Data link network guarantee demand processing method and system

Abstract

The invention provides a data link network guarantee demand processing method and system, and the method comprises the steps: setting a database, integrating a plurality of AI algorithm models based on service calling, and achieving the knowledge extraction, knowledge recommendation and knowledge query of a data link network guarantee demand; the database comprises a plurality of guarantee schemes and sample knowledge maps corresponding to the guarantee schemes; the AI algorithm model at least comprises a knowledge extraction model, a knowledge recommendation model and a knowledge query model; each AI algorithm model is packaged in a service form; creating a process pool, wherein one process in the process pool correspondingly loads one AI algorithm model; and selecting an idle process from the process pool through the service request to call the AI algorithm model, and feeding back a call result of the AI algorithm model through the service response. According to the method, extraction, recommendation and query are integrated, an intelligent auxiliary capability is provided for data link network design planners when data link network guarantee requirements are acquired and analyzed, and the working efficiency is effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of information processing technology, and in particular to a method and system for processing data link network security requirements. Background Art

[0002] The data link is a key system for realizing the deep connection between the command and control platform, weapon platform and sensor platform, and is the core system for improving the agility and accuracy of the "detection, control, attack and evaluation" combat loop in joint operations. Different from general military wireless communication systems, the data link is closely related to combat operations and is a communication system customized for combat missions. In order to provide real-time situation information sharing and rapid combat capabilities from sensors to weapon platforms, it is necessary to conduct communication demand analysis and scientific and detailed network design planning based on specific combat plans before the start of combat operations. Whether the data link network design planning can be completed efficiently and accurately is the key to realizing the networking and combat application of the data link system, and is the premise for applying the data link system to enhance the overall confrontation capability of the combat system and improve the combat effectiveness of the weapon system. The collection and analysis of data link network security requirements, as the starting point of data link network design planning, is an extremely important link in data link network design planning.

[0003] At present, some experts have proposed using LSTM neural networks to automatically extract data link network support requirements from combat plan documents. However, since data is the main factor affecting the performance of data-driven intelligent algorithms, and there is relatively little data in the field of intelligent extraction of data link network support requirements, directly using existing knowledge extraction methods will result in insufficient data and cause data link network performance loss.

[0004] Other experts proposed using graph similarity calculation to achieve intelligent recommendation of data link network assurance solutions. However, this method did not design a corresponding graph similarity calculation method based on the characteristics of the data link network assurance requirement knowledge graph, resulting in the problem of low recommendation accuracy when using a general graph similarity calculation method.

[0005] Some experts have also proposed using a machine reading model to provide intelligent query capabilities for data link network security needs. This method can only find answers in the text by extraction. If the answer does not exist in the text or cannot be obtained directly, and the answer can only be obtained after statistics, induction, reasoning, etc. based on the text information content, then this method cannot be used to query and obtain the correct answer.

[0006] In summary, the existing collection and analysis of data link network security requirements only considers partial processing of data link network security requirements, such as individual processing methods of intelligent extraction, intelligent recommendation and intelligent query, but lacks system design that integrates the three modules of intelligent extraction, intelligent recommendation and intelligent query; this results in data link network design and planning personnel lacking effective intelligent assistance when collecting and analyzing data link network security requirements, and resulting in low design and planning efficiency. Summary of the invention

[0007] The present invention provides a data link network security demand processing method and system, which are used to solve the problem that existing data link network design and planning personnel lack effective intelligent assistance when collecting and analyzing data link network security demands, and the design and planning efficiency is low.

[0008] On the one hand, the present invention provides a method for processing data link network assurance requirements, the method comprising: setting a database and integrating multiple AI algorithm models based on service calls to realize knowledge extraction, knowledge recommendation and knowledge query of data link network assurance requirements; The database includes multiple protection schemes and sample knowledge graphs corresponding to the protection schemes; the AI ​​algorithm model includes at least a knowledge extraction model, a knowledge recommendation model and a knowledge query model; each of the AI ​​algorithm models is packaged in the form of a service; A process pool is created, in which one process in the process pool corresponds to loading an AI algorithm model; an idle process is selected from the process pool through a service request to call the AI ​​algorithm model, and the calling result of the AI ​​algorithm model is fed back through a service response.

[0009] Furthermore, the knowledge extraction model is used to receive a target document, extract demand elements from the target document, and form a target knowledge graph; Among them, the construction of the knowledge extraction model includes: using the existing extraction model to extract entities from the target text, obtain triple data, and generate a triple knowledge base; using a graph representation learning method to generate entity vector encodings corresponding to each entity in the triple knowledge base; using the entity vector encoding as a supplementary enhancement means for entity semantic information to improve the existing extraction model and obtain the knowledge extraction model.

