Network security situation awareness and early warning method and system
By arranging switching node probes in the network, identifying abnormal events and training multi-task models, and evaluating business risks, the problem of high resource occupation in the existing technology is solved and the network security of small and micro enterprises is improved.
Patent Information
- Application Number
- CN202411038998.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-31
- Publication Date
- 2024-11-26
AI Technical Summary
The existing network security situation awareness relies on a learning model of data analysis. When facing a large number of data attacks, learning and training is required for different anomalies, resulting in a large demand for device performance resources and a high deployment threshold, which is not suitable for micro-micro enterprise.
Collect multi-form operation data by arranging network probes for exchange nodes, identify abnormal events, train the Multi-task Learning multi-task model, evaluate the abnormal status of the equipment and business risks, determine whether the security threshold is exceeded, and implement repair or business transfer strategies.
Reduces the resource usage of situational awareness, improves business security, and lowers the deployment threshold, making it suitable for small and micro enterprises.
Smart Images

Figure CN119030747A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of network attack and defense technology, and in particular to a network security situation perception and early warning method and system. Background Art
[0002] Network security situation awareness is a monitoring and analysis technology for the field of network security. It collects, analyzes and processes various information in the network environment to understand the network security status in real time, and promptly detect and respond to potential threats and attacks.
[0003] Since the existing network security situation awareness relies on learning models for data analysis, when facing large amounts of data attacks, different application models need to be learned and trained for different abnormal events. When the final situation awareness platform is deployed, the performance resource requirements of the equipment are large, the deployment threshold is high, and it is not suitable for small and micro enterprises. Summary of the invention
[0004] In order to solve the above technical problems, a network security situation awareness and early warning method and system are provided. This technical solution solves the above-mentioned problem that since the existing network security situation awareness relies on a learning model of data analysis, when facing a large amount of data attacks, it is necessary to learn and train different application models for different abnormal events. When the final situation awareness platform is deployed, the performance resource requirements of the equipment are large, the deployment threshold is high, and it is not suitable for small and micro enterprises.
[0005] In order to achieve the above purpose, the technical solution adopted by the present invention is:
[0006] A network security situation awareness and early warning method, comprising:
[0007] Deploy network probes at several switching nodes to collect multi-modal operation data of several switching nodes;
[0008] Based on the polymorphic operation data of several switching nodes, identify abnormal events in the polymorphic operation data to evaluate the abnormal status of each device;
[0009] Analyze the risk of business data loss under abnormal conditions of each device and evaluate the risk indicators of each business;
[0010] Determine whether the risk indicators of each business exceed the safety threshold. If not, implement a repair strategy; if so, implement a business transfer strategy.
[0011] Preferably, based on the polymorphic operation data of a plurality of switching nodes, identifying abnormal events in the polymorphic operation data to evaluate the abnormal state of each device specifically includes:
[0012] Based on a third-party black and white database of cyber attacks, train a Multi-task Learning model;
[0013] Substitute the multi-modal operation data of several switching nodes into the Multi-task Learning model to obtain the probability of occurrence of different abnormal events on each switching node;
[0014] Assign impact scores to different abnormal events according to the probability of occurrence of different abnormal events on each switching node and the degree of impact on the connected devices;
[0015] The abnormality coefficient of each device on the switching node is calculated based on the occurrence probability of different abnormal events on each switching node and the impact score of different abnormal events;
[0016] The calculation of the abnormal coefficient of each device on the switching node is as follows:
[0017]
[0018] Where P k is the abnormal coefficient of the kth device on the switching node, V k is the kth device on the switching node, G i is the probability of occurrence of the ith abnormal event, U i is the impact score of the ith abnormal event, L is the total number of devices, and n is the total number of abnormal events.
[0019] Preferably, based on a third-party black-and-white database of cyber attacks, training a Multi-task Learning model specifically includes:
[0020] Based on a third-party black-and-white database of cyber attacks, we determine the feature data and label data of all abnormal events and package them into a black-and-white data training set.
