Power sensitive data identification method and system based on knowledge base

Abstract

The invention provides an electric power sensitive data identification method and system based on a knowledge base, and the method comprises the steps: obtaining electric power data uploaded by a user, carrying out the graph structure conversion of the electric power data based on a label of the electric power data, and obtaining an input information graph; carrying out relevance matching on the input information graph and a sensitive graph in a sensitive information knowledge base by utilizing a lightweight VF3 algorithm to obtain a data relevance relationship between the input information graph and the sensitive graph, and further determining whether the power data is power sensitive data or not; the sensitive information knowledge base is constructed on the basis of a plurality of sensitive graphs, and the sensitive graphs are obtained on the basis of a plurality of data elements forming sensitive information in the power transaction data and the relation between the data elements; the graph in the sensitive information knowledge base and the input information graph are subjected to graph matching by utilizing a lightweight VF3 algorithm, so that complex association between data can be mined, and sensitive information generated by data combination can be quickly and accurately identified.

Description

Technical Field

[0001] The present invention relates to the field of power data security technology, and in particular to a knowledge base-based power sensitive data identification method and system. Background Art

[0002] With the rapid development of smart grids and new power systems, massive amounts of power data are exposing more and more security and privacy issues during transmission, processing and storage. Power data contains a large amount of sensitive information, such as power load, equipment operating conditions, user behavior, etc. If leaked or attacked, it will have a significant impact on power grid operation and user privacy. Therefore, the identification and protection of sensitive power data has become an important research topic in the current field of power information security.

[0003] In response to the above problems, researchers have proposed a variety of methods for automatically identifying sensitive information in power data. One is a method based on machine learning: using algorithms such as support vector machines, decision trees, random forests, and deep learning, by constructing feature engineering, extracting key features such as power load and power fluctuations to identify sensitive information. Machine learning methods can process large-scale power data, have high recognition efficiency, and are applicable to a variety of data types. The other is a method based on data labeling: the rule matching method identifies sensitive information through pre-defined rules (such as keyword matching, data type screening), and the semantic analysis method uses natural language processing technology to automatically analyze potential sensitive information in power data. For example, in "CN118312992A", sensitive data is identified by extracting features from semi-structured power data and training a convolutional neural network model, which can effectively process semi-structured power data.

[0004] However, due to the high degree of coupling between power businesses, there is a close correlation between power data. Through integration and correlation analysis, data with a low security level may be used to derive information with a higher security level. The above method does not fully consider the situation where different key field combinations may generate new sensitive fields. In a complex power data environment, such field combinations may lead to new sensitive information leakage.

[0005] In summary, both machine learning methods and data labeling methods lack the ability to jointly analyze large-scale and complex power data, that is, they fail to fully consider the data association caused by power business coupling, thereby ignoring the new sensitive information generated by data combination, resulting in inaccurate identification results. Summary of the invention

[0006] In order to overcome the above-mentioned deficiencies of the prior art, the present invention proposes a method for identifying power sensitive data based on a knowledge base, comprising:

[0007] Obtaining power data uploaded by a user, and converting the power data into a graph structure based on tags corresponding to the power data to obtain an input information graph;

[0008] Using a lightweight VF3 algorithm, the input information graph is matched with each sensitive graph in the sensitive information knowledge base to obtain a data association relationship between the input information graph and each sensitive graph;

[0009] Determining whether the power data is power sensitive data based on a data association relationship between the input information graph and each of the sensitive graphs;

[0010] The sensitive information knowledge base is constructed based on a plurality of the sensitive graphs, and the sensitive graphs are obtained based on a plurality of data elements constituting sensitive information in the power transaction data and the connections between the plurality of the data elements.

[0011] Optionally, the sensitive information knowledge base is constructed in the following manner:

[0012] Collecting electricity transaction data in the power system;

[0013] Classifying a plurality of data elements constituting sensitive information in the power transaction data according to the relationship between the data elements, to obtain a group of the data elements corresponding to each piece of sensitive information;

[0014] Based on each group of the data elements, each of the data elements in the group is used as a node, and the connection between each of the data elements in the group is used as an edge to form a sensitivity graph;

[0015] The sensitive graph formed by each group of data elements is stored in a sensitive information knowledge base.

[0016] Optionally, the plurality of data elements constituting sensitive information in the power transaction data and the relationship between the plurality of data elements are obtained in the following manner:

[0017] Extracting power transaction entities from power transaction data;

[0018] Classifying the power transaction data into multiple types according to their importance, and classifying each type of power transaction data into multiple levels according to sensitive information contained in the power transaction data;

[0019] Using an entity-relationship joint extraction algorithm in combination with the power transaction subject, type and level of the power transaction data, identifying entities and relationships in the power transaction data;

[0020] Determine, using a graph neural network based on entities and relationships in the power transaction data, a number of data elements constituting sensitive information in the power transaction data;

[0021] Based on the plurality of data elements, and entities and relationships in the power transaction data, connections between the plurality of data elements are obtained.

[0022] Optionally, the using of a lightweight VF3 algorithm to perform correlation matching on the input information graph and each sensitive graph in the sensitive information knowledge base to obtain a data correlation relationship between the input information graph and each of the sensitive graphs includes:

[0023] Based on each of the sensitive graphs, the node sorting function is used to sort the nodes of the sensitive graph to obtain the node order corresponding to the sensitive graph; the pattern graph preprocessing is performed on the sensitive graph to generate the state and parent node structure of the sensitive graph; based on the data structure representation, state, node order, parent node structure of the sensitive graph and the data structure representation of the input information graph, a matching function is used to perform recursive matching to obtain the matching result of the sensitive graph and the input information graph; the matching result of the sensitive graph and the input information graph is stored in the matching result set of the sensitive graph;

[0024] Storing the matching result set of each of the sensitive graphs into the total matching result set;

[0025] According to the total set of matching results, a data association relationship between the input information graph and each of the sensitive graphs is obtained.

[0026] Optionally, performing recursive matching based on the data structure representation, state, node order, parent node structure of the sensitive graph and the data structure representation of the input information graph using a matching function to obtain a matching result between the sensitive graph and the input information graph includes:

[0027] Check whether the current state of the sensitive map is the target state;

[0028] If it is the target state, storing the current matching result of the sensitive map into the matching result set;

[0029] If it is not the target state, then based on the candidate node pairs generated by the current parameters, update the current state to obtain a new state, pass the new state and the current matching result to the matching function for recursive call processing to obtain a new matching result; restore the new state to the current state, reselect the candidate node pairs to generate a new state, and continue to pass the new state and the new matching result to the matching function for recursive call processing until the termination condition is reached;

[0030] The termination condition is that the new state is the target state, or all candidate node pairs are traversed;

[0031] The current parameters are the data structure representation, state, node order, parent node structure of the sensitive graph and the data structure representation of the input information graph.