[0010] Furthermore, the knowledge extraction method includes: Receiving a target document, reading each text paragraph, generating a text paragraph list, and dividing the text paragraph list into a plurality of equal parts to obtain a plurality of text paragraph sublists; Traversing multiple processes in the process pool in sequence, passing a text paragraph sublist to a process, and having the current process call the knowledge extraction model to extract sublist requirement elements from the current text paragraph sublist; After waiting for all processes in the process pool to complete the extraction, the triple data lists of all sub-list requirement elements are concatenated to obtain the requirement elements of the target document.

[0011] Furthermore, the knowledge recommendation model is used to calculate the similarity value between the target knowledge graph and each of the sample knowledge graphs in the database, and output the protection plan corresponding to the sample knowledge graph with the highest similarity value.

[0012] Furthermore, calculating the similarity value between the target knowledge graph and each of the sample knowledge graphs in the database includes: Based on the basic framework of the SimGNN model, the edges in the knowledge graph are set as special nodes in the knowledge graph; and on the basis of the graph topology structure difference characteristics of the SimGNN model, the combat area difference characteristics and the combat force difference characteristics are added.

[0013] Furthermore, the added difference characteristics of the combat area and the difference characteristics of the participating forces include: Input the target knowledge graph and any sample knowledge graph, and perform first-path processing, second-path processing, and third-path processing; The first processing path includes: after edge-to-node processing in the knowledge graph, edge coding and adding relationship information processing on the edge, inputting into the SimGNN model, and obtaining the topological structure similarity between the two knowledge graphs based on the SimGNN model; The second processing path includes: extracting geographic information from the knowledge graph, setting any geographic location as a reference point, respectively calculating the distance between each geographic location in the knowledge graph and the reference point, and obtaining a distance vector of the reference point; and inputting the distance vector into the Transformer network, extracting the geographic spatial information of the two knowledge graphs, and obtaining the geographic spatial similarity between the two knowledge graphs; The third processing path includes: extracting the force information in the knowledge graph, classifying and counting the number of participating troops of each type of arms, and obtaining a force quantity vector; and inputting the force quantity vector into the Transformer network to obtain the force similarity between the two knowledge graphs; The topological structure similarity, the geographic space similarity and the military force similarity are spliced ​​and input into the Dense network to obtain the similarity value between the target knowledge graph and the sample knowledge graph.

[0014] Furthermore, the knowledge recommendation model is used to receive natural language questions from users, convert the natural language questions into the query language of the database, execute the query language on the database, generate answers to the natural language questions, and complete the query of the guarantee plan and the corresponding sample knowledge graph.

[0015] Furthermore, converting the natural language question into the query language of the database includes: Establishing a template library, wherein the template library includes standard question templates of multiple natural language questions; Processing the semantic information of the natural language question sentence to obtain question sentence semantic coding information, and determining which standard question sentence template the natural language question sentence belongs to based on the semantic coding information; Extracting multiple key entities from the natural language question; The key entity is filled into the standard question template to generate an instantiated query language of the database.

[0016] Furthermore, filling the key entity into the standard question template includes: Determine whether the key entity exists in the sample knowledge graph; If it exists, directly fill the key entity into the standard question template; If it does not exist, the edit distance between the key entity and each sample entity in the sample knowledge graph is calculated, and the key entity is replaced by the sample entity with the smallest edit distance in the sample knowledge graph, and then the sample entity is filled into the standard question template.

[0017] On the other hand, the present invention also provides a data link network assurance demand intelligent processing system, which includes at least a database, a knowledge extraction module, a knowledge query module, and a knowledge recommendation module, and is used to execute the steps of any of the above methods.

[0018] In general, the technical solution conceived by the present invention can achieve the following beneficial effects compared with the prior art: (1) The present invention provides a data link network security demand processing method and system, which integrates three modules: intelligent extraction, intelligent recommendation and intelligent query, and provides intelligent assistance capabilities for data link network design planners when collecting and analyzing data link network security requirements, effectively improving the work efficiency of data link network design planners. In addition, the operating efficiency of the AI ​​model can be improved by multi-core concurrency and the shared process pool can be used to effectively improve the operating efficiency of the system.

[0019] (2) The present invention provides a data link network security demand processing method and system, which utilizes a knowledge-enhanced knowledge extraction model to improve the performance of subsequent knowledge extraction models through the knowledge base knowledge formed after knowledge extraction, thereby improving the performance of the knowledge extraction model.

[0020] (3) The present invention provides a data link network security demand processing method and system, which realizes intelligent knowledge recommendation through hybrid graph embedding coding, making the output security plan faster and more accurate.

