, gradually reducing the error between the model output and the true label; in the decision tree model, the optimal splitting features and thresholds are determined by calculating information gain or the Gini coefficient, constructing a tree structure that can accurately distinguish between malicious and normal samples.

[0048] To ensure the model's generalization ability and prediction accuracy, testing is interspersed throughout the training process, using test data independent of the training set to evaluate the performance of each branch. Test metrics include accuracy, recall, and F1 score. For example, accuracy reflects the proportion of samples correctly predicted by the model, while recall measures the model's ability to detect malicious samples. When the accuracy of each branch on the test set reaches a convergence threshold (e.g., accuracy fluctuation less than 1% over multiple consecutive training cycles), and other performance metrics meet business requirements, training is considered complete. At this point, each branch has the ability to predict malicious behavior based on unknown data link information. Through this process, the system constructs a malicious behavior prediction branch system based on multi-node data features and machine learning technology, providing technical support for subsequent risk assessment of data link information.

[0049] In a preferred embodiment, based on the data interaction degree between the target node and other nodes and the similarity of branch training data, multiple malicious prediction confidence scores between the target node and other nodes are calculated, including: obtaining the number of data transmissions between the target node and other nodes within a historical time period to obtain multiple historical data transmission counts; calculating the ratio of the minimum value among the multiple historical data transmission counts to each historical data transmission count to obtain multiple interaction confidence scores; calculating the similarity of the training data of the malicious behavior prediction branch of the target node with other nodes to obtain multiple branch training data similarities, and calculating multiple training confidence scores; and calculating multiple malicious prediction confidence scores based on the multiple interaction confidence scores and the multiple training confidence scores.

[0050] Specifically, when assessing the reliability of other nodes' predictions of malicious behavior regarding the target node's data link information, the number of data transmissions between the target node and multiple other nodes in the blockchain network over a historical time frame is statistically analyzed, forming multiple sets of historical data transmission counts. For example, in a logistics traceability blockchain scenario, the target node is the core record node for a certain batch of goods. Its data interactions with other nodes (such as supplier nodes, transporter nodes, and warehousing nodes) at different frequencies over the past month, including operations such as updating goods status and synchronizing location information, will be recorded one by one.

[0051] To quantify the interaction strength between each node and the target node and its impact on the confidence level of malicious prediction, the minimum number of historical data transmissions is calculated, and then compared with the historical data transmission count of each node to obtain multiple interaction confidence levels. The logic is that a higher number of interaction transmissions suggests that nodes may form interest-based relationships or operational dependencies due to frequent data exchanges, increasing the potential probability of colluding to commit malicious acts, thus lowering the confidence level of the prediction of malicious behavior by that node. Conversely, fewer interaction transmissions indicate a lower correlation between nodes, a lower risk of collusion, and a relatively higher prediction confidence level. For example, if the target node has 50 historical data transmissions with node A and 10 with node B, and the minimum number of transmissions is 5, then the interaction confidence level for node A is 5 / 50 = 0.1, and for node B it is 5 / 10 = 0.5. Clearly, node B has a higher interaction confidence level, and its prediction result is more reliable.

[0052] Subsequently, focusing on the training data level of the malicious behavior prediction branch, the similarity between the training data used by the malicious behavior prediction branch of the target node and the training data of multiple other node branches is calculated, generating multiple branch training data similarity indicators. Similarity calculation can be achieved through methods such as feature vector cosine similarity and data distribution overlap. For example, if the training data of two node branches are highly consistent in terms of data type, field structure, and numerical distribution, the similarity is high. Based on the positive correlation between similarity and training confidence, the similarity value is directly mapped to the training confidence. That is, the greater the similarity, the closer the source of the training data is to the actual scenario of the target node, the stronger the adaptability of the prediction model to the target node data, and the higher the training confidence. For example, if the similarity between the training data of the target node branch and the training data of node C is 0.8, and with node D it is 0.3, then the training confidence of node C is higher than that of node D.