[0032] Optionally, the candidate node pair generated based on the current parameters updates the current state to obtain a new state, including:

[0033] Using a candidate node generation function, a plurality of candidate node pairs are generated based on current parameters of the sensitivity graph;

[0034] Using a preset feasibility condition to screen the candidate node pairs, and obtain a plurality of candidate node pairs that meet the feasibility condition;

[0035] A candidate node pair is selected from a plurality of candidate node pairs that meet feasibility conditions, and a current state of the sensitive graph is expanded to obtain a new state.

[0036] Optionally, the data structure representation of the input information graph and each of the sensitivity graphs is obtained according to the following method:

[0037] Based on the node set and edge set of the input information graph, a data structure representation of the input information graph is constructed;

[0038] Mapping the nodes of each sensitive graph in the input information graph and the sensitive information knowledge base to the node label set of the sensitive graph to obtain a point mapping function;

[0039] Mapping the edges of each sensitive graph in the input information graph and the sensitive information knowledge base to the edge label set of the sensitive graph to obtain an edge mapping function;

[0040] For each sensitive graph, a data structure representation of the sensitive graph is constructed based on a node set, an edge set, a node label set, an edge label set, a point mapping function, and an edge mapping function of the sensitive graph.

[0041] Optionally, determining whether the power data is power sensitive data based on the data association relationship between the input information graph and each of the sensitive graphs includes:

[0042] If the data association relationship includes at least one pair of matching node pairs formed by nodes in the sensitive graph and the input information graph, the power data is determined to be power sensitive data.

[0043] Based on the same inventive concept, the present invention proposes a knowledge base-based power sensitive data identification system, comprising:

[0044] A graph structure conversion module, used to obtain the power data uploaded by the user, and based on the tags corresponding to the power data, convert the power data into a graph structure to obtain an input information graph;

[0045] A graph structure matching module, used to use a lightweight VF3 algorithm to perform correlation matching between the input information graph and each sensitive graph in the sensitive information knowledge base, so as to obtain a data correlation relationship between the input information graph and each sensitive graph;

[0046] A sensitive data identification module, used to determine whether the power data is power sensitive data based on the data association relationship between the input information graph and each of the sensitive graphs;

[0047] The sensitive information knowledge base is constructed based on a plurality of the sensitive graphs, and the sensitive graphs are obtained based on a plurality of data elements constituting sensitive information in the power transaction data and the connections between the plurality of the data elements.

[0048] Optionally, the graph structure matching module is specifically used to:

[0049] Collecting electricity transaction data in the power system;

[0050] Classifying a plurality of data elements constituting sensitive information in the power transaction data according to the relationship between the data elements, to obtain a group of the data elements corresponding to each piece of sensitive information;

[0051] Based on each group of the data elements, each of the data elements in the group is used as a node, and the connection between each of the data elements in the group is used as an edge to form a sensitivity graph;

[0052] The sensitive graph formed by each group of data elements is stored in a sensitive information knowledge base.

[0053] Optionally, the graph structure matching module is specifically used to:

[0054] Extracting power transaction entities from power transaction data;

[0055] Classifying the power transaction data into multiple types according to their importance, and classifying each type of power transaction data into multiple levels according to sensitive information contained in the power transaction data;

[0056] Using an entity-relationship joint extraction algorithm in combination with the power transaction subject, type and level of the power transaction data, identifying entities and relationships in the power transaction data;

[0057] Determine, using a graph neural network based on entities and relationships in the power transaction data, a number of data elements constituting sensitive information in the power transaction data;

[0058] Based on the plurality of data elements, and entities and relationships in the power transaction data, connections between the plurality of data elements are obtained.

[0059] Optionally, the graph structure matching module is specifically used to:

[0060] Based on each of the sensitive graphs, the node sorting function is used to sort the nodes of the sensitive graph to obtain the node order corresponding to the sensitive graph; the pattern graph preprocessing is performed on the sensitive graph to generate the state and parent node structure of the sensitive graph; based on the data structure representation, state, node order, parent node structure of the sensitive graph and the data structure representation of the input information graph, a matching function is used to perform recursive matching to obtain the matching result of the sensitive graph and the input information graph; the matching result of the sensitive graph and the input information graph is stored in the matching result set of the sensitive graph;

[0061] Storing the matching result set of each of the sensitive graphs into the total matching result set;

[0062] According to the total set of matching results, a data association relationship between the input information graph and each of the sensitive graphs is obtained.

[0063] Optionally, the graph structure matching module is specifically used to:

[0064] Check whether the current state of the sensitive map is the target state;

[0065] If it is the target state, storing the current matching result of the sensitive map into the matching result set;

[0066] If it is not the target state, then based on the candidate node pairs generated by the current parameters, update the current state to obtain a new state, pass the new state and the current matching result to the matching function for recursive call processing to obtain a new matching result; restore the new state to the current state, reselect the candidate node pairs to generate a new state, and continue to pass the new state and the new matching result to the matching function for recursive call processing until the termination condition is reached;

[0067] The termination condition is that the new state is the target state, or all candidate node pairs are traversed;

[0068] The current parameters are the data structure representation, state, node order, parent node structure of the sensitive graph and the data structure representation of the input information graph.

[0069] Optionally, the graph structure matching module is specifically used to:

[0070] Using a candidate node generation function, a plurality of candidate node pairs are generated based on current parameters of the sensitivity graph;

[0071] Using a preset feasibility condition to screen the candidate node pairs, and obtain a plurality of candidate node pairs that meet the feasibility condition;

[0072] A candidate node pair is selected from a plurality of candidate node pairs that meet feasibility conditions, and a current state of the sensitive graph is expanded to obtain a new state.

[0073] Optionally, the graph structure matching module is specifically used to:

[0074] Based on the node set and edge set of the input information graph, a data structure representation of the input information graph is constructed;

[0075] Mapping the nodes of each sensitive graph in the input information graph and the sensitive information knowledge base to the node label set of the sensitive graph to obtain a point mapping function;

[0076] Mapping the edges of each sensitive graph in the input information graph and the sensitive information knowledge base to the edge label set of the sensitive graph to obtain an edge mapping function;

[0077] For each sensitive graph, a data structure representation of the sensitive graph is constructed based on a node set, an edge set, a node label set, an edge label set, a point mapping function, and an edge mapping function of the sensitive graph.

[0078] Optionally, the sensitive data identification module is specifically used to:

[0079] If the data association relationship includes at least one pair of matching node pairs formed by nodes in the sensitive graph and the input information graph, the power data is determined to be power sensitive data.