[0021] (4) The present invention provides a data link network security requirement processing method and system, which realizes the knowledge query of data link network security requirements based on template matching and key information extraction, can convert user questions into database query statements, and generate answers to user questions after executing the database query statements. It not only solves the problem that the answer cannot be given if it is not in the text; but also can use the database query statement to explain and trace the generated answer, so that it has better interpretability and can improve the user's credibility of the answer. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0023] Figure 1 It is a schematic diagram of constructing a knowledge extraction model of a data link network assurance requirement processing method and system provided by the present invention; Figure 2 It is a principle schematic diagram of an existing relationship classification model of a data link network assurance requirement processing method and system provided by the present invention; Figure 3 It is a principle schematic diagram of a relational classification model of a data link network assurance requirement processing method and system provided by the present invention; Figure 4 It is a schematic diagram of the knowledge extraction process of a data link network assurance demand processing method and system provided by the present invention; Figure 5 It is a schematic diagram of a knowledge recommendation process of a data link network assurance requirement processing method and system provided by the present invention; Figure 6 It is a similarity calculation schematic diagram of a data link network assurance requirement processing method and system provided by the present invention; Figure 7 It is a schematic diagram of a specific embodiment of a knowledge query model of a data link network assurance demand processing method and system provided by the present invention; Figure 8 It is a system structure diagram of a data link network assurance demand processing method and system provided by the present invention. DETAILED DESCRIPTION

[0024] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings and embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work belong to the scope of protection of the present invention.

[0025] It should be noted that, in the description of the embodiments of the present invention, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method or system including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method or system. In the absence of further restrictions, an element defined by the sentence "includes a ..." does not exclude the existence of other identical elements in the process, method or system including the element.

[0026] The present invention proposes a method for processing data link network security requirements, which includes: setting a database and integrating multiple AI algorithm models based on service calls to achieve knowledge extraction, knowledge recommendation and knowledge query of data link network security requirements.

[0027] The database includes multiple protection schemes and sample knowledge graphs corresponding to the protection schemes. In the process of forming each protection scheme in the database, protection schemes are extracted to generate graphs, preferably based on the knowledge extraction model mentioned later.

[0028] The AI ​​algorithm model includes at least a knowledge extraction model, a knowledge recommendation model and a knowledge query model; each AI algorithm model is encapsulated in the form of a service.

[0029] It should be noted that the data link network security requirement knowledge extraction tool automatically extracts the requirement elements from the target document by calling the knowledge extraction model, and constructs the data link network security requirement knowledge graph; among them, the requirement elements include eight types of data link security requirement elements, such as task type, force allocation, command relationship, coordination relationship, force deployment, combat area, battlefield environment, and action plan. The data link network security requirement knowledge recommendation tool automatically calculates the matching degree between the target document and each data link network security requirement knowledge graph in the database by calling the knowledge recommendation model, and realizes the intelligent recommendation of the best data link network security solution in the database. The data link network security requirement knowledge query tool realizes intelligent question and answer on the data link network security requirement knowledge graph by calling the knowledge query model, and provides flexible database query and management capabilities for data link network design planners.

[0030] In the intelligent processing method of data link network security requirements, multiple AI algorithm models are needed to realize the data link network security requirements knowledge extraction, knowledge query and knowledge recommendation respectively. Since each AI algorithm model relies on many underlying libraries, which may be different from each other, integration together may cause conflicts in the underlying neural network modules. Therefore, during the integration design, the deployment process is complicated. The AI ​​algorithm model integration based on service call is adopted to encapsulate each AI algorithm model in the form of a service, isolate the operating environment of each AI algorithm model, and make each AI algorithm model run in its own service process space, thereby avoiding various conflicts that may occur in the same process space.

[0031] In addition, as an embodiment of the present invention, the structural design of the B / S mode simplifies the difficulty of software system deployment; the system interface is presented by relying on the browser, which further simplifies the difficulty of interface development work.

[0032] Create a process pool, in which one process in the process pool corresponds to loading an AI algorithm model; select an idle process from the process pool through a service request to call the AI ​​algorithm model, and feedback the calling result of the AI ​​algorithm model through a service response.

[0033] The intelligent processing method for data link network security needs requires the use of multiple AI algorithm models based on deep neural networks, which results in a large amount of computation and a long time consumption. As an embodiment of the present invention, the operating efficiency of the AI ​​model can be improved by multi-core concurrency; for example, the performance can be improved by using the 64-core domestic Feiteng 2000+ CPU through the advantages of CPU multi-core concurrency.

[0034] In addition, the loading process of the AI ​​algorithm model requires initializing the neural network and importing the network parameter file, and this process takes seconds. Therefore, the shared process pool is adopted to eliminate the time-consuming problem of initializing the AI ​​algorithm model when the AI ​​algorithm model is called. Specifically, a shared process pool is created at startup, and each process in the process pool loads an AI algorithm model and completes the initialization of the AI ​​algorithm model. Thereafter, each time the AI ​​algorithm model is called, only idle processes are selected from the shared process pool, thereby effectively improving operating efficiency.