[0021] Build a multi-task learning model with multiple architectures;
[0022] Based on the Multi-task Learning model, the black and white data training set is used as the input layer;
[0023] Based on the black and white data training set inputted by the input layer, CNN convolution, pooling component, RNN loop component and Word2Vec component are encapsulated in the embedding layer to identify and extract the characteristic parameters of different abnormal events;
[0024] Based on the feature parameters of different abnormal events extracted, they are input into the fully connected shared layer. The fully connected shared layer extracts the feature parameters of the specified abnormal event by capturing common features across tasks and performing feature fusion to obtain fused feature parameters.
[0025] Based on the fusion feature parameters input to a specific layer, classification detection is performed through the softmax component inside the specific layer to obtain the probability value of the abnormal event;
[0026] Based on the probability value of the abnormal event, the final abnormal data label data is determined as the output result of the output layer.
[0027] Preferably, the risk of business data loss under abnormal conditions of each device is analyzed, and the risk indicators of each business are evaluated, including:
[0028] Determine the attack targets of different abnormal events;
[0029] Conduct correlation analysis based on the business communication records of the device and the attack targets of different abnormal events to obtain the correlation index between the abnormal event and the business;
[0030] Evaluate the business data loss risk index based on the abnormal coefficient of each device on the switching node and the correlation index between the abnormal event and the business;
[0031] The business data loss risk index calculation expression is:
[0032]
[0033] Where D is the business data loss risk index, P k is the abnormal coefficient of the kth device on the switching node, R ij is the correlation index between the ith abnormal event and the jth business, W i,j,k is the weighting factor between the ith abnormal event and the jth service on the kth device.
[0034] Preferably, the correlation analysis is performed based on the business communication records of the device and the attack targets of different abnormal events to obtain the correlation index between the abnormal event and the business, which specifically includes:
[0035] Get the communication frequency between each abnormal event and business on the device;
[0036] Determine the relevance of the attack target of each abnormal event on the device and the business communication transmission protocol, and determine the relevance of the abnormal event and the business communication transmission protocol;
[0037] Mark whether the initiating interface of each abnormal event meets the permission to access business communication resources, and determine the matching degree of access rights between the abnormal event and the business;
[0038] Based on the communication frequency between each abnormal event on the device and the business, the correlation between the communication transmission protocol of the abnormal event and the business, and the matching degree of the access rights of the abnormal event and the business, the correlation index between the abnormal event and the business is calculated through the normalized scoring function;
[0039] The calculation expression of the correlation index between the abnormal event and the business is:
[0040]
[0041] In the formula, R ij is the correlation index between the ith abnormal event and the jth business, F ij is the communication frequency between the i-th abnormal event and the j-th business, C ij is the communication transmission protocol correlation between the i-th abnormal event and the j-th business, A ij is the access permission matching degree between the i-th abnormal event and the j-th business, F, C, and A represent the communication frequency, communication transmission protocol relevance, and access permission matching degree respectively, and α, β, and γ represent the correlation coefficient of communication frequency, the correlation coefficient of communication transmission protocol relevance, and the access permission matching degree correlation coefficient respectively.
[0042] Furthermore, a network security situation awareness and early warning system is proposed, which is used to implement the network security situation awareness and early warning method as described above, including:
[0043] A data collection module is used to arrange network probes of several switching nodes and collect multi-form operation data of several switching nodes;
[0044] The abnormality assessment module is wirelessly connected to the data acquisition module. The abnormality assessment module is used to identify abnormal events in the multi-form operation data based on the multi-form operation data of several switching nodes to evaluate the abnormal state of each device;
[0045] A risk assessment module, which is electrically connected to the abnormality assessment module. The risk assessment module is used to analyze the risk of business data loss under the abnormal state of each device and assess the risk index of each business;
[0046] The decision-making module is electrically connected to the risk assessment module. The decision-making module is used to determine whether the risk index of each business exceeds the safety threshold. If not, a repair strategy is implemented. If so, a business transfer strategy is implemented.
