A communication data protection system

Abstract

The present invention relates to the field of data encryption technology, and specifically to a communication data protection system. The system includes: a data receiving and decomposing module that receives communication data, classifies the data into multi-dimensional tensors according to time, protocol type, and traffic, refines the original data structure through feature hierarchical analysis, and generates decomposed data tensors. In the present invention, by refining the communication data, multi-dimensional tensor classification of data features is achieved, the detailed analysis ability of the data structure is improved. Through the real-time update of the dynamic graph model, the system can quickly adapt to changes in data features and the environment, ensure that security policies are synchronized with real threats, dynamically adjust data flow classification and decision-making logic, allow the system to adjust security measures according to real-time analysis, improve the accuracy of decision-making. In addition, the strategy for optimizing the data transmission path is dynamically selected based on actual security and transmission requirements, significantly improving the security and efficiency of data during transmission.

Description

Technical Field

[0001] The present invention relates to the technical field of data encryption, and particularly to a communication data protection system. Background Art

[0002] Data encryption technology is an important part of computer security and information security, aiming to protect the privacy and integrity of data. This technology prevents unauthorized access by converting data into a form that can only be interpreted by authorized users, namely ciphertext. Data encryption can be applied to both data transmission and data storage. Encryption algorithms, such as symmetric key encryption and asymmetric key encryption, as well as hash functions, are key technologies in this field. In symmetric key encryption, the same key is used for both encryption and decryption, while in asymmetric key encryption, a pair of public and private keys is used. Data encryption technology is widely applied in multiple fields such as network communication security, e-commerce, and financial services to ensure the secure transmission of sensitive information.

[0003] Among them, a communication data protection system refers to using data encryption technology to ensure the security and privacy of data during communication. The main purpose of such systems is to prevent data from being intercepted or tampered with during transmission, ensuring that only the intended recipient can access and understand the data content. By encrypting the data during transmission, the communication data protection system can effectively resist hacker attacks and the risk of data leakage, and is widely applied in multiple scenarios such as enterprise internal communication, Internet services, and mobile communication.

[0004] In the prior art, static data encryption methods usually cannot respond in a timely manner to the rapid changes in the network environment, which reduces the effect of data protection under dynamic network conditions. For example, when the network threat pattern is updated or the environment changes, fixed encryption policies may no longer be applicable, increasing the risk of data intrusion. Similarly, traditional data transmission path selection lacks the ability to react in real time to the network state, which may lead to data being transmitted via insecure paths, increasing the possibility of data leakage or tampering. Summary of the Invention

[0005] The purpose of the present invention is to solve the disadvantages existing in the prior art, and to propose a communication data protection system.

[0006] To achieve the above purpose, the present invention adopts the following technical solution: A communication data protection system includes:

[0007] A data receiving and decomposing module receives communication data, classifies the data into multi-dimensional tensors according to time, protocol type, and traffic, refines the original data structure through feature hierarchical analysis, and generates decomposed data tensors;

[0008] The feature extraction and graph construction module extracts key features from the decomposed data tensor, identifies the distribution and density of each feature data, calculates the weights of the edges based on feature interactions, and establishes a preliminary graphical structure. The graph nodes represent multiple features, and the edges reflect the relationships between features, generating an encrypted feature dynamic graph;

[0009] The dynamic graph model update module monitors the node weights in the encrypted feature dynamic graph, dynamically adjusts the nodes and connections according to new data, updates the graph structure to match the current changes in communication features, and generates an updated dynamic graph model;

[0010] The classification and decision-making module uses the updated dynamic graph model to classify data streams, monitors the weight changes between nodes, analyzes the changing trends of data features, adjusts the decision-making logic to match the data features according to the changing trends, and generates real-time classification results;

[0011] The path optimization and data transmission module, based on the real-time classification results, performs data fragmentation, evaluates available transmission paths according to the security and transmission requirements of each piece, and evaluates the security and transmission rate. It dynamically adjusts the path selection according to the evaluation results and generates an optimized transmission path.

[0012] The decomposed data tensor specifically includes time, protocol type, and traffic. The encrypted feature dynamic graph specifically refers to the preliminary graphical structure, graph nodes, and edges. The updated dynamic graph model includes node weights, connections, and graph structure. The real-time classification results specifically include data stream classification, weight changes, and changing trends of data features. The optimized transmission path includes data fragmentation, transmission paths, security evaluation, and transmission rate evaluation.

[0013] As a further solution of the present invention, the steps for obtaining the decomposed data tensor are specifically as follows:

[0014] Receive communication data, perform multi-dimensional tensor classification on the data according to time, protocol type, and traffic, set classification criteria, divide the data according to different dimensions, obtain multiple types of labels and assign categories to data points to obtain a multi-dimensional tensor data set;

[0015] Perform feature hierarchical analysis on the multi-dimensional tensor data set. By analyzing the data features of the three dimensions of time, protocol type, and traffic, extract the key features in each dimension, and perform refinement and normalization processing on the data features to obtain a refined feature data set;

[0016] Based on the refined feature data set, perform clustering analysis. By calculating the distances of data points in multiple dimensions, select the matching clustering center points and classify the data points to generate a multi-dimensional tensor decomposition result and obtain a decomposed data structure;

[0017] Based on the decomposed data structure, according to the relationships between multiple tensor elements, use the formula:

[0018] ;

[0019] Calculate the tensor decomposition value of each data point to generate a decomposed data tensor;

[0020] Wherein, represents the element value of the decomposed data tensor, represents the parameter of the first dimension in the tensor decomposition, represents the parameter of the second dimension, represents the parameter of the third dimension, represents an adjustment factor for adjusting the relationship between the first dimension and the clustered data points, represents a normalization parameter for normalizing the relationship between the second dimension and the clustered data points, is the maximum value of the tensor dimension.