[0035] The knowledge extraction model is used to receive the target document, extract the required elements from the target document, and form a target knowledge graph.

[0036] In the field of data link network security demand knowledge extraction, there is a problem of insufficient data, which leads to the loss of accuracy of the existing data-driven knowledge extraction model. In order to solve this problem, the present invention uses a knowledge-enhanced knowledge extraction model and uses the entity pairs with objective meanings and the relationships between them in the triple knowledge base formed after knowledge extraction to improve the performance of the knowledge extraction model.

[0037] Specifically, the knowledge extraction model is constructed as follows: Figure 1 As shown, it specifically includes: first, using the existing extraction model to extract entities from the target text, obtain triple data, and generate a triple knowledge base; then, using a graph representation learning method, such as TransE, to generate entity vector encodings corresponding to each entity in the triple knowledge base; finally, using entity vector encodings as a supplementary enhancement method for entity semantic information to improve the existing extraction model and obtain a knowledge extraction model.

[0038] It should be noted that existing extraction models are usually implemented in two steps: named entity recognition and relationship classification.

[0039] Among them, the named entity recognition model is responsible for extracting entity pairs from the text, namely a subject word and an object word, and then using the relational classification model to determine the relationship between the entity pairs in the current text environment.

[0040] For example, in the text "the air force early warning aircraft and fighter jets, and the navy missile destroyers and minesweepers constitute a command relationship and a coordination relationship respectively", after using the named entity recognition model, two entity pairs are obtained, namely [early warning aircraft, fighter jets] and [missile destroyers, minesweepers]; then, the relational classification model is used to determine the relationship between these two entity pairs.

[0041] like Figure 2 As shown in the figure, when determining the relationship of an entity pair, the existing relationship classification model mainly determines the semantic information of the text in which the entity pair is located, the semantic information of the subject word, and the semantic information of the object word. The semantic information of the text can be obtained by concatenating the word vectors, character vectors, and part-of-speech vectors in the text, or by using pre-trained language models such as Google BERT and Baidu ERNIE to extract the semantic information of the text; and then using CNN convolutional neural network and LSTM long short-term memory neural network to process it. The semantic information of the subject word and the semantic information of the object word are usually directly expressed by the text semantic information of the subject word and the object word at the position of the text.

[0042] However, in the text "the air force early warning aircraft and fighter jets, the navy missile destroyers and minesweepers respectively form a command relationship and a coordination relationship", the existing relationship classification model is difficult to accurately determine the relationship between the two entity pairs [early warning aircraft, fighter jets] and [missile destroyers, minesweepers], because the two relationship words "command relationship" and "coordination relationship" are close together, and only the word "respectively" indicates that the two relationships correspond to the above two entity pairs. But if combined with the knowledge in the knowledge base extracted previously, a common sense knowledge will be obtained, that is, the early warning aircraft and fighter jets will not form a coordination relationship, and the missile destroyers and minesweepers will not form a command relationship.

[0043] Therefore, the present invention proposes to improve the relationship classification model, such as Figure 3 As shown in the figure, graph representation learning methods such as TransE are used to encode the knowledge vectors of each entity in the knowledge base, and then these entity vector encoding information is used as a supplement to the entity semantic information to realize the knowledge-enhanced relational classification model. In this way, it is easy to determine that the relationship between the early warning aircraft and the fighter is a command relationship, and the relationship between the missile destroyer and the minesweeper is a collaborative relationship.

[0044] In addition, in order to make full use of multi-core concurrency and improve the performance of the data link network assurance demand knowledge extraction model, the knowledge extraction method includes: receiving the target document, reading each text paragraph, generating a text paragraph list, and dividing the text paragraph list into multiple equal parts to obtain multiple text paragraph sublists; traversing multiple processes in the process pool in turn, passing a text paragraph sublist to a process, and the current process calls the knowledge extraction model to extract the sublist demand elements from the current text paragraph sublist; after waiting for all processes in the process pool to complete the extraction, the triple data lists of all sublist demand elements are spliced ​​together to obtain the demand elements of the target document.

[0045] It should be noted that, based on the number of processes in the process pool, the text paragraph list is divided into equal parts that match the number of processes in the process pool.

[0046] As a specific embodiment of the present invention, Python's multiprocessing library is called to create a process pool containing 8 processes; each process in the process pool loads the knowledge extraction model and completes the initialization work of loading the relevant deep learning model and network parameters. Figure 4As shown. Receive the user's request to call the knowledge extraction model, call the Python-docx library, read each text paragraph from the docx format target document imported by the user, and generate a text paragraph list; and divide the text paragraph list read from the target document into 8 equal parts, that is, form 8 text paragraph sublists; traverse the 8 processes in the process pool in turn, pass one of the 8 text paragraph sublists to a process, and the process calls the knowledge extraction model to extract the demand elements from the text paragraph sublist; after waiting for all processes in the process pool to complete the extraction task, splice the triple data list of the demand elements extracted by each process and return it to the knowledge extraction module.