[0047] Optionally, the anomaly assessment module includes:
[0048] The multi-task model building unit trains the Multi-taskLearning multi-task model based on a third-party black and white database of cyber attacks;
[0049] The abnormal probability analysis unit substitutes the multi-modal operation data of several switching nodes into the Multi-task Learning multi-task model to obtain the probability of occurrence of different abnormal events on each switching node;
[0050] A scoring unit assigns impact scores to different abnormal events according to the impact of the occurrence probability of different abnormal events on each switching node on the connected devices;
[0051] The first calculation unit calculates an abnormality coefficient of each device on the switching node according to the occurrence probability of different abnormal events on each switching node and the impact scores of different abnormal events.
[0052] Optionally, the multi-task model building unit includes:
[0053] Architecture model unit, building multi-architecture Multi-task Learning multi-task models;
[0054] The input layer uses the black and white data training set as the input layer based on the Multi-task Learning model;
[0055] The embedding layer encapsulates CNN convolution, pooling components, RNN loop components, and Word2Vec components based on the black-and-white data training set input by the input layer to identify and extract the characteristic parameters of different abnormal events;
[0056] The shared layer extracts feature parameters of different abnormal events and inputs them into the fully connected shared layer. The fully connected shared layer extracts feature parameters of specified abnormal events by capturing common features across tasks and performing feature fusion to obtain fused feature parameters.
[0057] Specific layer: Based on the fusion feature parameters input into the specific layer, the softmax component inside the specific layer is used for classification detection to obtain the probability value of abnormal events;
[0058] The output layer determines the final abnormal data label data based on the probability value of the abnormal event as the output result of the output layer.
[0059] Optionally, the risk assessment module includes:
[0060] Target confirmation unit, which determines the attack targets of different abnormal events;
[0061] The correlation analysis unit performs correlation analysis based on the business communication records of the device and the attack targets of different abnormal events to obtain the correlation index between the abnormal event and the business;
[0062] The second calculation unit evaluates the business data loss risk index based on the abnormal coefficient of each device on the switching node and the correlation index between the abnormal event and the business;
[0063] The correlation analysis unit includes:
[0064] The communication frequency subunit obtains the communication frequency of each abnormal event and service on the device;
[0065] The protocol correlation unit determines the correlation between the attack target of each abnormal event on the device and the business communication transmission protocol, and determines the correlation between the abnormal event and the business communication transmission protocol;
[0066] The permission matching sub-unit marks whether the initiating interface of each abnormal event meets the permission to access business communication resources, and determines the matching degree of the access permission between the abnormal event and the business;
[0067] The third calculation unit calculates the correlation index between the abnormal event and the business through a normalized scoring function based on the communication frequency between each abnormal event on the device and the business, the correlation between the communication transmission protocol between the abnormal event and the business, and the matching degree of access rights between the abnormal event and the business.
[0068] Compared with the prior art, the present invention has the following beneficial effects:
[0069] The present invention proposes a network security situation awareness and early warning solution, which collects operation data by arranging switching node probes in the network, intelligently identifies abnormal events to evaluate equipment status, and then analyzes business risks. Repairs are implemented for risks that do not exceed the threshold, and business transfer is prioritized for risks that exceed the threshold, thereby reducing resource usage of situation awareness and improving business security. BRIEF DESCRIPTION OF THE DRAWINGS
[0070] Figure 1 It is a flow chart of a network security situation awareness and early warning method;
[0071] Figure 2 A flow chart of the method for evaluating the abnormal status of each device;
[0072] Figure 3 Flowchart of the method for training Multi-task Learning multi-task model;
[0073] Figure 4 A flow chart of the risk indicator methodology for assessing each business;
[0074] Figure 5The flowchart of the correlation index analysis method between abnormal events and business;
[0075] Figure 6 This is a framework diagram of a network security situation awareness and early warning system. DETAILED DESCRIPTION
[0076] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are only examples, and those skilled in the art may think of other obvious variations.