[0053] Finally, multiple interaction confidence scores and multiple training confidence scores are combined to calculate multiple malicious prediction confidence scores for other nodes and the target node. Specific calculation methods can employ weighted summation, product operations, etc. For example, if the weights of the interaction confidence score and the training confidence score are 0.4 and 0.6 respectively, then the malicious prediction confidence score for node E is (interaction confidence score E × 0.4 + training confidence score E × 0.6). Through this process, combining the interaction correlation between nodes and the adaptability of the training data, the reliability of each node's prediction of malicious behavior of the target node is quantitatively evaluated, providing a basis for subsequent selection of high-confidence prediction branches. For example, if node F has an interaction confidence score of 0.7 and a training confidence score of 0.9, its malicious prediction confidence score after weighted calculation is 0.82, while node G's confidence score is 0.65, then the system prioritizes the branch of node F to participate in the malicious behavior prediction of the target node's data link information.

[0054] In a preferred embodiment, the consensus node configuration of the data link information is performed based on the malicious coefficient and malicious behavior rate to obtain the number of consensus nodes, including: calculating the comprehensive malicious rate based on the malicious coefficient and malicious behavior rate; and calculating the number of consensus nodes based on the comprehensive malicious rate and the number of other nodes, wherein the number of consensus nodes is greater than or equal to 1 and less than or equal to the number of other nodes.

[0055] Optionally, in the process of data link information storage and consensus mechanism construction in the blockchain network, a comprehensive malicious rate is first calculated based on the malicious coefficient and malicious behavior rate of the target node. The malicious coefficient reflects the degree of malicious tendency in the target node's historical behavior. For example, if the target node has repeatedly tampered with data, forged identities, or refused to respond to consensus requests in past transactions, its malicious coefficient value will be high. The malicious behavior rate is derived from the malicious behavior prediction results of the target node's current data link information. For example, if the data link information is evaluated through multiple malicious behavior prediction branches and it is found that 60% of the prediction results point to malicious behavior, then the malicious behavior rate is 60%. To quantify the security protection strength required in the process of storing data link information of the target node, the malicious coefficient and malicious behavior rate are regarded as equally important evaluation indicators, and the comprehensive malicious rate is obtained by averaging, i.e., (malicious coefficient + malicious behavior rate) / 2. For example, if the malicious coefficient of the target node is 0.4 and the malicious behavior rate is 0.6, then the comprehensive malicious rate is (0.4 + 0.6) / 2 = 0.5. This value comprehensively reflects the malicious risk level of the target node in both historical and current dimensions.

[0056] Based on the overall malice rate and the number of other nodes, the number of consensus nodes is further calculated to ensure the security and reliability of data link information storage. Consensus nodes undertake core functions in the blockchain network, such as data verification, consensus achievement, and ledger maintenance. Their number needs to be dynamically adjusted according to network size, node trust levels, and malice risk. The specific calculation logic is as follows: multiply the overall malice rate by the number of other nodes, and round the product to the nearest integer. The resulting value is the number of consensus nodes, strictly limited to between 1 and the number of other nodes. For example, if the overall malice rate is 0.5, and there are 20 other nodes in the network besides the target node, then the calculated number of consensus nodes is 10 (0.5 × 20 = 10). This calculation method ensures that the number of consensus nodes is positively correlated with the malice risk level of the target node; that is, the higher the overall malice rate, the more consensus nodes are required, thus enhancing the multi-party verification and redundancy protection capabilities during data storage.

[0057] This process dynamically configures the number of consensus nodes based on the malicious risk characteristics of the target node and the scale of the network nodes, providing a mechanism to ensure the secure storage and reliable transmission of data link information. For example, in a financial transaction blockchain scenario, if the target node is a high-risk trading party (with a high overall malicious rate), the system will automatically increase the number of consensus nodes to ensure that transaction data is stored under multi-party verification, reducing the risk of ledger tampering or data loss due to malicious behavior.

[0058] In a preferred embodiment, at least one consensus node is configured based on the number of consensus nodes to upload the data link information to the data link blockchain. When the data link information is subsequently modified, consensus judgment is performed through the at least one consensus node to achieve trusted data link notarization. This includes: obtaining the number of data transmissions between the target node and multiple other nodes within a historical period, and calculating multiple interaction confidence levels; determining whether the data link information includes multiple other nodes; if so, assigning a relevant confidence level of 0 to the corresponding node; if not, assigning a relevant confidence level of 1 to the corresponding node; calculating multiple consensus confidence levels based on the multiple interaction confidence levels and multiple relevant confidence levels; selecting the node with the largest number of consensus nodes and the highest consensus confidence level among the other nodes as at least one consensus node; processing the data link information into a data link block and uploading it to the data link blockchain built based on multiple nodes; and performing consensus judgment through the at least one consensus node to achieve trusted data link notarization when the data link information is subsequently modified.