[0021] As a further solution of the present invention, the step of obtaining the preliminary graphic structure is specifically:

[0022] Receive the feature data extracted from the decomposed data tensor, and for each feature, apply statistical analysis methods to calculate its distribution and density, obtain the distribution information of multiple features, and generate feature distribution data;

[0023] Based on the feature distribution data, analyze the mutual relationships between features, calculate the fitness between multiple features, and use the distance metric method to quantify the relationships between features to generate a feature relationship matrix;

[0024] Using the feature relationship matrix, use the formula:

[0025] ;

[0026] Calculate the edge weights between features, and based on the edge weights, construct a preliminary graphic structure;

[0027] Wherein, represents the edge weight between features, is the fitness between features, represents the feature density, represents the feature density, represents the feature correlation degree, represents the feature correlation degree, represents the total number of features, represents the index of the feature, represents the feature and the fitness between representing feature and feature between represents a normalization factor, indicating the relative degree of the difference in fitness between features and among all feature differences.

[0028] As a further aspect of the present invention, the steps for obtaining the encrypted feature dynamic graph are specifically as follows:

[0029] Based on the feature nodes in the preliminary graphic structure, use the edge weight matrix to analyze the relationship between each pair of feature nodes. Based on the node relationship and weight distribution, calculate the importance of each node in the graph to generate a node importance score;

[0030] Based on the node importance score, screen the feature nodes that are most critical to the graph structure, select the connection relationships between the key nodes and the edges to form a subgraph, and encrypt its structure to generate an encrypted feature subgraph;

[0031] Through the encrypted feature subgraph, based on the weights between nodes and the encrypted features of the edges, use the formula:

[0032] ;

[0033] Calculate the node change value and the graphic structure evolution to generate an encrypted feature dynamic graph;

[0034] where represents the encrypted feature dynamic graph, is the edge weight between features, is the fitness between features, represents the feature node density, represents the feature node density, is a small constant used to avoid division by zero errors, represents and the larger value of, represents the difference between the fitness between nodes and 1, represents the total number of all feature nodes in the graph.

[0035] As a further aspect of the present invention, the steps for obtaining the updated dynamic graph model are specifically as follows:

[0036] Based on the encrypted feature dynamic graph structure, obtain multiple nodes and their connection relationships. By dynamically monitoring the changes of nodes and connection changes, analyze the weight changes of each node, and update the weight values of nodes and edges in real time to generate the result of node weight changes;

[0037] According to the result of node weight changes, combined with the criticality and connection degree of nodes, screen the nodes associated with communication feature changes, and re-adjust the edge weights in the graph based on the connection relationships between nodes and edges, update the dynamic changes of nodes and connections, and generate the adjustment result of the graph;

[0038] Based on the adjustment result of the graph, dynamically analyze and calculate the impact of new data on the node connection structure. Based on the correction of weight changes and node criticality, use the formula:

[0039] ;

[0040] Calculate the changes of nodes and edges at time , update the encrypted feature dynamic graph, and generate an updated graph structure model;

[0041] Wherein, is the change of the edge weight between nodes, represents the weight of the edge between node and node at time , represents the weight value of node at time , represents the weight value of node at time , represents the connection degree value between node and node , represents the updated value of the connection degree between node and node , represents the difference in node weights, is the connection degree change value, represents the total number of nodes in the network.

[0042] As a further solution of the present invention, the specific steps for obtaining the real-time classification result are as follows:

[0043] Based on the updated dynamic graph model, monitor the weight changes between each node, analyze the law of node weight fluctuations, and extract the key change points of the data stream according to the feature change trend to generate the result of node weight changes;

[0044] By iteratively analyzing the results of the node weight changes, combining the connection relationships and communication characteristics among multiple nodes in the graph, judging the changing trend of the data flow, identifying the associated characteristic change patterns, and generating the result of the changing trend of the data flow characteristics;

[0045] According to the result of the changing trend of the data flow characteristics, evaluate the influence of multiple characteristics on the decision-making logic, adjust the decision-making logic rules, and according to the changing trend, use the formula:

[0046] ;

[0047] Calculate the influence of the characteristic change of each node on the decision-making logic at time , obtain the adjusted decision-making logic rules and apply them to the data flow classification to generate the real-time classification result;

[0048] Among them, is the influence of the adjusted decision-making logic at time , is the weight of node at time , is the change value of the node characteristic at time , is the target value of the characteristic change, is the state value of the node at time , is the expected state value of the node, represents the total number of nodes, that is, the number of nodes participating in the data flow classification in the graph.

[0049] As a further solution of the present invention, the specific steps for obtaining the optimized transmission path are as follows:

[0050] Based on the real-time classification result, perform sharding of the data flow, divide the data into multiple segments, record the transmission requirements and security parameters of each segment, and combine the demand characteristics of each segment to generate a data sharding scheme;

[0051] According to the data sharding scheme, evaluate the transmission requirements and security of each piece of data, screen the candidate set of available transmission paths by comparing the security differences and transmission requirements between segments, and obtain the preliminary characteristics of the candidate paths to generate a path evaluation set;

[0052] According to the security and transmission requirements of each piece in the data sharding scheme, combined with the candidate path network characteristics and the current load in the path evaluation set, calculate the weight of each path, and perform dynamic selection on the candidate paths, using the formula:

[0053] ;

[0054] Calculate the optimization score of the path, select the optimal path according to the score, and generate an optimized transmission path;

[0055] Among them, is the optimized path, is the security score of the path, is the transmission rate of the path, is the network latency of the path, is the current load of the path, is the transmission requirement of the path.