[0077] Reference Figure 1 As shown, a network security situation awareness and early warning method includes:
[0078] Deploy network probes at several switching nodes to collect multi-modal operation data of several switching nodes;
[0079] Based on the polymorphic operation data of several switching nodes, identify abnormal events in the polymorphic operation data to evaluate the abnormal status of each device;
[0080] Analyze the risk of business data loss under abnormal conditions of each device and evaluate the risk indicators of each business;
[0081] Determine whether the risk indicators of each business exceed the safety threshold. If not, implement a repair strategy; if so, implement a business transfer strategy.
[0082] This solution collects operating data by deploying switching node probes in the network, intelligently identifies abnormal events to evaluate device status, and then analyzes business risks. It repairs risks that do not exceed the threshold, and prioritizes business transfer for risks that exceed the threshold to ensure business security.
[0083] Reference Figure 2 As shown, based on the multi-modal operation data of several switching nodes, identifying abnormal events in the multi-modal operation data to evaluate the abnormal state of each device specifically includes:
[0084] Based on a third-party black and white database of cyber attacks, train a Multi-task Learning model;
[0085] Substitute the multi-modal operation data of several switching nodes into the Multi-task Learning model to obtain the probability of occurrence of different abnormal events on each switching node;
[0086] Assign impact scores to different abnormal events according to the probability of occurrence of different abnormal events on each switching node and the degree of impact on the connected devices;
[0087] The abnormality coefficient of each device on the switching node is calculated based on the occurrence probability of different abnormal events on each switching node and the impact score of different abnormal events;
[0088] The calculation of the abnormal coefficient of each device on the switching node is as follows:
[0089]
[0090] Where P k is the abnormal coefficient of the kth device on the switching node, V k is the kth device on the switching node, G i is the probability of occurrence of the ith abnormal event, U i is the impact score of the ith abnormal event, L is the total number of devices, and n is the total number of abnormal events.
[0091] It is understandable that the existing situational awareness usually builds different learning models for different abnormal events and deploys them to the monitoring platform. However, doing so will make the monitoring platform have higher overall performance requirements. Therefore, by building a multi-architecture multi-task analysis model, the performance requirements can be reduced while ensuring the recognition capability.
[0092] Reference Figure 3 As shown in the figure, based on the third-party black and white database of network attacks, training the Multi-task Learning multi-task model specifically includes:
[0093] Based on a third-party black-and-white database of cyber attacks, we determine the feature data and label data of all abnormal events and package them into a black-and-white data training set.
[0094] Build a multi-task learning model with multiple architectures;
[0095] Based on the Multi-task Learning model, the black and white data training set is used as the input layer;
[0096] Based on the black and white data training set inputted by the input layer, CNN convolution, pooling component, RNN loop component and Word2Vec component are encapsulated in the embedding layer to identify and extract the characteristic parameters of different abnormal events;
[0097] Based on the feature parameters of different abnormal events extracted, they are input into the fully connected shared layer. The fully connected shared layer extracts the feature parameters of the specified abnormal event by capturing common features across tasks and performing feature fusion to obtain fused feature parameters.
[0098] Based on the fusion feature parameters input to a specific layer, classification detection is performed through the softmax component inside the specific layer to obtain the probability value of the abnormal event;
[0099] Based on the probability value of the abnormal event, the final abnormal data label data is determined as the output result of the output layer.
[0100] It is understandable that based on the third-party database, most of the features of known abnormal events can be obtained, making the model training more perfect and avoiding the update and iteration of the worm virus that leads to the decline of the model's detection performance.