[0059] Specifically, after configuring the number of consensus nodes based on malicious risk assessment, a specific configuration process for consensus nodes is initiated for the target node and multiple other nodes to ensure the credibility and tamper-resistance of data link information storage. First, the number of data transmissions between the target node and multiple other nodes within a historical timeframe is statistically analyzed. The interaction confidence level is then calculated based on the aforementioned logic, quantifying the interaction strength by the ratio of the minimum number of historical data transmissions to the total number of transmissions for each node. For example, if the target node and node X have had 20 data interactions in the past month, and the minimum number of transmissions is 5, then the interaction confidence level for node X is 5 / 20 = 0.25. This value reflects the degree of collusion risk that node X may generate due to frequent interactions.

[0060] Subsequently, the data link information is structurally parsed to determine whether it directly contains the identity identifiers, data signatures, or operation records of other nodes. If the data link information explicitly involves a node (e.g., containing the node's data editing records, transmission path node identifiers, etc.), then that node is considered to have a risk of colluding to tamper with data and is assigned a confidence level of 0. Conversely, if the data link information does not involve a node, it is assigned a confidence level of 1. For example, in a medical data sharing blockchain, if a data link records the patient data processing flow of hospital A, but also contains the equipment operation logs of supplier B, then supplier B is assigned a confidence level of 0 because it directly participates in the data link, while the unmentioned third-party auditing institution C is assigned a confidence level of 1. The logic here is that nodes directly involved in the data link operation are more likely to maliciously tamper with data through data association, while nodes that do not participate are considered independent verifiers.

[0061] Furthermore, based on the interaction confidence and correlation confidence, the consensus confidence of each node is calculated using methods such as weighted summation or product operations. For example, if the interaction confidence weight is 0.6 and the correlation confidence weight is 0.4, then the consensus confidence of node Y is (interaction confidence Y × 0.6 + correlation confidence Y × 0.4). For example, if node Y has an interaction confidence of 0.3 and a correlation confidence of 1, then its consensus confidence is 0.3 × 0.6 + 1 × 0.4 = 0.58. This value comprehensively reflects the trustworthiness of the node in both interaction correlation and data link independence dimensions.

[0062] Among the other nodes, they are sorted in descending order of consensus confidence, and the top N nodes are selected as consensus nodes, where N is the number of consensus nodes determined in the previous steps. For example, if the number of consensus nodes is 5, then the 5 nodes with the highest consensus confidence are selected from all nodes to participate in the subsequent consensus process. This strategy ensures that high-confidence nodes dominate the data storage process and reduces the risk of malicious node interference.

[0063] Subsequently, the data link information of the target node is encapsulated into a data link block. The block header contains metadata such as a timestamp and hash pointer, while the block body records the original data link information and its signature. This block is broadcast to a data link blockchain network built on multiple nodes through consensus nodes. Each consensus node verifies and stores the block content based on consensus algorithms such as PBFT and Raft. When a subsequent request to change the data link information is made, a consensus judgment process is triggered. That is, the change request is first submitted to the consensus node cluster, and each consensus node independently verifies the legality of the change request (such as permission verification, signature verification, etc.) and reaches a consensus through multiple rounds of voting. For example, in a supply chain finance blockchain, if it is necessary to modify the transportation status record of a certain shipment, approval from more than 2 / 3 of the consensus nodes is required before the blockchain ledger can be updated.

[0064] Through this process, combining node interaction history, data link correlation and consensus mechanism, trusted evidence storage and dynamic protection of data link information are achieved, ensuring that the blockchain network can maintain data consistency and immutability even when faced with malicious behavior.

[0065] The blockchain-driven trusted data link evidence storage method provided in this invention has at least the following technical effects:

[0066] 1. By integrating the historical malicious coefficients of target nodes, data interaction levels, and branch training data similarity, the risk of malicious behavior can be quantified, malicious behavior prediction branches and prediction coefficients can be dynamically configured, and the malicious behavior rate can be accurately output. For example, in financial transaction scenarios, high-risk nodes can be identified and early warnings can be issued, reducing the risk of fraudulent transaction evidence storage and ensuring data link security.

[0067] 2. The number of consensus nodes is dynamically determined based on the overall malice rate and the number of nodes. Node selection is optimized by combining interaction confidence and data link relevance to form a highly reliable consensus cluster. For example, in supply chain logistics scenarios, by excluding nodes with the risk of collusion in tampering, the evidence preservation process is ensured to be led by independent and trusted nodes, significantly improving the credibility and tamper resistance of data evidence preservation.