[0080] Compared with the closest prior art, the present invention has the following beneficial effects:

[0081] The present invention provides a knowledge base-based power sensitive data identification method and system, comprising: obtaining power data uploaded by a user, and converting the power data into a graph structure based on a label corresponding to the power data to obtain an input information graph; using a lightweight VF3 algorithm, performing correlation matching between the input information graph and each sensitive graph in a sensitive information knowledge base to obtain a data correlation relationship between the input information graph and each of the sensitive graphs; determining whether the power data is power sensitive data based on the data correlation relationship between the input information graph and each of the sensitive graphs; the sensitive information knowledge base is constructed based on a plurality of sensitive graphs, and the sensitive graphs are constructed based on a plurality of data elements constituting sensitive information in power transaction data and a plurality of sensitive graphs. The connection between the data elements is obtained; in this scheme, the lightweight VF3 algorithm is used to match the sensitive graph in the sensitive information knowledge base with the input information graph to perform correlation matching on the graph structure, so as to determine whether the power data contains sensitive information. Compared with the traditional keyword matching, this scheme matches and compares through the intrinsic connection between data elements, so as to mine the complex connection between data, identify the sensitive information generated by the combination of data, and improve the accuracy of sensitive information data identification. At the same time, it avoids the situation where the sensitive data generated by the integration of data elements with complex connections is not identified, thereby avoiding the leakage of sensitive information data; and the lightweight VF3 algorithm can quickly and accurately identify sensitive information in massive power data. BRIEF DESCRIPTION OF THE DRAWINGS

[0082] Figure 1 A schematic diagram of a flow chart of a method for identifying power sensitive data based on a knowledge base provided by the present invention;

[0083] Figure 2 A schematic diagram of the process of converting the electric power data graph structure provided by the present invention;

[0084] Figure 3 A schematic diagram of business categories of power data provided by the present invention;

[0085] Figure 4 A detailed framework diagram of the classification and grading of power transaction data provided by the present invention;

[0086] Figure 5 Schematic diagram of the sensitivity graph composed of data elements provided by the present invention Figure 1 ;

[0087] Figure 6 Schematic diagram of the sensitivity graph composed of data elements provided by the present invention Figure 2 ;

[0088] Figure 7 Schematic diagram of the sensitivity graph composed of data elements provided by the present invention Figure 3 ;

[0089] Figure 8 The execution time experimental result diagram of different recognition algorithms provided by the present invention;

[0090] Fig. 9 A schematic diagram of the structure of a power sensitive data identification system based on a knowledge base provided by the present invention;

[0091] Fig.10 The present invention provides a schematic structural diagram of an electronic device. DETAILED DESCRIPTION

[0092] As an important part of the power system, sensitive data on power dispatching and market operations involve detailed data on power transactions, market operations, and power dispatching. These data are closely related to the trading strategies, price information, and safe operation of the power system of market participants. With the advancement of power market reform and smart grid technology, the risk of leakage of sensitive data on power dispatching and market operations is also increasing. Leakage of sensitive data on power dispatching and market operations may lead to the following serious consequences:

[0093] 1. Increased system security risks: Power dispatch data contains information such as the real-time status of power grid operation and dispatch plans. If this data is leaked, attackers may launch targeted network attacks by analyzing the vulnerabilities of the power grid, threatening the security and stability of the power system and even causing serious consequences such as regional power outages.

[0094] 2. Market manipulation and unfair competition: Data leakage in the electricity market may enable market participants to obtain sensitive data such as competitors’ trading strategies and price information, thereby manipulating market prices or engaging in unfair competition, undermining market fairness and transparency and harming the healthy development of the market.

[0095] 3. User privacy leakage: Power dispatch and market data often contain information such as users' power consumption habits and contract terms. If these data are obtained by criminals, they may be used to commit illegal activities such as power fraud and identity theft, seriously infringing on users' privacy rights.

[0096] Therefore, it is often necessary to classify and grade power data including power dispatching operation and market sensitive data, and use corresponding security levels to protect them to avoid leakage of sensitive data.

[0097] At present, the classification and grading of power data faces two core problems. First, the security level of power data needs to be adjusted in real time as external conditions change, such as when new regulatory standards require the protection or disclosure of certain data, or when it is discovered that certain data may reveal key sensitive information. Second, due to the high degree of coupling between power businesses, there is a close correlation between power data, and data with low security levels may be used to derive information with higher security levels through integration and correlation analysis. Existing manual classification and grading methods, machine learning methods, and data labeling methods are difficult to dynamically analyze the complex relationships between power data, and are unable to automatically and quickly update the security level of power data. In addition, the effectiveness of existing machine learning methods depends on the construction of large-scale annotated data sets, and the model's generalization ability for new types of data still needs to be further improved; in data labeling methods, as the complexity of the power system continues to increase, existing rules need to be frequently updated to cope with unknown or newly emerging sensitive information.

[0098] For example, CN115758172A achieves rapid identification of sensitive data by generating a minimum connected dominating set and performing simplified graph matching. Although it improves the matching speed and simplifies the computational complexity, it does not mention the knowledge base for storing sensitive data templates, resulting in a lack of system support for sensitive data template management and maintenance, and an inability to effectively manage or update sensitive data templates during data identification; and CN118427244A identifies sensitive information in energy data through association mining. This method does not fully consider the processing and real-time update capabilities of large-scale power data in sensitive data identification.

[0099] Therefore, it is of great practical significance to construct a sensitive data identification model that can be updated in real time and is based on the correlation of power data.

[0100] The specific implementation modes of the present invention are further described in detail below with reference to the accompanying drawings.

[0101] Example 1

[0102] The present invention provides a method for identifying power sensitive data based on a knowledge base, such as Figure 1 As shown, including:

[0103] S1. Obtaining power data uploaded by a user, and converting the power data into a graph structure based on tags corresponding to the power data to obtain an input information graph;

[0104] S2. Using a lightweight VF3 algorithm, the input information graph is matched with each sensitive graph in the sensitive information knowledge base to obtain a data association relationship between the input information graph and each sensitive graph;

[0105] The sensitive information knowledge base is constructed based on a plurality of the sensitive graphs, and the sensitive graphs are obtained based on a plurality of data elements constituting sensitive information in the power transaction data and a relationship between a plurality of the data elements;

[0106] S3. Determine whether the power data is power sensitive data based on the data association relationship between the input information graph and each of the sensitive graphs.

[0107] In step S1, the power data uploaded by the user is obtained, and based on the tags corresponding to the power data, the power data is converted into a graph structure to obtain an input information graph.

[0108] Obtain the power data uploaded by the user, and convert the power data information uploaded by the user into a graph structure composed of nodes and edges according to its corresponding labels, such as Figure 2 Specifically, the conversion scenario considered in this embodiment is, for example, on a web page, a specific input box is given, and a specified type of power data information is input, and the background can automatically obtain the label corresponding to the power data information. If the power data information uploaded by the user is a large amount of text data, it is necessary to use the entity-relationship joint extraction algorithm to extract the structure of the input power data information.