[0101] Reference Figure 4 As shown in the figure, the business data loss risk of each device under abnormal conditions is analyzed, and the risk indicators of each business are evaluated, including:
[0102] Determine the attack targets of different abnormal events;
[0103] Conduct correlation analysis based on the business communication records of the device and the attack targets of different abnormal events to obtain the correlation index between the abnormal event and the business;
[0104] Evaluate the business data loss risk index based on the abnormal coefficient of each device on the switching node and the correlation index between the abnormal event and the business;
[0105] The business data loss risk index calculation expression is:
[0106]
[0107] Where D is the business data loss risk index, P k is the abnormal coefficient of the kth device on the switching node, R ij is the correlation index between the ith abnormal event and the jth business, W i,j,k is the weighting factor between the ith abnormal event and the jth service on the kth device.
[0108] It should be noted that the weight value of the weighting factor needs to be determined through an experience pool based on the importance of the business and the criticality of the equipment. The specific selection of the value based on the type of business will not be described in detail here.
[0109] Referring to Figure 5, a correlation analysis is performed based on the business communication records of the device and the attack targets of different abnormal events, and the correlation index between the abnormal event and the business is obtained, which specifically includes:
[0110] Get the communication frequency between each abnormal event and business on the device;
[0111] Determine the relevance of the attack target of each abnormal event on the device and the business communication transmission protocol, and determine the relevance of the abnormal event and the business communication transmission protocol;
[0112] Mark whether the initiating interface of each abnormal event meets the permission to access business communication resources, and determine the matching degree of access rights between the abnormal event and the business;
[0113] Based on the communication frequency between each abnormal event on the device and the business, the correlation between the communication transmission protocol of the abnormal event and the business, and the matching degree of the access rights of the abnormal event and the business, the correlation index between the abnormal event and the business is calculated through the normalized scoring function;
[0114] The calculation expression of the correlation index between the abnormal event and the business is:
[0115]
[0116] In the formula, R ij is the correlation index between the ith abnormal event and the jth business, F ij is the communication frequency between the i-th abnormal event and the j-th business, C ij is the communication transmission protocol correlation between the i-th abnormal event and the j-th business, A ij is the access permission matching degree between the i-th abnormal event and the j-th business, F, C, and A represent the communication frequency, communication transmission protocol relevance, and access permission matching degree respectively, and α, β, and γ represent the correlation coefficient of communication frequency, the correlation coefficient of communication transmission protocol relevance, and the access permission matching degree correlation coefficient respectively.
[0117] It can be understood that by considering the communication frequency between abnormal events and business, the relevance of the communication transmission protocol used, and the matching degree of access rights of the abnormal event initiation interface to the business communication resources, the potential impact or correlation of these abnormal events on the business can be quantitatively evaluated. The level of correlation index directly reflects the closeness between abnormal events and business, providing core parameters for business security analysis.
[0118] Reference Figure 6 As shown, based on the same inventive concept of a network security situation awareness and early warning method, a network security situation awareness and early warning system is proposed, comprising:
[0119] A data collection module is used to arrange network probes of several switching nodes and collect multi-form operation data of several switching nodes;
[0120] The abnormality assessment module is wirelessly connected to the data acquisition module. The abnormality assessment module is used to identify abnormal events in the multi-form operation data based on the multi-form operation data of several switching nodes to evaluate the abnormal state of each device;
[0121] A risk assessment module, which is electrically connected to the abnormality assessment module. The risk assessment module is used to analyze the risk of business data loss under the abnormal state of each device and assess the risk index of each business;
[0122] The decision-making module is electrically connected to the risk assessment module. The decision-making module is used to determine whether the risk index of each business exceeds the safety threshold. If not, a repair strategy is implemented. If so, a business transfer strategy is implemented.
[0123] The abnormal assessment module includes:
[0124] The multi-task model building unit trains the Multi-taskLearning multi-task model based on a third-party black and white database of cyber attacks;
[0125] The abnormal probability analysis unit substitutes the multi-modal operation data of several switching nodes into the Multi-task Learning multi-task model to obtain the probability of occurrence of different abnormal events on each switching node;
[0126] A scoring unit assigns impact scores to different abnormal events according to the impact of the occurrence probability of different abnormal events on each switching node on the connected devices;
[0127] The first calculation unit calculates an abnormality coefficient of each device on the switching node according to the occurrence probability of different abnormal events on each switching node and the impact scores of different abnormal events.