[0056] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0057] In the present invention, by refining communication data, multi-dimensional tensor classification of data features is realized, and the detailed analysis ability of the data structure is improved. Through the real-time update of the dynamic graph model, the system can quickly adapt to changes in data features and the environment, ensuring that security policies are synchronized with real threats. Dynamically adjust the data flow classification and decision-making logic, allowing the system to adjust security measures according to real-time analysis and improving the accuracy of decision-making. In addition, the strategy for optimizing the data transmission path is dynamically selected based on actual security and transmission requirements, significantly improving the security and efficiency of data during transmission. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 is the system flow chart of the present invention;

[0059] Figure 2 is the flow chart of the steps for obtaining the decomposed data tensor of the present invention;

[0060] Figure 3 is the flow chart of the steps for obtaining the preliminary graphic structure of the present invention;

[0061] Figure 4 is the flow chart of the steps for obtaining the encrypted feature dynamic graph of the present invention;

[0062] Figure 5 is the flow chart of the steps for obtaining the updated dynamic graph model of the present invention;

[0063] Figure 6 is the flow chart of the steps for obtaining the real-time classification result of the present invention;

[0064] Figure 7 is the flow chart of the steps for obtaining the optimized transmission path of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0065] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0066] In the description of the present invention, it should be understood that the orientation or positional relationship indicated by the terms "length", "width", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc. is based on the orientation or positional relationship shown in the drawings. It is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus should not be construed as a limitation to the present invention. In addition, in the description of the present invention, the meaning of "a plurality of" is two or more, unless otherwise specifically defined.

[0067] Embodiment 1: Please refer to Figure 1 , a communication data protection system includes:

[0068] The data reception and decomposition module receives communication data, classifies the data in a multi-dimensional tensor according to time, protocol type, and traffic, refines the original data structure through feature hierarchical analysis, and generates a decomposed data tensor.

[0069] The feature extraction and graph construction module extracts key features from the decomposed data tensor, identifies the distribution and density of each feature data, calculates the weights of the edges according to the feature interactions, and establishes a preliminary graph structure. The graph nodes represent multiple features, and the edges reflect the relationships between the features, generating an encrypted feature dynamic graph.

[0070] The dynamic graph model update module monitors the node weights in the encrypted feature dynamic graph, dynamically adjusts the nodes and connections according to the new data, updates the graph structure to match the current changes in the communication features, and generates an updated dynamic graph model.

[0071] The classification and decision-making module uses the updated dynamic graph model to classify the data stream, monitors the weight changes between the nodes, analyzes the changing trend of the data features, adjusts the decision-making logic to match the data features according to the changing trend, and generates a real-time classification result.

[0072] The path optimization and data transmission module, based on the real-time classification result, performs data fragmentation, evaluates the available transmission paths according to the security and transmission requirements of each piece, and evaluates the security and transmission rate. It dynamically adjusts the path selection according to the evaluation results and generates an optimized transmission path.

[0073] The decomposed data tensor specifically includes time, protocol type, and traffic. The encrypted feature dynamic graph specifically refers to the preliminary graph structure, graph nodes, and edges. The updated dynamic graph model includes node weights, connections, and graph structure. The real-time classification result specifically includes data stream classification, weight changes, and changing trends of data features. The optimized transmission path includes data fragmentation, transmission paths, security evaluation, and transmission rate evaluation.

[0074] Please refer to Figure 2 , the steps for obtaining the decomposed data tensor are specifically as follows:

[0075] Receive communication data, classify the data into multi-dimensional tensors according to time, protocol type, and traffic, set classification criteria, divide the data according to different dimensions, obtain multiple types of labels and assign categories to data points to obtain a multi-dimensional tensor data set;

[0076] By analyzing the characteristics of communication data such as time, protocol type, and traffic one by one, timestamp extraction and classification can be adopted, calculate the data traffic and protocol type in each time period, and perform traffic standardization. First, extract time features from each time period, calculate the distribution of different protocol types, divide the data into different categories according to the protocol type, and perform preprocessing on the traffic data. For traffic data, normalization processing is adopted and missing value filling is performed to ensure the integrity of the traffic data. On this basis, the data in the three dimensions of time, protocol type, and traffic are tensorized to form a multi-dimensional tensor data structure, and category labels are assigned to each data point to obtain a multi-dimensional tensor data set that meets the analysis requirements.

[0077] Perform feature hierarchical analysis on the multi-dimensional tensor data set. By analyzing the data characteristics of the three dimensions of time, protocol type, and traffic, extract the key features in each dimension, and perform refinement and normalization processing on the data features to obtain a refined feature data set;

[0078] When analyzing the data characteristics of the three dimensions of time, protocol type, and traffic, first, it is necessary to extract the timing information of the data from the time dimension, such as the daily, weekly, and monthly fluctuations of the data. The protocol type dimension needs to count the distribution of each protocol and its change trend. The traffic dimension needs to calculate the fluctuation range and fluctuation period of the traffic value for each traffic data point. After extracting these features, the most representative features are screened out through feature selection techniques and normalized processing. For example, perform logarithmic transformation on the traffic value or map all features to the same scale through standardization to obtain a more refined and normalized feature data set.