[0047] Furthermore, an idle process is selected from the process pool through a service request. Specifically: Determine whether there is an idle process; if so, call the current idle process to complete the extraction of demand elements from a text paragraph sublist; if not, continue to determine whether there are unfinished extraction tasks. If there are tasks, continue to wait until there are no tasks; if there are no tasks, concatenate the triple data lists of demand elements extracted by each process and feed them back to the knowledge extraction module.

[0048] The knowledge recommendation model is used to calculate the similarity between the target knowledge graph and each sample knowledge graph in the database, and output the security plan corresponding to the sample knowledge graph with the highest similarity value. In this way, the matching degree between the target document and each sample knowledge graph in the database is automatically calculated, and the intelligent recommendation of the best security plan in the database is realized.

[0049] Specifically, if Figure 5 As shown in the figure: after the user inputs the target document, the target document is first processed through the knowledge extraction model to generate the target knowledge graph A; then each guarantee plan in the database is traversed to obtain the sample knowledge graph B corresponding to the current guarantee plan; then the similarity between the target knowledge graph A and the sample knowledge graph B is calculated by calling the knowledge recommendation algorithm; it should be noted that if the similarity between the current target knowledge graph A and the sample knowledge graph B is higher than the similarity calculated previously, indicating that a more matching guarantee plan is obtained, then based on the most matching similarity value, the current guarantee plan that best matches the target document is updated until each guarantee plan in the database is traversed and the guarantee plan that best matches the current target document is obtained.

[0050] Currently, the calculation of the similarity between two knowledge graphs is based on the SimGNN model, which implements graph embedding encoding through graph topology information to calculate graph similarity. However, when the SimGNN model performs graph embedding encoding, it only considers the nodes in the graph and does not consider the information on the edges, which leads to information loss when embedding the graph on the knowledge graph, affecting the accuracy of graph similarity calculation.

[0051] In order to more accurately calculate the similarity between two knowledge graphs, the present invention performs calculations based on a knowledge recommendation model of hybrid graph embedding coding.

[0052] On the one hand, the edge encoding method in the SimGNN model is improved. Specifically, based on the basic framework of the SimGNN model, the edges in the knowledge graph are set as special nodes in the knowledge graph.

[0053] In the prior art, when processing the edges in the graph, an edge is encoded by the two node information on the edge, but the specific relationship information represented by the edge is discarded. For example, there is a triple <escort mission, start time, January 1, 2023> in the target knowledge graph A, and a triple <escort mission, end time, January 1, 2023> in the sample knowledge graph B. Obviously, there are certain differences between the target knowledge graph A and the sample knowledge graph B, one with a start time of January 1, 2023 and the other with an end time of January 1, 2023. However, when using the SimGNN model for graph embedding encoding, the edge encoding input of the target knowledge graph A is [<escort mission, January 1, 2023>], and the edge encoding input of the sample knowledge graph B is also [<escort mission, January 1, 2023>], which will result in the target knowledge graph A and the sample knowledge graph B having completely consistent encoding information, so the calculated similarity values ​​are also exactly the same. The reason for this problem is that the SimGNN model mainly relies on node encoding information. The encoding information of an edge is only composed of the encoding information of the two nodes associated with the edge, and the specific relationship information on the edge is discarded, resulting in the above two knowledge graphs becoming indistinguishable.

[0054] In order to solve the problem that the target knowledge graph A and the sample knowledge graph B are different but their encoding information is completely consistent, the present invention regards the edges as special nodes in the graph based on the basic framework of the SimGNN model.

[0055] For example, in the triple <escort mission, end time, January 1, 2023> of the sample knowledge graph B, the relationship name "end time" is also regarded as a node in the graph. Therefore, the edge encoding represented by the triple is changed from the original [<escort mission, January 1, 2023>] to [<escort mission, end time>, <end time, January 1, 2023>, <escort mission, January 1, 2023>]; similarly, the triple <escort mission, start time, January 1, 2023> of the target knowledge graph A is encoded into [<escort mission, start time>, <start time, January 1, 2023>, <escort mission, January 1, 2023>]; thus, the encoding inputs of the two knowledge graphs become different and more in line with the actual situation.

[0056] That is, the edge coding in SimGNN is changed from [<edge node 1, edge node 2>] to [<edge node 1, edge relationship name>, <edge relationship name, edge node 2>, <edge node 1, edge node 2>]. It should be noted that the present invention only regards the edge relationship name as a special node on the graph, and its coding information is still consistent with the node coding method in the existing graph.