[0128] The multi-task model building unit includes:
[0129] Architecture model unit, building multi-architecture Multi-task Learning multi-task models;
[0130] The input layer uses the black and white data training set as the input layer based on the Multi-task Learning model;
[0131] The embedding layer encapsulates CNN convolution, pooling components, RNN loop components, and Word2Vec components based on the black-and-white data training set input by the input layer to identify and extract the characteristic parameters of different abnormal events;
[0132] The shared layer extracts feature parameters of different abnormal events and inputs them into the fully connected shared layer. The fully connected shared layer extracts feature parameters of specified abnormal events by capturing common features across tasks and performing feature fusion to obtain fused feature parameters.
[0133] Specific layer: Based on the fusion feature parameters input into the specific layer, the softmax component inside the specific layer is used for classification detection to obtain the probability value of abnormal events;
[0134] The output layer determines the final abnormal data label data based on the probability value of the abnormal event as the output result of the output layer.
[0135] The risk assessment module includes:
[0136] Target confirmation unit, which determines the attack targets of different abnormal events;
[0137] The correlation analysis unit performs correlation analysis based on the business communication records of the device and the attack targets of different abnormal events to obtain the correlation index between the abnormal event and the business;
[0138] The second calculation unit evaluates the business data loss risk index based on the abnormal coefficient of each device on the switching node and the correlation index between the abnormal event and the business;
[0139] The correlation analysis unit includes:
[0140] The communication frequency subunit obtains the communication frequency of each abnormal event and service on the device;
[0141] The protocol correlation unit determines the correlation between the attack target of each abnormal event on the device and the business communication transmission protocol, and determines the correlation between the abnormal event and the business communication transmission protocol;
[0142] The permission matching sub-unit marks whether the initiating interface of each abnormal event meets the permission to access business communication resources, and determines the matching degree of the access permission between the abnormal event and the business;
[0143] The third calculation unit calculates the correlation index between the abnormal event and the business through a normalized scoring function based on the communication frequency between each abnormal event on the device and the business, the correlation between the communication transmission protocol between the abnormal event and the business, and the matching degree of access rights between the abnormal event and the business.
[0144] The use process of a network security situation awareness and early warning system is as follows:
[0145] Step 1: deploy network probes of several switching nodes to collect multi-modal operation data of several switching nodes;
[0146] Step 2: Train the Multi-task Learning model based on a third-party black and white database of cyber attacks;
[0147] Step 3: Substitute the multi-modal operation data of several switching nodes into the Multi-task Learning model to obtain the probability of occurrence of different abnormal events on each switching node;
[0148] Step 4: Assign impact scores to different abnormal events according to the probability of occurrence of different abnormal events on each switching node and the degree of impact on the connected devices;
[0149] Step 5: Calculate the abnormal coefficient of each device on the switching node according to the occurrence probability of different abnormal events on each switching node and the impact score of different abnormal events;
[0150] Step 6: Determine the attack targets of different abnormal events;
[0151] Step 7: Perform correlation analysis based on the business communication records of the device and the attack targets of different abnormal events to obtain the correlation index between the abnormal event and the business;
[0152] Step 8: Based on the abnormal coefficient of each device on the switching node and the correlation index between the abnormal event and the business, evaluate the business data loss risk index;
[0153] Step 9: Determine whether the risk indicator of each business exceeds the safety threshold. If not, implement a repair strategy. If so, implement a business transfer strategy.
[0154] To sum up, the advantages of the present invention are: by arranging switching node probes in the network to collect operation data, intelligently identifying abnormal events to evaluate equipment status, and then analyzing business risks, repairing risks that do not exceed the threshold, and giving priority to business transfer when the threshold is exceeded, reducing the resource usage of situational awareness and improving business security.