[0079] Based on the refined feature data set, perform clustering analysis. By calculating the distances of data points in multiple dimensions, select the matching clustering center points and classify the data points to generate the multi-dimensional tensor decomposition result and obtain the decomposed data structure;

[0080] Using the K-means clustering method, the refined feature dataset is input into the clustering algorithm. First, an initial clustering center is selected, and then the Euclidean distance from each data point to the clustering center is calculated. The data points are assigned to the nearest clustering center, and the clustering center is iteratively updated until the algorithm converges. When calculating the distance for each data point, considering the contributions of dimensions such as time, protocol type, and traffic to the distance, the clustering category of each data point is obtained, and the multi-dimensional tensor decomposition of the data is performed according to the category information to form a new data structure, ensuring the representativeness of the dataset in each dimension.

[0081] Based on the decomposed data structure, according to the relationship between multiple tensor elements, the formula:

[0082] ;

[0083] Calculate the tensor decomposition value of each data point to generate a decomposed data tensor;

[0084] Among them, represents the element value of the decomposed data tensor, represents the parameter of the first dimension in the tensor decomposition, represents the parameter of the second dimension, represents the parameter of the third dimension, represents the adjustment factor, which adjusts the relationship between the first dimension and the clustering data points, represents the normalization parameter, which normalizes the relationship between the second dimension and the clustering data points, is the maximum value of the tensor dimension.

[0085] Formula:

[0086] ;

[0087] The benefit of the formula is that by performing a weighted sum of the parameters of different dimensions ( , , ) and the adjustment factors ( , ), the mutual relationship between data dimensions can be fully reflected, increasing the accuracy and reliability of the data decomposition result, thereby enhancing the analysis ability of the multi-dimensional data tensor.

[0088] Assume that , , , , , for a sample data , then:

[0089] ;

[0090] The result shows that by weighted summation of each dimension, the value of the tensor element obtained is 1.298, which reflects the comprehensive decomposition result of the data points in three dimensions and can be used for subsequent analysis and decision-making.

[0091] Please refer to Figure 3 , and the steps for obtaining the preliminary graphic structure are specifically as follows:

[0092] Receive the feature data extracted from the decomposed data tensor. For each feature, apply statistical analysis methods to calculate its distribution and density, obtain the distribution information of multiple features, and generate feature distribution data;

[0093] Receive the feature data extracted from the decomposed data tensor. Based on the data sets of each feature, conduct statistical analysis, calculate the distribution and density of each feature, integrate the numerical values and statistical results of each feature through data aggregation methods, further identify data anomalies using distribution and density information, eliminate the data items that do not meet the expected range, generate a feature data set that meets the requirements, calculate the standard deviation, mean, and skewness of each feature to evaluate its degree of dispersion and central tendency, and use these indicators as the basic data for subsequent analysis to provide support for the next step of constructing the preliminary graphic structure, and output the feature distribution data.

[0094] Based on the feature distribution data, analyze the mutual relationships between features, calculate the goodness of fit between multiple features, and use the distance metric method to quantify the relationships between features to generate a feature relationship matrix;

[0095] Based on the feature distribution data, apply the correlation analysis method to evaluate the mutual relationships between features, select an appropriate similarity metric method, such as using the Pearson correlation coefficient or Euclidean distance to pair features, generate a similarity matrix for the relationship of each pair of features, use the numerical values in the matrix for feature correlation analysis, by calculating the influence of each feature on other features one by one, concretize it into a series of numerical values to describe the dependence relationship between features, calculate the correlation coefficient between each pair of features, and generate a feature relationship matrix.

[0096] Using the feature relationship matrix, adopt the formula:

[0097] ;

[0098] Calculate the edge weights between features, and based on the edge weights, construct a preliminary graphic structure;

[0099] Among them, represents the edge weight between features, is the goodness of fit between features, represents feature 's density, Represents a feature The density of Represents a feature The degree of correlation of Represents a feature The degree of correlation of Represents the total number of features Represents the index of a feature Represents a feature And the feature The goodness of fit between Represents a feature And the feature The goodness of fit between Represents a normalization factor, indicating the relative degree of difference in the goodness of fit between the feature And Among all the differences between features.

[0100] Formula:

[0101] ;

[0102] The advantage of the formula is that it takes into account the similarity, density difference, and correlation difference between features, optimizes the calculation of edge weights, and enables the graph structure to better reflect the true relationship between features.

[0103] Detailed explanation of the formula and the derivation process of formula calculation:

[0104] First, the similarity matrix obtained through the aforementioned steps , feature density And the correlation And other data are successively substituted into the formula for calculation. For each pair of features And , first calculate their similarity , then adjust according to the difference in feature density And , and finally further optimize the result according to the difference in the correlation between features.

[0105] For example, set , , , , , , assuming that the differences between all other features are 0, the formula calculation process is:

[0106] ;

[0107] ;

[0108] ;

[0109] ;

[0110] The result shows that the edge weight between features is 0.982, indicating a high correlation between the two, and a graphical structure is constructed using this edge weight.

[0111] Please refer to Figure 4 , and the specific steps for obtaining the encrypted feature dynamic graph are as follows:

[0112] Based on the feature nodes in the preliminary graphical structure, use the edge weight matrix to analyze the relationship between each pair of feature nodes. Based on the node relationship and weight distribution, calculate the importance of each node in the graph and generate a node importance score.

[0113] Based on the feature nodes in the preliminary graphical structure, first collect the feature data of each node, including attributes such as the category, density, and morphology of the node. Standardize the features of each node through data processing methods to make them conform to a unified quantization standard, and then perform weight assignment. Using a similarity-based calculation method, after numerically processing the similarity between nodes, a preliminary similarity matrix is obtained; then, analyze the relationship between each pair of nodes in this similarity matrix and calculate the weight of the edge for each pair of nodes. These edge weights are calculated from the similarity, distance, and relationship density of the nodes. By further optimizing the edge weights, ensure that the connection strength between nodes can be truly reflected during the calculation process, generate an edge weight matrix and correspond it to the feature nodes one by one to form a graph structure, and obtain the node importance score of each feature node. This score reflects the influence of the node on the graphical structure and provides an important basis for subsequent graph structure analysis.