[0068] 3. Covering the entire lifecycle of data link information from upload to subsequent changes, the system uses a consensus node cluster to perform multiple rounds of verification and judgment on evidence storage requests and change operations. For example, in a medical data sharing scenario, when it is necessary to update patient medical record information, the blockchain ledger can only be modified with the approval of the consensus node cluster, ensuring that the data is tamper-proof and the change process is traceable, thus achieving trusted evidence storage and dynamic protection throughout the entire lifecycle of the data link.

[0069] Example 2:

[0070] like Figure 2 As shown, based on the same inventive concept as the blockchain-driven trusted data link evidence storage method provided in Embodiment 1, this embodiment of the invention also provides a blockchain-driven trusted data link evidence storage system, the system comprising:

[0071] The coefficient acquisition module 11 is used to acquire the malicious coefficient of the target node within a historical time period when the target node uploads data link information.

[0072] The behavior prediction module 12 is used to configure the malicious behavior rate prediction coefficient according to the malicious coefficient, select the malicious behavior prediction branch of at least one node according to the data interaction degree between the target node and other nodes and the similarity of the branch training data, and perform malicious behavior rate prediction on the data link information to obtain the malicious behavior rate.

[0073] The consensus configuration module 13 is used to configure the consensus nodes of the data link information according to the malicious coefficient and malicious behavior rate, and to obtain the number of consensus nodes.

[0074] The trusted evidence storage module 14 is used to configure at least one consensus node according to the number of consensus nodes, upload the data link information to the data link blockchain, and perform consensus judgment through the at least one consensus node when the data link information is subsequently changed, so as to perform trusted evidence storage of the data link. The consensus node is configured according to the data interaction degree between the target node and other nodes and the correlation between the data link information and other nodes.

[0075] Furthermore, the coefficient acquisition module 11 is also used to perform the following steps:

[0076] When the target node uploads data link information, the data link information is obtained, including the data ontology, the data transmission node sequence, and the data editing content sequence; the number of Byzantine actions of the target node within a historical time period is obtained; the maximum number of Byzantine actions of multiple nodes within the data link blockchain is extracted and calculated; the ratio of the number of Byzantine actions to the maximum number of Byzantine actions is calculated to obtain the malice coefficient.

[0077] Furthermore, the behavior prediction module 12 is also used to perform the following steps:

[0078] The malicious coefficient is used as the malicious behavior rate prediction coefficient; multiple malicious behavior prediction branches of other pre-trained nodes are invoked; based on the data interaction degree between the target node and other nodes and the similarity of branch training data, multiple malicious prediction confidence scores between the other nodes and the target node are calculated; based on the malicious behavior rate prediction coefficient and the total number of multiple malicious behavior prediction branches, the number of malicious prediction branches is calculated, wherein the number of malicious prediction branches is greater than or equal to 1 and less than or equal to the total number; the multiple malicious prediction confidence scores are sorted from high to low, and a malicious behavior prediction branch with the same number of malicious prediction branches is selected according to the sorting result; the data link information is input respectively, and multiple malicious behavior prediction results are predicted and output, wherein each malicious behavior prediction result includes yes or no; the proportion of the multiple malicious behavior prediction results that are yes is calculated to obtain the malicious behavior rate.

[0079] Furthermore, the behavior prediction module 12 is also used to perform the following steps:

[0080] Within multiple nodes, multiple sets of sample data link information are collected, and each sample data link information is labeled according to whether subsequent malicious behavior occurs, to obtain multiple sets of sample malicious behavior results; machine learning is used to construct multiple malicious behavior prediction branches; the multiple sets of sample data link information and multiple sets of sample malicious behavior results are used to conduct supervised training and testing on the multiple malicious behavior prediction branches, and training is completed after the test accuracy converges.

[0081] Furthermore, the behavior prediction module 12 is also used to perform the following steps:

[0082] The process involves obtaining the number of data transmissions between the target node and multiple other nodes over a historical period, resulting in multiple historical data transmission counts; calculating the ratio of the minimum value among these historical data transmission counts to the count of each historical data transmission count, resulting in multiple interaction confidence scores; calculating the similarity between the training data of the malicious behavior prediction branch of the target node and multiple other nodes, resulting in multiple branch training data similarities, and calculating multiple training confidence scores; and finally, calculating multiple malicious prediction confidence scores based on the multiple interaction confidence scores and the multiple training confidence scores.