[0155] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions only describe the principles of the present invention. The present invention may be subject to various changes and improvements without departing from the spirit and scope of the present invention. These changes and improvements fall within the scope of the present invention. The scope of protection claimed by the present invention is defined by the attached claims and their equivalents.
Claims
1. A network security situation awareness and early warning method, characterized in that: include: Deploy network probes at several switching nodes to collect multi-modal operation data of several switching nodes; Based on the multi-modal operation data of several switching nodes, identify abnormal events in the multi-modal operation data to evaluate the abnormal status of each device; Analyze the risk of business data loss under abnormal conditions of each device and evaluate the risk indicators of each business; Determine whether the risk indicators of each business exceed the safety threshold. If not, implement a repair strategy; if so, implement a business transfer strategy.
2. A network security situation awareness and early warning method according to claim 1, characterized in that: Based on the multi-modal operation data of several switching nodes, identify abnormal events in the multi-modal operation data for evaluation. The abnormal state of each device specifically includes: Based on a third-party black and white database of cyber attacks, train a Multi-task Learning model; Substitute the multi-modal operation data of several switching nodes into the Multi-task Learning model to obtain the probability of occurrence of different abnormal events on each switching node; Assign impact scores to different abnormal events according to the probability of occurrence of different abnormal events on each switching node and the degree of impact on the connected devices; The abnormality coefficient of each device on the switching node is calculated based on the occurrence probability of different abnormal events on each switching node and the impact score of different abnormal events; The calculation of the abnormal coefficient of each device on the switching node is as follows: Where P k is the abnormal coefficient of the kth device on the switching node, V k is the kth device on the switching node, G i is the probability of occurrence of the ith abnormal event, U i is the impact score of the ith abnormal event, L is the total number of devices, and n is the total number of abnormal events.
3. A network security situation awareness and early warning method according to claim 2, characterized in that: Based on a third-party black and white database of cyber attacks, training the Multi-task Learning model specifically includes: Based on a third-party black-and-white database of cyber attacks, we determine the feature data and label data of all abnormal events and package them into a black-and-white data training set. Build a multi-task learning model with multiple architectures; Based on the Multi-task Learning model, the black and white data training set is used as the input layer; Based on the black and white data training set inputted by the input layer, CNN convolution, pooling component, RNN loop component and Word2Vec component are encapsulated in the embedding layer to identify and extract the characteristic parameters of different abnormal events; Based on the feature parameters of different abnormal events extracted, they are input into the fully connected shared layer. The fully connected shared layer extracts the feature parameters of the specified abnormal event by capturing common features across tasks and performing feature fusion to obtain fused feature parameters. Based on the fusion feature parameters input to a specific layer, classification detection is performed through the softmax component inside the specific layer to obtain the probability value of the abnormal event; Based on the probability value of the abnormal event, the final abnormal data label data is determined as the output result of the output layer.
4. A network security situation awareness and early warning method according to claim 3, characterized in that: Analyze the risk of business data loss under abnormal conditions of each device and evaluate the risk indicators of each business, including: Determine the attack targets of different abnormal events; Conduct correlation analysis based on the business communication records of the device and the attack targets of different abnormal events to obtain the correlation index between the abnormal event and the business; Evaluate the business data loss risk index based on the abnormal coefficient of each device on the switching node and the correlation index between the abnormal event and the business; The business data loss risk index calculation expression is: Where D is the business data loss risk index, P k is the abnormal coefficient of the kth device on the switching node, R ij is the correlation index between the ith abnormal event and the jth business, W i,j,k is the weighting factor between the ith abnormal event and the jth service on the kth device.