[0109] Convert the power data information uploaded by the user into a directed graph, such as Figure 2 As shown, each node (Node1, Node2, Node3, Node4, Node5) represents a data element (V) in the input information, that is, Figure 2 In V1-V5, each edge (E) represents the relationship between different data elements (considering the above web page input scenario, the corresponding backend database stores the relationship between different data element types), that is, Figure 2 In E1-E4, the same type label of the input information is used as the node label of the graph (Node type). Figure 2 The node labels in the graph include A, C, D, E and G. The relationship between different types of node labels is used as the edge label of the graph structure (Edge Type). Figure 2 The edge labels in include c, e, d, n. Using the directed graph (input information graph) G input =(V input ,E input ) to represent the input information, where V input represents the node set in the input information graph, E input Represents the set of edges in the input information graph.

[0110] In step S2, S21, before performing relevance matching, a sensitive information knowledge base is pre-built.

[0111] Specifically, S21-1, collecting power transaction data in the power system; obtaining a number of data elements constituting sensitive information in the power transaction data and the connections between the data elements.

[0112] S21-1-1. Extract the power transaction subject from the power system's power transaction data.

[0113] The power system equipment is widely distributed, there are many participants, and the interaction methods are complex. At the same time, the data in the new power system is large, diverse, real-time, high-speed, complex and has high security requirements. This has an important impact on power data access control. The amount of data is huge and the sources are diverse, which requires effective processing and management. Real-time and high speed require timely response and processing of large amounts of real-time data. The complexity of the data requires processing of multiple types and levels of data. Due to the criticality and sensitivity of the power system, the security and privacy protection of data are crucial, so access control requires security measures such as encryption, identity authentication and permission management. First, classify the power data according to its business category, such as Figure 3 As shown, the categories of power business include: engineering construction, equipment operation and maintenance, operation and scheduling, customer marketing and safety supervision.

[0114] Collect power system transaction data from customer marketing categories. In this solution, we take the power transaction data of the green power trading market as an example to build a sensitive data knowledge base. Analyze the data circulation needs in the power trading scenario. Figure 4 The power trading entities and power trading data involved in the green power trading market were sorted out. Institutions such as the National Renewable Energy Management Center and the Beijing Power Trading Center are responsible for power trading management, policy formulation and market supervision across the country to ensure the smooth and sustainable development of power trading. Provincial power trading centers are responsible for managing and supervising power trading activities within their administrative regions, implementing the policies of national institutions, and making appropriate adjustments and supplements based on local actual conditions. Market entities participating in power trading include power sales companies, power purchasing companies, virtual aggregators and individual power purchasers. Power sales companies are responsible for selling electricity to end users, while power purchasing companies are companies or organizations that purchase electricity. Virtual aggregators participate in market transactions on a larger scale and improve their bargaining power by aggregating the demand of multiple small power purchasers. Individual power purchasers are individual consumers who directly purchase and use electricity.

[0115] Obtain the power trading entity corresponding to each power trading data from the power trading market.

[0116] S21-1-2. Divide the power transaction data into multiple types according to their importance, and divide each type of power transaction data into multiple levels according to sensitive information contained in the power transaction data.

[0117] Figure 4 This paper describes a detailed framework for the classification and grading of power trading data in the power trading platform, aiming to achieve refined management and control of different data types. The power trading data is divided into five categories according to their importance, namely core data, sensitive data, business data, operational data and other non-core data. Each type of data is further divided into different levels according to its sensitivity (i.e. the sensitive information contained therein) and importance, and is given different openness, including four levels: non-sharing, conditional sharing, partial opening after approval, and opening to the public. It supports the construction of a sensitive information knowledge base for sensitive data monitoring, dangerous behavior identification and access control.

[0118] S21-1-3. Utilize an entity-relationship joint extraction algorithm to combine the power transaction subject, type and level of the power transaction data to identify entities and relationships in the power transaction data.

[0119] Using algorithms such as sensitive word discovery, entity recognition, and neural networks, sensitive data subcategories, i.e., data elements that constitute sensitive information, are discovered. Specifically, in this solution, the Joint Entity and Relation Extraction algorithm is used to identify and extract entities in power trading data, as well as the relationships between entities, based on the power trading subject, type, and level.

[0120] S21-1-4. Determine the data elements that constitute sensitive information in the power trading data based on the entities and relationships in the power trading data using a graph neural network; obtain the connections between the data elements based on the data elements and the entities and relationships in the power trading data.

[0121] S21-2, constructing a plurality of sensitive graph structures based on the plurality of data elements and the connections between the data elements and storing them in a sensitive information knowledge base. Specifically:

[0122] Classifying a plurality of data elements constituting sensitive information in the power transaction data according to the relationship between the data elements, to obtain a group of the data elements corresponding to each piece of sensitive information;

[0123] Based on each group of data elements, each data element in the group is used as a node, and the connection between each data element in the group is used as an edge to form a graph structure of a sensitive graph, such as Figure 5-Figure 7 As shown in the figure, the circular nodes are data elements, each node has a point label (Node Type), the nodes are connected by edges, and each edge has an edge label (EdgeType). The three data elements and the edge relationships between them together constitute a piece of sensitive information, which is represented by Figure 5-Figure 7 Graph data structure representation of ;

[0124] The sensitive graph structure composed of each group of data elements is stored in a sensitive information knowledge base; the sensitive information knowledge base is composed of multiple graph data structures, covering all permutations and combinations of data elements that may constitute sensitive information, and is stored using a Neo4j graph database. The sensitive information knowledge base stores sensitive information in a graph structure, which is composed of multiple (1) data elements. A single data element may be sensitive, but the combined graph structure must be sensitive.

[0125] In this embodiment, the constructed sensitive information knowledge base contains a total of 162 types of sensitive data and the sensitive information combinations they constitute.

[0126] Since the security level of power data needs to be adjusted in real time as external conditions change, the sensitive information knowledge base needs to be updated in real time. The time interval can be set to obtain new power data and the sensitive information knowledge base can be updated based on the new security level. That is, this solution can provide a real-time updated recognition algorithm.

[0127] S22. Using the lightweight VF3 algorithm, the input information graph is matched with each sensitive graph in the sensitive information knowledge base to obtain a data association relationship between the input information graph and each of the sensitive graphs.

[0128] S22-1. Before performing relevance matching, pre-acquire the parameters required for the execution of the lightweight VF3 algorithm, including the data structure representation of the input information graph, the data structure representation of the sensitive graph, and the mapping matching conditions of the nodes when executing the algorithm. Specifically:

[0129] Based on the node set and edge set of the input information graph, a data structure representation of the input information graph is formed. The input information graph G input =(V input ,E input ).

[0130] The nodes of the input information graph and each sensitive graph in the sensitive information knowledge base are mapped to the node label set of the sensitive graph to obtain a point mapping function; the edges of the input information graph and each sensitive graph in the sensitive information knowledge base are mapped to the edge label set of the sensitive graph to obtain an edge mapping function.