[0083] Furthermore, the consensus configuration module 13 is also used to perform the following steps:

[0084] Based on the malicious coefficient and malicious behavior rate, the comprehensive malicious rate is calculated; based on the comprehensive malicious rate and the number of other nodes, the number of consensus nodes is calculated, wherein the number of consensus nodes is greater than or equal to 1 and less than or equal to the number of other nodes.

[0085] Furthermore, the trusted evidence storage module 14 is also used to perform the following steps:

[0086] The process involves: acquiring the number of data transmissions between the target node and multiple other nodes over a historical period; calculating multiple interaction confidence levels; determining whether the data link information includes multiple other nodes; assigning a correlation confidence level of 0 to the corresponding node if yes, and a correlation confidence level of 1 if no; calculating multiple consensus confidence levels based on the multiple interaction confidence levels and multiple correlation confidence levels; selecting the node with the highest number of consensus nodes among the other nodes as at least one consensus node; processing the data link information into a data link block and uploading it to a data link blockchain built based on multiple nodes; and performing consensus judgment through the at least one consensus node when the data link information is subsequently modified, thus ensuring trusted data link storage.

[0087] Through the foregoing detailed description of a blockchain-driven trusted data link evidence storage method, those skilled in the art can clearly understand that this embodiment is a blockchain-driven trusted data link evidence storage system. As the system disclosed in the embodiment corresponds to the method disclosed in the embodiment, the description is relatively simple, and relevant parts can be referred to the method section description.

[0088] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A blockchain-driven trusted evidence storage method for data links, characterized in that, The method includes: When the target node uploads data link information, the malicious coefficient of the target node within a historical time period is obtained; Based on the malicious coefficient, configure the malicious behavior rate prediction coefficient, select at least one malicious behavior prediction branch of the target node and other nodes and the similarity of the branch training data based on the data interaction degree between the target node and other nodes, and perform malicious behavior rate prediction on the data link information to obtain the malicious behavior rate. Based on the malicious coefficient and malicious behavior rate, the consensus node configuration for the data link information is performed to obtain the number of consensus nodes, including: The overall malice rate is calculated based on the aforementioned malice coefficient and malice behavior rate. Based on the overall malice rate and the number of other nodes, the number of consensus nodes is calculated, wherein the number of consensus nodes is greater than or equal to 1 and less than or equal to the number of other nodes. Based on the number of consensus nodes, at least one consensus node is configured to upload the data link information to the data link blockchain. When the data link information is subsequently modified, consensus is determined through the at least one consensus node to perform trusted data link notarization. The consensus node configuration is based on the data interaction degree between the target node and other nodes, and the relevance of the data link information to other nodes, including: The number of data transmissions between the target node and multiple other nodes within a historical time period is obtained, and multiple interaction confidence scores are calculated. Determine whether the data link information includes multiple other nodes. If yes, assign a correlation confidence score of 0 to the corresponding node; otherwise, assign a correlation confidence score of 1 to the corresponding node. Multiple consensus confidence levels are calculated based on multiple interaction confidence levels and multiple correlation confidence levels. Among the other nodes, select the node with the highest consensus confidence as at least one consensus node; The data link information is processed into data link blocks and uploaded to a data link blockchain built on multiple nodes. When the data link information is subsequently modified, consensus is determined through at least one consensus node to perform trusted data link notarization.

2. The blockchain-driven trusted data link evidence storage method according to claim 1, characterized in that, When the target node uploads data link information, the maliciousness coefficient of the target node over a historical period is obtained, including: When uploading data link information at the target node, the data link information is obtained, wherein the data link information includes the data body, the data transmission node sequence, and the data editing content sequence; Obtain the number of Byzantine actions of the target node within a historical time period; Extract and calculate the maximum number of Byzantine actions across multiple nodes within the data link blockchain; The ratio of the number of Byzantine actions to the maximum number of Byzantine actions is calculated to obtain the malice coefficient.