[0114] Based on the node importance score, screen the feature nodes that are most critical to the graph structure, select the connection relationships between the key nodes and the edges to form a subgraph, and encrypt its structure to generate an encrypted feature subgraph.

[0115] Based on the node importance score, first sort these scores in descending order, and select the top several nodes as the core nodes of the subgraph. These nodes play a key role in the graph structure; then, according to the edge weights between these key nodes and other nodes, calculate the edge set that forms the subgraph to ensure that the relationship between the selected nodes and other nodes is reflected in the graph, and then perform encryption processing. Modify the weight values of some nodes and edges, and encrypt these edges through an encryption algorithm to make the connection information between nodes more secure and difficult to reverse calculate, forming an encrypted feature subgraph. This subgraph contains the most influential nodes and the encrypted edge structure, ensuring that key information and structural changes can be effectively reflected during the subsequent generation process of the feature dynamic graph.

[0116] By encrypting the feature subgraph, based on the weights between nodes and the encrypted features of edges, using the formula:

[0117] ;

[0118] Calculate the node change value and the graph structure evolution to generate an encrypted feature dynamic graph;

[0119] Among them, represents the encrypted feature dynamic graph, is the edge weight between features, is the fitness between features, represents the feature node 's density, represents the feature node 's density, is a small constant used to avoid division-by-zero errors, represents and 's larger value, represents the difference between the fitness between nodes and 1, represents the total number of all feature nodes in the graph.

[0120] Formula:

[0121] ;

[0122] The advantage of the formula is that by combining the weights, density differences, and similarities of nodes, it can effectively reflect the importance and changes of nodes in the graph, and through encryption processing, it ensures that the features of the graph will not leak, thereby improving the security and reliability of the feature graph.

[0123] Assume that in the calculation example, the density of node is , the density of node is , the similarity between nodes is , the edge weight , the constant . After substituting into the formula, the calculation process is as follows:

[0124] ;

[0125] ;

[0126] ;

[0127] ;

[0128] ;

[0129] The result shows that during the generation of the encrypted feature dynamic graph, the changes in the relationships between nodes are reflected as large negative values in the graph, reflecting the encryption and evolution states between nodes.

[0130] Please refer to Figure 5 , and the steps for obtaining the updated dynamic graph model are specifically as follows:

[0131] Based on the structure of the encrypted feature dynamic graph, obtain multiple nodes and their connection relationships. By dynamically monitoring the changes in nodes and connection changes, analyze the weight changes of each node, and update the weight values of nodes and edges in real time to generate the result of node weight changes;

[0132] First, obtain the initial weight of each node, which can be calculated through node attributes such as transmission frequency, connection quality, data traffic, etc. The initial value of the node weight is usually directly related to these attributes, forming a basic weight vector. Then, monitor the changes in the connection status between the node and other nodes, such as connection stability, latency, and bandwidth changes, etc., and incorporate these factors into the change formula of the node weight for dynamic adjustment. In this process, network performance data collected in real time can be introduced, and by comparing historical data and the current network state through algorithms, the weight adjustment value can be calculated, and the node weight can be dynamically corrected according to these changes to ensure that the encrypted feature dynamic graph matches the actual communication environment. The updated node weight change data will be used as the basis for the next analysis, providing a real-time basis for subsequent graph structure adjustment.

[0133] According to the result of node weight changes, combined with the criticality and connectivity of nodes, screen the nodes associated with communication feature changes, and based on the connection relationships between nodes and edges, readjust the edge weights in the graph, update the dynamic changes of nodes and connections, and generate the adjustment result of the graph;

[0134] First, evaluate the importance of nodes by calculating the connectivity of each node, that is, the number of connections and connection strength with other nodes. These connection strengths can be weighted and calculated through multiple factors such as signal strength, transmission quality, data exchange frequency, etc. Then, by comparing the weight of the node with the connectivity, screen out the nodes that have the greatest impact on communication feature changes, and use these nodes as core nodes to further analyze their relationships with other nodes. During this process, the edge weights of the nodes will be readjusted. Combining the connection changes of the nodes, the edge weights can be modified through parameters such as dynamically measured network load and transmission latency to reflect the changes in the current network state. After this series of adjustments, the adjustment result of the graph is generated.

[0135] Based on the adjustment result of the graph, dynamically analyze and calculate the impact of new data on the node connection structure. Based on the correction of weight changes and node criticality, use the formula:

[0136] ;

[0137] Calculate the changes of nodes and edges at time to update the encrypted feature dynamic graph and generate an updated graph structure model;

[0138] Among them, is the change of the edge weight between nodes, represents the edge weight between node and node at time , represents the weight value of node at time , represents the weight value of node at time , represents the connectivity value between node and node , represents the updated value of the connectivity between node and node , represents the difference in node weights, is the change value of connectivity, represents the total number of nodes in the network.

[0139] Formula:

[0140] ;

[0141] The benefit of the formula is that by comprehensively considering the changes in node weights and the relative importance between nodes, the edge weights are dynamically adjusted, so as to ensure that the interaction relationship between nodes can be more accurately reflected in the communication network environment. In particular, by introducing various factors such as connectivity and delay differences, the process of updating the graph structure is optimized, and the adaptability of the graph structure is improved.