[0131] Specifically, we define the sensitive graphs G1, G2, …, G representing sensitive information in the sensitive information knowledge base. n The node set (V1, V2, ..., V n ) and the input information diagram G representing the input information input The node set V input To node label set Point mapping function λ v for:

[0132]

[0133] Among them, V i Represents the node set of the i-th sensitive graph, and the label corresponding to each node can be represented by the point mapping function λ v : OK, V represents a node.

[0134] Similarly, define the sensitive graphs G1, G2, …, G representing sensitive information in the sensitive information knowledge base. n The edge set (E1, E2, ..., E n ) and input information G representing the data uploaded by the user input The edge set E input To edge label set The edge mapping function λ e for:

[0135]

[0136] The node label sets of the input information graph and the sensitive graph in the sensitive knowledge information database should be consistent and share the same set of node label sets; and the edge label sets of the input information graph and the sensitive graph in the sensitive knowledge information database should also be consistent and share the same set of edge label sets. The label corresponding to each edge can be obtained by the edge mapping function λ e : OK, E represents an edge.

[0137] For each sensitive graph G in the sensitive information knowledge base, based on the node set, edge set, node label set, edge label set, the node mapping function and edge mapping function obtained above, a data structure representation of the sensitive graph is constructed. Here, V and E represent the node set and edge set of the sensitive graph, respectively, rather than simply referring to a data element and an edge.

[0138] Based on the above basic definition of data structure representation, the identification of sensitive information can be achieved by graph matching algorithm. Next, the association matching process between sensitive graph and input information graph is defined. For the n sensitive graphs (G1, G2, …, G n ), these sensitive graphs need to be respectively input Perform subgraph isomorphism detection, that is, to explore the deep correlation matching between the sensitive graph and the input information graph. The matching process can be achieved by finding a mapping function P:V that satisfies point label matching, edge label matching, and precise graph structure matching. i →V input, which means that the nodes in the sensitive graph can be mapped to the input information graph. In order to find a mapping function P that meets the above requirements, the points of the graph participating in the matching must meet the following mapping conditions:

[0139]

[0140] Where n represents the number of sensitive graphs in the sensitive information knowledge base, u represents the nodes in the sensitive graph, and V n represents the node set in the sensitive graph; μ n (u) represents a sub-mapping function in the mapping function P, indicating u∈V n →V input , which means the node in the input information graph corresponding to node u, the node u in the sensitive graph has been mapped to the input information graph by the mapping function P; λ v (u) represents the node label determined by the point mapping function for node u, λ v (μ n (u)) represents the node label determined by the point mapping function in the input information graph corresponding to node u; λ v (u)≈λ v (μ n (u)) indicates that the node labels between the mapped nodes are semantically compatible. Formula (1) indicates that for each node in the sensitive graph, there must be a corresponding node in the input information graph, and the node labels of each pair of corresponding nodes must be semantically compatible. ≈ indicates the compatibility between two labels. It does not require that the two labels are completely equal, but they should be semantically consistent.

[0141] u′ represents another node in the sensitive graph, μ n (u′) represents the node in the input information graph corresponding to the node u′ in the sensitive graph; Formula (2) represents that for two different nodes in the sensitive graph, the corresponding nodes in the input information graph cannot be the same.

[0142] The edges of the sensitive graph and the input information graph involved in the matching should satisfy the following mapping conditions:

[0143]

[0144] Where n represents the number of sensitive graphs in the sensitive information knowledge base, u represents a node in the sensitive graph, (u,u′) represents an edge in the sensitive graph, and E n represents the edge set in the sensitive graph, (μ n (u),μ n (u′)) indicates that the edge (u,u′) in the sensitive graph maps to the corresponding edge in the input information graph, E input represents the edge set in the input information graph, λ e((u,u′)) represents the edge label of edge (u,u′) determined based on the edge mapping function; λ e ((μ n (u),μ n (u′))) represents the edge label of the edge in the input information graph corresponding to the edge (u,u′) based on the edge mapping function; Formula (3) indicates that for each edge in the sensitive graph, there must be an edge corresponding to it in the input information graph, and the labels of each pair of corresponding edges must be semantically compatible to ensure that the structural relationship of the sensitive graph is preserved in the input information graph.

[0145] (v,v′) represents the edge in the input information graph, μ n -1 (v) represents μ n The inverse of (v), μ n -1 (v′) represents μ n The inverse of (v′), that is, the corresponding node in the sensitive graph of the node in the input information graph, (μ n -1 (v),μ n -1 (v′)) represents the edge in the input information graph corresponding to the edge in the sensitive graph, λ e ((v,v′)) represents the edge label of edge (v,v′) determined based on the edge mapping function; λ e ((μ n (v),μ n (v′))) represents the edge label of the edge in the sensitive graph corresponding to the edge (v, v′) determined based on the edge mapping function; Formula (4) indicates that for each edge in the input information graph, there must be an edge corresponding to it in the sensitive graph, and the labels of each pair of corresponding edges must be semantically compatible. ≈ represents the compatibility of label values, which is a less strict matching relationship than equality.

[0146] S22-2, based on the data structure representation of the input information graph, the data structure representation of the sensitive graph, and the mapping matching conditions of the nodes when executing the algorithm, the lightweight VF3 algorithm is used to first sort the nodes of the sensitive graph in the sensitive information knowledge base, and then the preprocessing pattern graph is used to generate the initial state and parent node tree. Next, the algorithm matches through recursive depth-first search, uses feasibility rules to ensure the legitimacy of each state, and finally returns all the matching results found. Specifically:

[0147] The overall process of the lightweight VF3 sensitive information identification algorithm is described as follows:

[0148]

[0149] Among them, Algorithm represents the algorithm; VF3_Light_sensitive Data Identification represents the sensitive information identification of the lightweight VF3 algorithm, and Require and Ensure represent the input conditions and output conditions of the algorithm respectively. The main process of identifying sensitive data based on the lightweight VF3 algorithm is divided into the following steps:

[0150] First, initialize an empty set All Results to store the matching results of each sensitive graph. Then, list For each sensitivity map, the i-th sensitivity map G i For example, i represents the i-th sensitive graph in the sensitive information knowledge base. First, Compute Ordering (G i ,G input ) Node sorting function for sensitive graph G i Sort the nodes and generate the sensitive graph G i The node order Next, use PreprocessPattern For sensitive graph G i Perform pattern map preprocessing to generate sensitive map G i Then, initialize an empty matching result set Results to store the current sensitive graph G. i Based on the data structure representation, state, node order, parent node structure of the sensitive graph and the data structure representation of the input information graph, a matching function is used to perform recursive matching, that is, through the matching function Match(S0,G i ,G input , Parent, Results) performs a recursive matching process to obtain a matching result between the sensitive map and the input information map, and stores the matching result between the sensitive map and the input information map in a matching result set Results;

[0151] After the current sensitive graph completes recursive matching, return the current sensitive graph G i The results are stored in AllResults[G i ]middle;

[0152] Finally, after processing all sensitive maps, the matching result set of each sensitive map is stored in the total matching result set, that is, all matching result sets are returned and stored in All Results;

[0153] According to the total set of matching results All Results, a data association relationship between the input information graph and each of the sensitive graphs is obtained.