[0057] On the other hand, since the existing SimGNN model simply uses the graph topology structure in the knowledge graph to calculate graph similarity, when it is directly applied to the knowledge graph similarity calculation, it does not involve the required domain knowledge, which will lead to poor accuracy of similarity calculation.

[0058] Therefore, the present invention adds operational area difference characteristics and combat force difference characteristics based on the graph topology difference characteristics of the SimGNN model.

[0059] Specifically, if Figure 6 As shown in the figure, the added characteristics of differences in combat areas and participating forces include: Input the target knowledge graph and any sample knowledge graph, and perform first-path processing, second-path processing, and third-path processing; The first processing path includes: edge-to-node processing in the knowledge graph, edge coding to add relationship information processing on the edge, input into the SimGNN model, and obtain the topological structure similarity between the two knowledge graphs based on the SimGNN model; The second processing path includes: extracting geographic information from the knowledge graph, setting any geographic location as a reference point, calculating the distance between each geographic location in the knowledge graph and the reference point, and obtaining the distance vector of the reference point; and inputting the distance vector into the Transformer network, extracting the geographic spatial information of the two knowledge graphs, and obtaining the geographic spatial similarity between the two knowledge graphs; The third processing path includes: extracting the force information in the knowledge graph, classifying and counting the number of participating troops of each branch, and obtaining the force quantity vector; and inputting the force quantity vector into the Transformer network to obtain the force similarity between the two knowledge graphs; The topological structure similarity, geographic space similarity and military strength similarity are concatenated and input into the Dense network to obtain the similarity value between the target knowledge graph and the sample knowledge graph.

[0060] During the test, the security solution recommended by the method of the present invention was more accurate, reaching 92.37%, while the accuracy of the security solution recommended in the prior art was only 83.54%.

[0061] The knowledge recommendation model is used to receive natural language questions from users, convert natural language questions into database query language, execute query language on the database, generate answers to natural language questions, and complete the query of guarantee solutions and corresponding sample knowledge graphs. This enables intelligent question-answering on the knowledge graph of data link network guarantee requirements, and provides flexible database query and management capabilities for data link network design planners.

[0062] In the prior art, models based on machine reading comprehension can only find answers in text by extraction. If the answer does not exist in the text, the correct answer cannot be given.

[0063] For example, given a text and a user's question; the text is: "On the morning of January 1, 2023, the first, second, and third ships formed the 44th escort formation and set sail from a certain military port in a certain city to the first and second locations to replace the 43rd escort formation to perform escort missions. Among them, the first ship is the X-type a ship, hull number 1, and the service time is January 1, 2021; the second ship is the Y-type b ship, hull number 2, and the service time is January 1, 2020; the third ship is the Z-type c ship, hull number 3, and the service time is January 1, 2010." The user's question is: "Which ship has the longest service time in the 44th escort formation?" Since there is no such description in the text, the model based on machine reading comprehension cannot accurately answer such a question.

[0064] Therefore, the present invention is based on a knowledge query model of template matching and key information extraction, that is, converting the user's natural language question into the query language of the database, and then generating the answer to the user's question by executing the query language on the database.

[0065] For example, after completing the extraction of data link network assurance requirements knowledge from the text, a knowledge graph containing the following triple information will be obtained: <43rd escort formation, formation relationship, first ship>, <43rd escort formation, formation relationship, second ship>, <nth escort formation, formation relationship, third ship>, <first ship, hull number, 1>, <first ship, model, X-type a ship>, <first ship, service time, January 1, 2021>, <second ship, hull number, 2>, <second ship, model, Y-type b ship>, <second ship, service time, January 1, 2020>, <third ship, hull number, 3>, <third ship, model, Z-type c ship>, <third ship, service time, January 1, 2010>. By executing the query statement "MATCH (n)-[: formation relationship]->(m)-[: service time]->(q) WHERE n.name='escort formation' RETURN q.name ORDER BY q.name LIMIT 1" in the knowledge graph, the answer to the user's question can be obtained, which solves the existing technical problems.

[0066] Since the number of data samples in the field of intelligent processing of data link network security requirements is small, it is difficult to train a neural network model. Therefore, the present invention pre-sets the user's question as a question template according to the actual use of the user.

[0067] Specifically: converting a natural language question into a query language of a database includes: establishing a template library, the template library includes a variety of standard question templates of natural language questions; processing the semantic information of the natural language question to obtain question semantic coding information, and judging which standard question template the natural language question belongs to based on the question semantic coding information; extracting the natural language question to obtain multiple key entities; filling the key entities into the standard question template to generate an instantiated database query language.