[0142] Detailed explanation of the formula and the derivation process of formula calculation:

[0143] First, based on the node weight change results generated in the previous paragraph and the distance between node and and and , calculate the weight change of each edge. The distance is calculated based on the data transmission delay between nodes or other network metrics. Then, calculate the connectivity between nodes and The difference can be obtained by measuring the bandwidth difference, signal quality difference, etc. between nodes. During the calculation process, absolute value and square root operations are used to ensure that the formula can effectively reflect the dynamic relationship between network nodes.

[0144] During the derivation of the formula, specific values can be given to demonstrate the calculation process. For example, assume the initial weights of node and node , the distance between node and node is , the distance between node and node is , the connection degree is , , then there is:

[0145] ;

[0146] ;

[0147] ;

[0148] This result indicates that the edge weight change between node and node is 0.0267. After this calculation, the graph structure will reflect this change during the next update to ensure that the communication characteristics are consistent with the current network state.

[0149] Please refer to Figure 6 . The specific steps for obtaining the real-time classification result are as follows:

[0150] Based on the updated dynamic graph model, monitor the weight change between each node, analyze the pattern of node weight fluctuations, and extract the key change points of the data stream according to the characteristic change trend to generate the node weight change result;

[0151] Based on the updated dynamic graph model, first, it is necessary to monitor and track the node weight changes in the model in real time, collect the relevant data of each node in the model, monitor the fluctuations of the weights between nodes, further identify the weight change trend between nodes through dynamic analysis, and then obtain the change data set of node weights. Further, use graph analysis algorithms to identify the potential change patterns in the data stream, combine the characteristics of relevant data stream models, obtain the weight change pattern of each node, provide data support for subsequent analysis, and obtain the weight change result of the node through calculation and monitoring tools.

[0152] Through iterative analysis of the node weight change results, combined with the connection relationship and communication characteristics between multiple nodes in the graph, judge the data stream change trend, identify the associated characteristic change patterns, and generate the data stream characteristic change trend result;

[0153] By further analyzing the results of the weight changes between nodes, combining the connection relationships between nodes and the corresponding data flow control characteristics, further refining the mutual relationships between nodes, and using clustering analysis methods, graph theory algorithms, etc. to quantitatively describe the connections between nodes, the weight differences of different nodes at a specific time can be accurately calculated. Furthermore, a detailed model of the characteristic change trend between nodes can be derived. Through these analyses, the change trend data for each time period are obtained, and the specific results of the characteristic change trend of the data flow are further formed, providing a basis for decision-making adjustment.

[0154] According to the results of the characteristic change trend of the data flow, evaluate the influence of multiple characteristics on the decision-making logic, adjust the decision-making logic rules, and based on the change trend, use the formula:

[0155] ;

[0156] Calculate the influence of the characteristic change of each node on the decision-making logic at time Obtain the adjusted decision-making logic rules and apply them to the data flow classification to generate real-time classification results;

[0157] Among them, is the adjusted decision-making logic influence at time , is the weight of node at time , is the change value of the node characteristic at time , is the target value of the characteristic change, is the state value of the node at time , is the expected state value of the node, represents the total number of nodes, that is, the number of nodes participating in the data flow classification in the graph.

[0158] Formula:

[0159] ;

[0160] The benefit of the formula is that by combining the node weight , the characteristic change and the target value , the node state and the expected state in the calculation, the decision-making logic can be dynamically adjusted to improve the accuracy and real-time performance of the data flow classification, avoiding the static processing method that only considers the weight in the traditional model.

[0161] Detailed explanation of the formula and the derivation process of the formula calculation:

[0162] 1. Node weight represents the weight value of node i at time , which is calculated by monitoring and calculating the real-time data of the node;

[0163] 2. Feature change The difference from the target value reflects the deviation between the state change of node i at time and the target state. The larger the difference value, the greater the change amplitude;

[0164] 3. State value The difference from the expected state represents the deviation between the current state of node i and the ideal state, which is used to further refine the adjustment strategy and reduce the system error.

[0165] By combining and calculating the above parameters, the decision logic can be predicted and adjusted more precisely, and the change trend of the data stream can be responded to in a timely manner, enhancing the adaptability of the system. The actual operation process of the formula is as follows:

[0166] First, based on the obtained node data, assume , , , , , and substitute them into the formula:

[0167] ;

[0168] This result shows that the decision logic adjustment of node i at time obtains an adjustment value of 0.4444 due to the influence of its weight, feature change and state deviation, which means that the decision logic needs to adjust the behavior of node i according to this weight.

[0169] Please refer to Figure 7 , and the specific steps for obtaining the optimized transmission path are as follows:

[0170] Based on the real-time classification results, the data stream is fragmented, the data is divided into multiple segments, and the transmission requirements and security parameters of each segment are recorded. Combining the demand characteristics of each segment, a data fragmentation scheme is generated;

[0171] Based on the real-time classification results, first collect and organize the sharded data, divide the data stream into multiple segments, and assign corresponding transmission requirements and security requirements to each segment according to its characteristics. Specifically, use the previous data statistics to calculate the average transmission rate and data size of each segment, and perform precise sharding planning to ensure that the transmission time of each segment matches the capacity requirements; according to the transmission requirements of each piece of data, combined with the security indicators of this piece of data, such as encryption strength, reliability of the transmission protocol, etc., perform preliminary allocation to ensure the integrity and availability of the data segments; by analyzing the demand characteristics of each segment, generate and record the optimal allocation plan for each piece of data, and then provide the necessary basis and reference for subsequent path selection and traffic control, forming the allocation plan for each data segment, and this result is convenient for subsequent execution in the matching of the path evaluation set and the transmission requirements.