[0154] This algorithm transforms the problem of identifying sensitive data into a subgraph isomorphism problem between multiple graphs. Through the above process, each graph in the graph list is processed in turn, matched with the input graph and the results are stored.

[0155] The recursive matching process in the above algorithm steps, i.e., the process of performing recursive matching using a matching function based on the data structure representation, state, node order, parent node structure of the sensitive graph and the data structure representation of the input information graph to obtain the matching result of the sensitive graph and the input information graph, is specifically as follows:

[0156]

[0157]

[0158] The above table is the code of the recursive matching procedure in the algorithm. This part performs isomorphic matching between the input information graph and multiple sensitive graphs in the sensitive information knowledge base through candidate node generation, feasibility check and state recovery.

[0159] procedure MATCH(s,G1,G2, Parent, out Results) is a recursive matching process function.

[0160] First, the Is Goal(s) function is used to check whether the current state s of the sensitive graph is the target state.

[0161] If it is the target state, the current mapping M(s) is added to the matching result set Results. The current mapping M(s) refers to the set of mapping nodes in the current state of the sensitive graph, that is, the matching node pairs in the current matching results.

[0162] If it is not the target state, the candidate node generation function is used to generate several candidate node pairs based on the current parameters, that is, Next Candidates(s, Parent, G1, G2) generates candidate node pairs (u n ,v n ); the current parameters are the state s of the sensitive graph, the data structure representation G1, and the node order The parent node structure Parent and the data structure representation G2 of the input information graph;

[0163] Based on the candidate node pairs generated by the current parameters, the current state is updated to obtain the new state. Specifically, for each candidate node pair, the algorithm checks whether it meets the preset feasibility condition (Is Feasible(s,u n ,v n ), that is, using the feasibility condition to screen the candidate node pairs, and obtaining multiple candidate node pairs that meet the feasibility condition; selecting a candidate node pair from the multiple candidate node pairs that meet the feasibility condition, and expanding the current state of the sensitive graph to obtain a new state S n :=Extend State(s,u n ,v n );

[0164] Next, the new state and the current matching result (if the matching function has not been used for matching, the current matching result is empty) are passed to the matching function Match (S n ,G1,G2,N G1 ,Parent,Results) to perform recursive call processing and continue to process the new state S n , and also pass the data structure representation of the current sensitive graph and the parent node structure to the matching function;

[0165] After the processing is completed, call the recovery function Restore State(s,u n ,v n ) restore the new state to the previous state, that is, the current state s, reselect the candidate node pair to generate a new state, and continue to pass the new state and the new matching result to the matching function for recursive calling until the termination condition is reached;

[0166] The termination condition is that the new state is the target state, or all candidate node pairs are traversed; if only the first solution is needed, the algorithm can stop executing when the first target state is found.

[0167] Through recursive expansion and backtracking, the algorithm can efficiently traverse the search space and find all possible subgraph isomorphic solutions.

[0168] In step S3, based on the data association relationship between the input information graph and each of the sensitive graphs, it is determined whether the power data is power sensitive data.

[0169] Specifically, if the data association relationship includes at least one pair of matching node pairs formed by nodes in the sensitive graph and the input information graph, the power data is determined to be power sensitive data.

[0170] In the context of the circulation of massive power data, the present invention provides a sensitive data identification method that can fully consider the association relationship of power data. The method analyzes the potential sensitive information in the data flow process by mining the complex associations between power data. A knowledge base is constructed by adopting a sensitive information storage method based on a graph structure, and a sensitive information identification algorithm based on graph matching is proposed on this basis. The algorithm can mine the intrinsic associations between data, thereby realizing real-time monitoring and protection of sensitive information in power data, and ensuring the security and privacy of sensitive data during transmission and processing.

[0171] Specifically, the present invention provides a sensitive data identification algorithm for power trading scenarios based on a knowledge base and a lightweight VF3 graph matching algorithm. Sensitive information of power transactions is stored in the form of a graph structure and stored in the knowledge base as graph structure sensitive information metadata. The data input by the user is compared and identified with the graph structure sensitive information metadata stored in the sensitive data knowledge base, and the input data is constructed into a graph data structure composed of points, edges, point labels and edge labels. The lightweight VF3 graph matching algorithm is used to mine the intrinsic connection between the information, and the various parts of the data that may constitute sensitive information after combination can be identified, that is, it is not limited to traditional keyword matching, and can mine the complex associations between power data, efficiently identify new sensitive fields composed of multiple keywords, and improve the accuracy of sensitive data identification; and when dealing with massive and complex data in the power system based on the lightweight VF3 algorithm, it can quickly and accurately match and identify sensitive data to ensure identification efficiency.

[0172] Example 2

[0173] Based on the knowledge base-based power sensitive data identification method provided by the present invention, this embodiment provides an experiment of the sensitive data identification method of this solution. The hardware and software environment of the experiment are shown in Table 1:

[0174] Table 1 Software experimental environment

[0175]

[0176]

[0177] The experiment compared the execution efficiency of sensitive information identification methods developed based on different graph matching algorithms when processing smaller-scale input information graphs. The experimental environment is shown in Table 1, and Python 3.11 is used for programming. This embodiment selects the VF3-Light lightweight algorithm and three commonly used graph matching algorithms, namely VF3, LAD (Local AlignmentDescriptor, a local alignment descriptor method for graph matching) and PathLAD (an extension of the LAD algorithm that combines path information for graph matching), and tests their average response time when processing target graphs with different numbers of nodes. The target graph of each group uses 10 different types of sensitive information with similar numbers of nodes, and the average delay is obtained after repeating the matching algorithm 50 times.

[0178] Figure 8 It is the execution time of different recognition algorithms (Execution Time ofDifferentRecognitionAlgorithms), the horizontal axis represents the number of nodes in the target graph (Node Number ofTargetGraph), and the vertical axis represents the execution time (Execution Time). Figure 8 The experimental results show that VF3-Light Recognition Algorithm performs well in most cases, especially when processing graphs with a large number of nodes, its performance advantage is more obvious. In contrast, the execution time of VF3 Recognition Algorithm is relatively stable, and it performs well when the number of nodes is small, but its response time gradually increases as the number of nodes increases. LAD Recognition Algorithm and PathLAD Recognition Algorithm take a long time to execute when processing graphs with more nodes, and are at a disadvantage in processing such sensitive information recognition tasks. In summary, VF3-Light algorithm is a better choice in practical applications, and it performs better when processing graph matching tasks with smaller and sparser graphs.