[0068] As a specific embodiment of the present invention, Figure 7 As shown, after entering the user's natural language question: (1) Extracting the user's semantic information to obtain question semantic encoding information. As an implementation of the present invention, the user's question semantic information is first extracted using the semantic extraction layer 1, for example, using the BERT pre-trained language model; and then the user's question semantic information is secondly extracted using the semantic extraction layer 2 to obtain question semantic encoding information; wherein the semantic extraction layer 2 includes a BiLSTM (bidirectional long short-term neural network) with 300 hidden neurons.

[0069] (2) In the user's question semantic encoding information, the output [CLS] tag position is subjected to question template classification operation to obtain the standard question template corresponding to the user's natural language question. For example, for the user's question "What is the action route of the escort formation?", the matching Cypher graph database query statement question template is "MATCH (n)-[: formation relationship]->(m)-[: service time]->(q) WHERE n.name='key entity' RETURN q.name ORDER BYq.name LIMIT 1"; it should be noted that all template libraries need to be defined in advance, and then the corresponding standard question template is searched from the template library list according to the classification results.

[0070] (3) Using the key entity start position classification layer, the user's question semantic encoding information is serialized and classified to obtain the key entity start position; using the key entity end position classification layer, the user's question semantic encoding information is serialized and classified to obtain the key entity end position; combining the key entity start position and the key entity end position, the key entity is obtained from the user's question, such as Figure 7 The "Escort Formation" in.

[0071] (4) Fill the key entities into the standard question template to generate an instantiated database query language. That is, assemble the key entities and question template to obtain the database query statement corresponding to the user's question; Figure 6 "MATCH (n)-[: formation relationship]->(m)-[: service time]->(q) WHERE n.name='escort formation' RETURN q.nameORDER BY q.name LIMIT 1"; (5) Finally, execute the query statement in the database to get the answer to the user's question.

[0072] It should be noted that when extracting key entities, if the extracted key entities do not exist in the knowledge graph, the generated query statement may not obtain any results after being executed on the graph database.

[0073] Therefore, the present invention further includes: determining whether the key entity exists in the sample knowledge graph; if so, directly filling the key entity into the standard question template; if not, calculating the edit distance between the key entity and each sample entity in the sample knowledge graph, and replacing the key entity with the sample entity with the smallest edit distance in the sample knowledge graph, and then filling the sample entity into the standard question template. In this way, the generated answer can be explained and traced using a database query statement, thereby having better interpretability and improving the user's credibility of the answer.

[0074] As a preferred embodiment of the present invention, the system uses the open source graph database Neo4j and the graph database query language Cypher in the selection of the database query language. Specifically, the specific process of executing the Cypher graph database query statement on the Neo4j graph database to obtain the result of the user query is as follows: Determine whether to traverse each triplet data in the data link network design planning guarantee requirement triplet data list: If yes, call the py2neo library to add the subject and object in the current knowledge graph triple data as nodes to the Neo4j graph database respectively; then call the py2neo library again to add the current knowledge graph triple data as an edge to the Neo4j graph database; until each triple data in the triple data list is traversed; If not, the py2neo library is called to input the Cypher graph database query statement corresponding to the user's natural language question into the Neo4j graph database, and the user's query result is obtained by executing the Cypher graph database query statement in the Neo4j graph database.

[0075] In addition, in the knowledge query module, a service-based call method is used to implement the call of the knowledge query model to avoid underlying module conflicts caused by the simultaneous loading of various AI algorithm models.

[0076] As a preferred embodiment of the present invention, a RESTFUL-based service call is implemented by calling the flask library, and the specific process is: before the knowledge query module runs, the flask library is called to start the data link network security demand knowledge query service, and monitor the service request from the user in real time; when the knowledge query request from the user is received, the natural language question of the current query is obtained from the user request data, and the knowledge query model is called to convert the user's natural language question into a Cypher graph database query statement; finally, the flask library is called to return the obtained Cypher graph database query statement to the user.

[0077] On the other hand, the present invention also provides a data link network security demand intelligent processing system, which at least includes a database, a knowledge extraction module, a knowledge query module, and a knowledge recommendation module, and is used to execute the steps of any of the above methods. Figure 8 , which is a specific embodiment of the present invention, is consistent with the technical features of the above method and will not be described in detail here.

[0078] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods of each embodiment or some parts of the embodiment.

[0079] 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 it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for processing data link network security requirements, characterized in that: The method includes: setting up a database and integrating multiple AI algorithm models based on service calls to realize knowledge extraction, knowledge recommendation and knowledge query of data link network assurance requirements; The database includes multiple protection schemes and sample knowledge graphs corresponding to the protection schemes; the AI ​​algorithm model includes at least a knowledge extraction model, a knowledge recommendation model and a knowledge query model; each of the AI ​​algorithm models is packaged in the form of a service; A process pool is created, in which one process in the process pool corresponds to loading an AI algorithm model; an idle process is selected from the process pool through a service request to call the AI ​​algorithm model, and the calling result of the AI ​​algorithm model is fed back through a service response.