[0172] According to the data sharding scheme, evaluate the transmission requirements and security of each piece of data. By comparing the security differences and transmission requirements between segments, filter the candidate set of available transmission paths, and obtain the preliminary characteristics of the candidate paths to generate the path evaluation set;

[0173] According to the data sharding scheme, evaluate the transmission requirements and its security of each piece of data. First, conduct a transmission requirement analysis on each piece of data, including calculating the average transmission bandwidth requirement, latency requirement, and transmission rate of each piece of data, and select candidate paths based on the current network status. Further analyze the reliability of the paths, considering the load situation, latency, bandwidth, and security of the paths. This information is obtained through the real-time network status acquisition system, and dynamic weight calculation is performed in combination with the network load data, and then generate the path evaluation set, filter out the candidate paths that meet the requirements, and select the most suitable path for the current data segment according to the parameters of the evaluation set, providing a basis for subsequent path selection and optimization, and forming the path evaluation set.

[0174] Based on the security and transmission requirements of each piece in the data sharding scheme, combined with the network characteristics and current load of the candidate paths in the path evaluation set, calculate the weight of each path, and perform dynamic selection of the candidate paths, using the formula:

[0175] ;

[0176] Calculate the optimization score of the path, and select the optimal path according to the score to generate the optimized transmission path;

[0177] Among them, is the optimized path, is the security score of the path, is the transmission rate of the path, is the network latency of the path, is the current load of the path, is the transmission requirement of the path.

[0178] Formula:

[0179] ;

[0180] The advantage of the formula is that by dynamically calculating factors such as the transmission rate, load, and delay of each path and comprehensively considering multiple characteristic factors, the path selection not only meets the transmission requirements but also optimizes the use of network resources, achieving an optimized effect with strong adaptability in a dynamically changing network environment.

[0181] Detailed explanation of the formula and the derivation process of formula calculation:

[0182] First, calculate the security score of each candidate path , transmission rate , network delay , current load and its transmission requirement , and the calculation process is as follows:

[0183] 1. Assume that the security score of a certain path , transmission rate , delay , current load , transmission requirement , substitute into the formula to calculate the path optimization score:

[0184] ;

[0185] 2. Calculate , substitute into the formula to get:

[0186] ;

[0187] This result indicates that after optimized calculation, the path optimization score is 184.26, which means that this path has a high transmission capacity under the current transmission requirements and network status and is suitable as the best selected transmission path to further support the efficient transmission and management of data streams.

[0188] The above is only a preferred embodiment of the present invention and does not impose other forms of limitations on the present invention. Any person skilled in the relevant art may use the disclosed technical content to make changes or modifications into equivalent embodiments with equivalent changes and apply them to other fields. However, as long as it does not depart from the technical solution content of the present invention, any simple modification, equivalent change, and modification made to the above embodiments based on the technical essence of the present invention still fall within the protection scope of the technical solution of the present invention.

Claims

1. A communication data protection system, characterized in that: The system comprises: The data receiving and decomposition module receives communication data, classifies the data into multi-dimensional tensors according to time, protocol type and flow rate, refines the original data structure through feature hierarchical analysis, and generates decomposed data tensors; The feature extraction and graph construction module extracts key features from the decomposed data tensor, identifies the distribution and density of each feature data, calculates the weight of the edge according to the interaction of the features, establishes a preliminary graph structure, the graph nodes represent multiple features, the edges reflect the relationship between the features, and generates an encrypted feature dynamic graph; The dynamic graph model update module monitors the node weights in the encrypted feature dynamic graph, dynamically adjusts nodes and connections according to new data, updates the graph structure to match the current changes in communication features, and generates an updated dynamic graph model; The classification and decision module uses the updated dynamic graph model to classify data streams, monitor weight changes between nodes, analyze data feature change trends, adjust decision logic to match data features according to change trends, and generate real-time classification results; The path optimization and data transmission module performs data sharding according to the real-time classification results, evaluates the available transmission paths according to the security and transmission requirements of each slice, evaluates the security and transmission rate, dynamically adjusts the path selection according to the evaluation results, and generates an optimized transmission path.

2. The communication data protection system according to claim 1, characterized in that: The decomposed data tensor specifically includes time, protocol type, and traffic; the encrypted feature dynamic graph specifically refers to the preliminary graph structure, graph nodes, and edges; the updated dynamic graph model includes node weights, connections, and graph structure; the real-time classification results specifically include data flow classification, weight changes, and data feature change trends; the optimized transmission path includes data segmentation, transmission path, security assessment, and transmission rate assessment.

3. The communication data protection system according to claim 2, characterized in that: The steps for obtaining the decomposed data tensor are specifically as follows: Receive communication data, classify the data into multi-dimensional tensors according to time, protocol type, and traffic, set classification standards, divide the data according to differentiation dimensions, obtain multi-class labels and assign categories to data points to obtain multi-dimensional tensor data sets; Performing feature hierarchical analysis on the multidimensional tensor data set, extracting key features in each dimension by analyzing data features in three dimensions: time, protocol type, and traffic, and performing refinement and normalization processing on the data features to obtain a refined feature data set; Based on the refined feature data set, cluster analysis is performed, by calculating the distances of data points in multiple dimensions, selecting matching cluster center points and classifying the data points, generating a multidimensional tensor decomposition result, and obtaining a decomposed data structure; Based on the decomposed data structure and the relationship between multiple tensor elements, the formula is adopted: ； Calculate the tensor decomposition value of each data point and generate a decomposed data tensor; in, Represents the element value of the decomposed data tensor, Represents the parameter of the first dimension in tensor decomposition, represents the parameter of the second dimension, represents the parameter of the third dimension, Represents the adjustment factor, which adjusts the relationship between the first dimension and the clustered data points. Represents the normalization parameter, which normalizes the relationship between the second dimension and the clustered data points. The maximum value of the tensor dimension.