[0179] Example 3

[0180] Based on the same inventive concept, the present invention also provides a knowledge base-based power sensitive data identification system, such as Fig. 9 As shown, including:

[0181] A graph structure conversion module, used to obtain the power data uploaded by the user, and based on the tags corresponding to the power data, convert the power data into a graph structure to obtain an input information graph;

[0182] A graph structure matching module, used to use a lightweight VF3 algorithm to perform correlation matching between the input information graph and each sensitive graph in the sensitive information knowledge base, so as to obtain a data correlation relationship between the input information graph and each of the sensitive graphs;

[0183] A sensitive data identification module, used to determine whether the power data is power sensitive data based on the data association relationship between the input information graph and each of the sensitive graphs;

[0184] The sensitive information knowledge base is constructed based on a plurality of the sensitive graphs, and the sensitive graphs are obtained based on a plurality of data elements constituting sensitive information in the power transaction data and the connections between the plurality of the data elements.

[0185] In a possible implementation, the graph structure matching module is specifically used to:

[0186] Collecting electricity transaction data in the power system;

[0187] Classifying a plurality of data elements constituting sensitive information in the power transaction data according to the relationship between the data elements, to obtain a group of the data elements corresponding to each piece of sensitive information;

[0188] Based on each group of the data elements, each of the data elements in the group is used as a node, and the connection between each of the data elements in the group is used as an edge to form a sensitivity graph;

[0189] The sensitive graph formed by each group of data elements is stored in a sensitive information knowledge base.

[0190] In a possible implementation, the graph structure matching module is specifically used to:

[0191] Extracting power transaction entities from power transaction data;

[0192] Classifying the power transaction data into multiple types according to their importance, and classifying each type of power transaction data into multiple levels according to sensitive information contained in the power transaction data;

[0193] Using an entity-relationship joint extraction algorithm in combination with the power transaction subject, type and level of the power transaction data, identifying entities and relationships in the power transaction data;

[0194] Determine, using a graph neural network based on entities and relationships in the power transaction data, a number of data elements constituting sensitive information in the power transaction data;

[0195] Based on the plurality of data elements, and entities and relationships in the power transaction data, connections between the plurality of data elements are obtained.

[0196] In a possible implementation, the graph structure matching module is specifically used to:

[0197] Based on each of the sensitive graphs, the node sorting function is used to sort the nodes of the sensitive graph to obtain the node order corresponding to the sensitive graph; the pattern graph preprocessing is performed on the sensitive graph to generate the state and parent node structure of the sensitive graph; based on the data structure representation, state, node order, parent node structure of the sensitive graph and the data structure representation of the input information graph, a matching function is used to perform recursive matching to obtain the matching result of the sensitive graph and the input information graph; the matching result of the sensitive graph and the input information graph is stored in the matching result set of the sensitive graph;

[0198] Storing the matching result set of each of the sensitive graphs into the total matching result set;

[0199] According to the total set of matching results, a data association relationship between the input information graph and each of the sensitive graphs is obtained.

[0200] In a possible implementation, the graph structure matching module is specifically used to:

[0201] Check whether the current state of the sensitive map is the target state;

[0202] If it is the target state, storing the current matching result of the sensitive map into the matching result set;

[0203] If it is not the target state, then based on the candidate node pairs generated by the current parameters, the current state is updated to obtain a new state, and the new state and the current matching result (if the matching function has not been used for matching, the current matching result is empty) are passed to the matching function for recursive call processing to obtain a new matching result; the new state is restored to the current state, the candidate node pairs are reselected to generate a new state, and the new state and the new matching result are continued to be passed to the matching function for recursive call until the termination condition is reached;

[0204] The termination condition is that the new state is the target state, or all candidate node pairs are traversed;

[0205] The current parameters are the data structure representation, state, node order, parent node structure of the sensitive graph and the data structure representation of the input information graph.

[0206] In a possible implementation, the graph structure matching module is specifically used to:

[0207] Using a candidate node generation function, a plurality of candidate node pairs are generated based on current parameters of the sensitivity graph;

[0208] Using a preset feasibility condition to screen the candidate node pairs, and obtain a plurality of candidate node pairs that meet the feasibility condition;

[0209] A candidate node pair is selected from a plurality of candidate node pairs that meet feasibility conditions, and a current state of the sensitive graph is expanded to obtain a new state.

[0210] In a possible implementation, the graph structure matching module is specifically used to:

[0211] Based on the node set and edge set of the input information graph, a data structure representation of the input information graph is constructed;

[0212] Mapping the nodes of each sensitive graph in the input information graph and the sensitive information knowledge base to the node label set of the sensitive graph to obtain a point mapping function;

[0213] Mapping the edges of each sensitive graph in the input information graph and the sensitive information knowledge base to the edge label set of the sensitive graph to obtain an edge mapping function;

[0214] For each sensitive graph, a data structure representation of the sensitive graph is constructed based on a node set, an edge set, a node label set, an edge label set, a point mapping function, and an edge mapping function of the sensitive graph.

[0215] In a possible implementation, the sensitive data identification module is specifically used to:

[0216] If the data association relationship includes at least one pair of matching node pairs formed by nodes in the sensitive graph and the input information graph, the power data is determined to be power sensitive data.

[0217] Example 4

[0218] like Fig.10 As shown, the present invention also provides an electronic device, which may be a computer device, a single-chip device, an intelligent mobile device, etc. The electronic device in this embodiment may include a processor, a memory, a transceiver component, etc. The memory, the processor, and the transceiver component are connected via a bus; the memory may be used to store an execution program, and an exemplary execution program may include instructions; the processor is used to execute the instructions stored in the memory. The memory may also be used to store data, which may be called and / or modified when the instructions are executed.

[0219] The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. It is the computing core and control core of the terminal, which is suitable for implementing one or more instructions, specifically suitable for loading and executing one or more instructions in a storage medium to implement the corresponding method flow or corresponding functions, so as to implement the steps of a knowledge base-based power-sensitive data identification method in the above-mentioned embodiment.

[0220] Example 5

[0221] Based on the same inventive concept, the present invention also provides a readable storage medium, specifically an electronic device readable storage medium (Memory), which is a memory device in an electronic device for storing programs and data. It can be understood that the storage medium here can include both built-in storage media in electronic devices and, of course, extended storage media supported by electronic devices. The storage medium provides a storage space, which stores the operating system of the terminal. In addition, one or more instructions suitable for being loaded and executed by a processor are also stored in the storage space, and these instructions can be one or more execution programs (including program codes). It should be noted that the storage medium here can be a high-speed RAM memory or a non-volatile memory, such as at least one disk storage. The processor loads and executes one or more instructions stored in the storage medium, which can implement the steps of a knowledge base-based power-sensitive data identification method in the above embodiment.