3. The blockchain-driven trusted data link evidence storage method according to claim 1, characterized in that, Based on the maliciousness coefficient, a malicious behavior rate prediction coefficient is configured. Based on the data interaction degree between the target node and other nodes and the similarity of the branch training data, a malicious behavior prediction branch for at least one node is selected. Malicious behavior rate prediction is then performed on the data link information to obtain the malicious behavior rate, including: The maliciousness coefficient is used as a predictor of the malicious behavior rate. Invoke multiple malicious behavior prediction branches from other pre-trained nodes; Based on the data interaction degree between the target node and other nodes and the similarity of branch training data, multiple malicious prediction confidences between the target node and other nodes are calculated. Based on the malicious behavior rate prediction coefficient and the total number of multiple malicious behavior prediction branches, the number of malicious prediction branches is calculated, wherein the number of malicious prediction branches is greater than or equal to 1 and less than or equal to the total number. The multiple malicious prediction confidence levels are sorted from high to low. Based on the sorting results, a malicious behavior prediction branch with the same number of malicious prediction branches is selected. The data link information is input separately, and multiple malicious behavior prediction results are output. Each malicious behavior prediction result includes yes or no. The percentage of predicted malicious behavior results that are true is calculated to obtain the malicious behavior rate.

4. The blockchain-driven trusted data link evidence storage method according to claim 3, characterized in that, The training steps for the multiple malicious behavior prediction branches include: Within multiple nodes, multiple sets of sample data link information are collected, and each sample data link information is labeled according to whether there is subsequent malicious behavior, to obtain multiple sets of sample malicious behavior results. Machine learning is used to construct multiple branches for predicting malicious behavior; Using the multiple sample data link information sets and multiple sample malicious behavior result sets, supervised training and testing are performed on multiple malicious behavior prediction branches. Training is completed after the test accuracy converges.

5. A blockchain-driven trusted data link evidence storage method according to claim 4, characterized in that, Based on the data interaction degree between the target node and other nodes and the similarity of branch training data, multiple malicious prediction confidence scores between the target node and other nodes are calculated, including: Obtain the number of data transmissions between the target node and multiple other nodes within a historical time period, and obtain multiple historical data transmission counts; Calculate the ratio of the minimum value within multiple historical data transmission counts to the number of each historical data transmission count to obtain multiple interaction confidence levels; Calculate the similarity between the training data of the malicious behavior prediction branch of the target node and multiple other nodes, obtain the similarity of the training data of multiple branches, and calculate multiple training confidence scores. Multiple malicious prediction confidences are calculated based on multiple interaction confidences and multiple training confidences.

6. A blockchain-driven trusted data link evidence storage system, characterized in that, The system is used to implement the blockchain-driven trusted data link evidence storage method according to any one of claims 1-5, the system comprising: The coefficient acquisition module is used to acquire the malicious coefficient of the target node over a historical period when the target node uploads data link information. The behavior prediction module is used to configure the malicious behavior rate prediction coefficient according to the malicious coefficient, select the malicious behavior prediction branch of at least one node according to the data interaction degree between the target node and other nodes and the similarity of the branch training data, and perform malicious behavior rate prediction on the data link information to obtain the malicious behavior rate. The consensus configuration module is used to configure the consensus nodes for the data link information based on the malicious coefficient and malicious behavior rate, and to obtain the number of consensus nodes, including: The overall malice rate is calculated based on the aforementioned malice coefficient and malice behavior rate. Based on the overall malice rate and the number of other nodes, the number of consensus nodes is calculated, wherein the number of consensus nodes is greater than or equal to 1 and less than or equal to the number of other nodes. A trusted evidence storage module is used to configure at least one consensus node based on the number of consensus nodes, upload the data link information to the data link blockchain, and perform consensus judgment through the at least one consensus node when the data link information is subsequently modified, thereby performing trusted evidence storage of the data link. The consensus node configuration is based on the data interaction degree between the target node and other nodes, and the relevance of the data link information to other nodes, including: The number of data transmissions between the target node and multiple other nodes within a historical time period is obtained, and multiple interaction confidence scores are calculated. Determine whether the data link information includes multiple other nodes. If yes, assign a correlation confidence score of 0 to the corresponding node; otherwise, assign a correlation confidence score of 1 to the corresponding node. Multiple consensus confidence levels are calculated based on multiple interaction confidence levels and multiple correlation confidence levels. Among the other nodes, select the node with the highest consensus confidence as at least one consensus node; The data link information is processed into data link blocks and uploaded to a data link blockchain built on multiple nodes. When the data link information is subsequently modified, consensus is determined through at least one consensus node to perform trusted data link notarization.

Images