2. A data link network security demand processing method as claimed in claim 1, characterized in that: The knowledge extraction model is used to receive a target document, extract demand elements from the target document, and form a target knowledge graph; Among them, the construction of the knowledge extraction model includes: using the existing extraction model to extract entities from the target text, obtain triple data, and generate a triple knowledge base; using a graph representation learning method to generate entity vector encodings corresponding to each entity in the triple knowledge base; using the entity vector encoding as a supplementary enhancement means for entity semantic information to improve the existing extraction model and obtain the knowledge extraction model.

3. A data link network security demand processing method as claimed in claim 1 or 2, characterized in that: The method for knowledge extraction comprises: Receiving a target document, reading each text paragraph, generating a text paragraph list, and dividing the text paragraph list into a plurality of equal parts to obtain a plurality of text paragraph sublists; Traversing multiple processes in the process pool in sequence, passing a text paragraph sublist to a process, and having the current process call the knowledge extraction model to extract sublist requirement elements from the current text paragraph sublist; After waiting for all processes in the process pool to complete the extraction, the triple data lists of all sub-list requirement elements are concatenated to obtain the requirement elements of the target document.

4. A data link network security demand processing method as claimed in claim 2, characterized in that: The knowledge recommendation model is used to calculate the similarity value between the target knowledge graph and each of the sample knowledge graphs in the database, and output the guarantee plan corresponding to the sample knowledge graph with the highest similarity value.

5. A data link network security demand processing method as claimed in claim 4, characterized in that: Calculating the similarity value between the target knowledge graph and each of the sample knowledge graphs in the database includes: Based on the basic framework of the SimGNN model, the edges in the knowledge graph are set as special nodes in the knowledge graph; and on the basis of the graph topology structure difference characteristics of the SimGNN model, the combat area difference characteristics and the combat force difference characteristics are added.

6. A method for processing data link network security requirements as claimed in claim 5, characterized in that: The addition of the operational area difference characteristics and the combat force difference characteristics include: Input the target knowledge graph and any sample knowledge graph, and perform first-path processing, second-path processing, and third-path processing; The first processing path includes: after edge-to-node processing in the knowledge graph, edge coding and adding relationship information processing on the edge, inputting into the SimGNN model, and obtaining the topological structure similarity between the two knowledge graphs based on the SimGNN model; The second processing path includes: extracting geographic information from the knowledge graph, setting any geographic location as a reference point, respectively calculating the distance between each geographic location in the knowledge graph and the reference point, and obtaining a distance vector of the reference point; and inputting the distance vector into the Transformer network, extracting the geographic spatial information of the two knowledge graphs, and obtaining the geographic spatial similarity between the two knowledge graphs; The third processing path includes: extracting the force information in the knowledge graph, classifying and counting the number of participating troops of each type of arms, and obtaining a force quantity vector; and inputting the force quantity vector into the Transformer network to obtain the force similarity between the two knowledge graphs; The topological structure similarity, the geographic space similarity and the military force similarity are spliced ​​and input into the Dense network to obtain the similarity value between the target knowledge graph and the sample knowledge graph.

7. A method for processing data link network security requirements as claimed in claim 1, characterized in that: The knowledge recommendation model is used to receive a natural language question from a user, convert the natural language question into a query language of the database, execute the query language on the database, generate an answer to the natural language question, and complete a query of the guarantee plan and the corresponding sample knowledge graph.

8. A method for processing data link network security requirements as claimed in claim 7, characterized in that: Converting the natural language question into the query language of the database includes: Establishing a template library, wherein the template library includes standard question templates of multiple natural language questions; Processing the semantic information of the natural language question sentence to obtain question sentence semantic coding information, and determining which standard question sentence template the natural language question sentence belongs to based on the semantic coding information; Extracting multiple key entities from the natural language question; The key entity is filled into the standard question template to generate an instantiated query language of the database.

9. A method for processing data link network security requirements as claimed in claim 8, characterized in that: Filling the key entity into the standard question template includes: Determine whether the key entity exists in the sample knowledge graph; If it exists, directly fill the key entity into the standard question template; If it does not exist, the edit distance between the key entity and each sample entity in the sample knowledge graph is calculated, and the key entity is replaced by the sample entity with the smallest edit distance in the sample knowledge graph, and then the sample entity is filled into the standard question template.

10. A data link network security demand intelligent processing system, characterized in that: The system at least comprises a database, a knowledge extraction module, a knowledge query module, and a knowledge recommendation module, and is used to execute the steps of the method described in any one of claims 1 to 9.

Images