[0222] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0223] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0224] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0225] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0226] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit its protection scope. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that after reading the present invention, those skilled in the art can still make various changes, modifications or equivalent substitutions to the specific implementation methods of the application, but these changes, modifications or equivalent substitutions are all within the protection scope of the claims to be approved.

Claims

1. A method for identifying power sensitive data based on a knowledge base, characterized in that: include: Obtaining power data uploaded by a user, and converting the power data into a graph structure based on tags corresponding to the power data to obtain an input information graph; Using a lightweight VF3 algorithm, the input information graph is matched with each sensitive graph in the sensitive information knowledge base to obtain a data association relationship between the input information graph and each sensitive graph; Determining whether the power data is power sensitive data based on a data association relationship between the input information graph and each of the sensitive graphs; The sensitive information knowledge base is constructed based on a plurality of the sensitive graphs, and the sensitive graphs are obtained based on a plurality of data elements constituting sensitive information in the power transaction data and the connections between the plurality of the data elements.

2. The method according to claim 1, characterized in that The sensitive information knowledge base is constructed in the following manner: Collect power transaction data in the power system; Classifying a plurality of data elements constituting sensitive information in the power transaction data according to the relationship between the data elements, to obtain a group of the data elements corresponding to each piece of sensitive information; Based on each group of the data elements, each of the data elements in the group is used as a node, and the connection between each of the data elements in the group is used as an edge to form a sensitivity graph; The sensitive graph formed by each group of data elements is stored in a sensitive information knowledge base.

3. The method according to claim 1 or 2, characterized in that The plurality of data elements constituting sensitive information in the power transaction data and the relationship between the plurality of data elements are obtained in the following manner: Extracting power transaction entities from power transaction data; Classifying the power transaction data into multiple types according to their importance, and classifying each type of power transaction data into multiple levels according to sensitive information contained in the power transaction data; Using an entity-relationship joint extraction algorithm in combination with the power transaction subject, type and level of the power transaction data, identifying entities and relationships in the power transaction data; Determine, using a graph neural network based on entities and relationships in the power transaction data, a number of data elements constituting sensitive information in the power transaction data; Based on the plurality of data elements, and entities and relationships in the power transaction data, connections between the plurality of data elements are obtained.

4. The method according to claim 1, characterized in that The lightweight VF3 algorithm is used to perform correlation matching between the input information graph and each sensitive graph in the sensitive information knowledge base to obtain a data correlation relationship between the input information graph and each sensitive graph, including: Based on each of the sensitive graphs, the node sorting function is used to sort the nodes of the sensitive graph to obtain the node order corresponding to the sensitive graph; the pattern graph preprocessing is performed on the sensitive graph to generate the state and parent node structure of the sensitive graph; based on the data structure representation, state, node order, parent node structure of the sensitive graph and the data structure representation of the input information graph, a matching function is used to perform recursive matching to obtain the matching result of the sensitive graph and the input information graph; the matching result of the sensitive graph and the input information graph is stored in the matching result set of the sensitive graph; Storing the matching result set of each of the sensitive graphs into the total matching result set; According to the total set of matching results, a data association relationship between the input information graph and each of the sensitive graphs is obtained.

5. The method according to claim 4, characterized in that The recursive matching is performed based on the data structure representation, state, node order, parent node structure of the sensitive graph and the data structure representation of the input information graph using a matching function to obtain a matching result between the sensitive graph and the input information graph, including: Check whether the current state of the sensitive map is the target state; If it is the target state, storing the current matching result of the sensitive map into the matching result set; If it is not the target state, then based on the candidate node pairs generated by the current parameters, update the current state to obtain a new state, pass the new state and the current matching result to the matching function for recursive call processing to obtain a new matching result; restore the new state to the current state, reselect the candidate node pairs to generate a new state, and continue to pass the new state and the new matching result to the matching function for recursive call processing until the termination condition is reached; The termination condition is that the new state is the target state, or all candidate node pairs are traversed; The current parameters are the data structure representation, state, node order, parent node structure of the sensitive graph and the data structure representation of the input information graph.

6. The method according to claim 5, characterized in that The candidate node pairs generated based on the current parameters update the current state to obtain a new state, including: Using a candidate node generation function, a plurality of candidate node pairs are generated based on current parameters of the sensitivity graph; Using a preset feasibility condition to screen the candidate node pairs, and obtain a plurality of candidate node pairs that meet the feasibility condition; A candidate node pair is selected from a plurality of candidate node pairs that meet feasibility conditions, and a current state of the sensitive graph is expanded to obtain a new state.

7. The method according to claim 4 or 5, characterized in that The data structure representation of the input information graph and each of the sensitivity graphs is obtained in the following manner: Based on the node set and edge set of the input information graph, a data structure representation of the input information graph is constructed; Mapping the nodes of each sensitive graph in the input information graph and the sensitive information knowledge base to the node label set of the sensitive graph to obtain a point mapping function; Mapping the edges of each sensitive graph in the input information graph and the sensitive information knowledge base to the edge label set of the sensitive graph to obtain an edge mapping function; For each sensitive graph, a data structure representation of the sensitive graph is constructed based on a node set, an edge set, a node label set, an edge label set, a point mapping function, and an edge mapping function of the sensitive graph.

8. The method according to claim 1, characterized in that The determining whether the power data is power sensitive data based on the data association relationship between the input information graph and each of the sensitive graphs includes: If the data association relationship includes at least one pair of matching node pairs formed by nodes in the sensitive graph and the input information graph, the power data is determined to be power sensitive data.

9. A knowledge base-based power sensitive data identification system, characterized in that: include: A graph structure conversion module, used to obtain the power data uploaded by the user, and based on the tags corresponding to the power data, convert the power data into a graph structure to obtain an input information graph; A graph structure matching module, used to use a lightweight VF3 algorithm to perform correlation matching between the input information graph and each sensitive graph in the sensitive information knowledge base, so as to obtain a data correlation relationship between the input information graph and each sensitive graph; A sensitive data identification module, used to determine whether the power data is power sensitive data based on the data association relationship between the input information graph and each of the sensitive graphs; The sensitive information knowledge base is constructed based on a plurality of the sensitive graphs, and the sensitive graphs are obtained based on a plurality of data elements constituting sensitive information in the power transaction data and the connections between the plurality of the data elements.

10. The system according to claim 9, characterized in that The graph structure matching module is specifically used for: Collect power transaction data in the power system; Classifying a plurality of data elements constituting sensitive information in the power transaction data according to the relationship between the data elements, to obtain a group of the data elements corresponding to each piece of sensitive information; Based on each group of the data elements, each of the data elements in the group is used as a node, and the connection between each of the data elements in the group is used as an edge to form a sensitivity graph; The sensitive graph formed by each group of data elements is stored in a sensitive information knowledge base.

Images