4. The communication data protection system according to claim 3, characterized in that: The steps for obtaining the preliminary graphic structure are specifically as follows: Receiving feature data extracted from the decomposed data tensor, applying a statistical analysis method to calculate the distribution and density of each feature, obtaining distribution information of multiple features, and generating feature distribution data; Based on the feature distribution data, the mutual relationship between the features is analyzed, the fit between multiple features is calculated, and the relationship between the features is quantified using a distance metric to generate a feature relationship matrix; Using the feature relationship matrix, the formula is adopted: ； Calculate the edge weights between features and build a preliminary graph structure based on the edge weights; in, represents the edge weight between features, is the fit between features, Representative features The density of Representative features The density of Representative features The correlation degree, Representative features The correlation degree, represents the total number of features, represents the index of the feature, Representative features and Features The fit between Representative features and Features The fit between Represents the normalization factor, indicating the characteristics and The difference in fit between the two features is relative to the difference between all features.

5. The communication data protection system according to claim 4, characterized in that: The steps for obtaining the encrypted feature dynamic graph are specifically as follows: According to the feature nodes in the preliminary graph structure, the relationship between each pair of feature nodes is analyzed using an edge weight matrix, and based on the node relationship and weight distribution, the importance of each node in the graph is calculated to generate a node importance score; Based on the node importance scores, the most critical feature nodes for the graph structure are screened, the connection relationships between the key nodes and the edges are selected to form a subgraph, and the structure is encrypted to generate an encrypted feature subgraph; Through the encrypted feature subgraph, based on the weights between nodes and the encrypted features of edges, the formula is adopted: ； Calculate the node change value and graph structure evolution to generate an encrypted feature dynamic graph; in, Represents the encrypted feature dynamic graph, is the edge weight between features, is the fit between features, Representative feature node The density of Representative feature node The density of is a small constant used to avoid division by zero errors, express and The larger value of It represents the difference between the degree of fit between nodes and 1. Represents the total number of all feature nodes in the graph.

6. The communication data protection system according to claim 5, characterized in that: The steps for obtaining the updated dynamic graph model are specifically as follows: Based on the encrypted feature dynamic graph structure, multiple nodes and their connection relationships are obtained, and the weight changes of each node are analyzed by dynamically monitoring the changes of nodes and connections, and the weight values ​​of nodes and edges are updated in real time to generate node weight change results; According to the node weight change result, combined with the node's criticality and connectivity, the nodes associated with the communication feature change are screened, and according to the connection relationship between the nodes and the edges, the edge weights in the graph are readjusted, the dynamic changes of the nodes and the connections are updated, and the graph adjustment result is generated; Based on the adjustment results of the graph, the impact of new data on the node connection structure is dynamically analyzed and calculated, and the formula is used based on the weight change and the correction of node criticality: ； Calculate the time between nodes and edges According to the change of the encrypted feature dynamic graph, the encrypted feature dynamic graph is updated to generate an updated graph structure model; in, is the change of edge weights between nodes, Representative time Time Node With Node The weight of the edge between Representative time Time Node The weight value of Representative time Time Node The weight value of Representative Node With Node The connectivity value between Representative Node With Node The updated value of the connectivity between represents the difference in node weights, is the connectivity change value, Represents the total number of nodes in the network.

7. The communication data protection system according to claim 6, characterized in that: The steps for obtaining the real-time classification results are specifically as follows: Based on the updated dynamic graph model, monitor the weight changes between each node, analyze the law of node weight fluctuations, and extract the key change points of the data flow according to the characteristic change trend to generate the node weight change results; By iteratively analyzing the node weight change results, combining the connection relationship and communication characteristics between multiple nodes in the graph, judging the data flow change trend, identifying the associated feature change pattern, and generating data flow feature change trend results; According to the change trend results of the data stream characteristics, evaluate the impact of multiple characteristics on the decision logic, adjust the decision logic rules, and use the formula based on the change trend: ； Calculate each node at time The impact of feature changes on decision logic is analyzed, and the adjusted decision logic rules are obtained and applied to data stream classification to generate real-time classification results; in, For time The impact of the adjusted decision logic, For time Time Node The weight of is the node feature at time The change value when is the target value of the feature change, For a node at time The state value at is the expected state value of the node, Represents the total number of nodes, that is, the number of nodes participating in data flow classification in the graph.

8. The communication data protection system according to claim 7, characterized in that: The steps of obtaining the optimized transmission path are specifically as follows: Based on the real-time classification results, data streams are segmented to divide the data into multiple segments, and the transmission requirements and security parameters of each segment are recorded. Combined with the demand characteristics of each segment, a data segmentation plan is generated; According to the data sharding scheme, the transmission requirements and security of each piece of data are evaluated, and by comparing the security differences and transmission requirements between the fragments, a candidate set of available transmission paths is screened, and preliminary features of the candidate paths are obtained to generate a path evaluation set; According to the security and transmission requirements of each shard in the data sharding scheme, combined with the network characteristics and current load of the candidate paths in the path evaluation set, the weight of each path is calculated, and the candidate paths are dynamically selected using the formula: ； Calculate the optimization score of the path, select the best path based on the score, and generate the optimized transmission path; in, is the optimized path, is the safety score of the path, is the transmission rate of the path, is the network delay of the path, is the current load of the path, The transmission demand of the path.

Images