5. A network security situation awareness and early warning method according to claim 4, characterized in that: Based on the correlation analysis between the business communication records of the device and the attack targets of different abnormal events, the correlation index between abnormal events and business is obtained, including: Get the communication frequency between each abnormal event and business on the device; Determine the relevance of the attack target of each abnormal event on the device and the business communication transmission protocol, and determine the relevance of the abnormal event and the business communication transmission protocol; Mark whether the initiating interface of each abnormal event meets the permission to access business communication resources, and determine the matching degree of access rights between the abnormal event and the business; Based on the communication frequency between each abnormal event on the device and the business, the correlation between the communication transmission protocol of the abnormal event and the business, and the matching degree of the access rights of the abnormal event and the business, the correlation index between the abnormal event and the business is calculated through the normalized scoring function; The calculation expression of the correlation index between the abnormal event and the business is: In the formula, R ij is the correlation index between the ith abnormal event and the jth business, F ij is the communication frequency between the i-th abnormal event and the j-th business, C ij is the communication transmission protocol correlation between the i-th abnormal event and the j-th business, A ij is the access permission matching degree between the i-th abnormal event and the j-th business, F, C, and A represent the communication frequency, communication transmission protocol relevance, and access permission matching degree respectively, and α, β, and γ represent the correlation coefficient of communication frequency, the correlation coefficient of communication transmission protocol relevance, and the access permission matching degree correlation coefficient respectively.
6. A network security situation awareness and early warning system, characterized in that: Used to implement a network security situation awareness and early warning method as described in any one of claims 1 to 5, comprising: A data collection module is used to arrange network probes of several switching nodes and collect multi-form operation data of several switching nodes; The abnormality assessment module is wirelessly connected to the data acquisition module. The abnormality assessment module is used to identify abnormal events in the multi-form operation data based on the multi-form operation data of several switching nodes to evaluate the abnormal state of each device; A risk assessment module, which is electrically connected to the abnormality assessment module. The risk assessment module is used to analyze the risk of business data loss under the abnormal state of each device and assess the risk index of each business; The decision-making module is electrically connected to the risk assessment module. The decision-making module is used to determine whether the risk index of each business exceeds the safety threshold. If not, a repair strategy is implemented. If so, a business transfer strategy is implemented.
7. A network security situation awareness and early warning system according to claim 6, characterized in that: The anomaly assessment module includes: The multi-task model building unit trains the Multi-task Learning multi-task model based on a third-party black and white database of cyber attacks; The abnormal probability analysis unit substitutes the multi-modal operation data of several switching nodes into the Multi-task Learning multi-task model to obtain the probability of occurrence of different abnormal events on each switching node; A scoring unit assigns impact scores to different abnormal events according to the impact of the occurrence probability of different abnormal events on each switching node on the connected devices; The first calculation unit calculates an abnormality coefficient of each device on the switching node according to the occurrence probability of different abnormal events on each switching node and the impact scores of different abnormal events.
8. A network security situation awareness and early warning system according to claim 6, characterized in that: The multi-task model building unit internally includes: Architecture model unit, building multi-architecture Multi-task Learning multi-task models; The input layer uses the black and white data training set as the input layer based on the Multi-task Learning model; The embedding layer encapsulates CNN convolution, pooling components, RNN loop components, and Word2Vec components based on the black-and-white data training set input by the input layer to identify and extract the characteristic parameters of different abnormal events; The shared layer extracts feature parameters of different abnormal events and inputs them into the fully connected shared layer. The fully connected shared layer extracts feature parameters of specified abnormal events by capturing common features across tasks and performing feature fusion to obtain fused feature parameters. Specific layer: Based on the fusion feature parameters input into the specific layer, the softmax component inside the specific layer is used for classification detection to obtain the probability value of abnormal events; The output layer determines the final abnormal data label data based on the probability value of the abnormal event as the output result of the output layer.
9. A network security situation awareness and early warning system according to claim 6, characterized in that: The risk assessment module includes: Target confirmation unit, which determines the attack targets of different abnormal events; The correlation analysis unit performs correlation analysis based on the business communication records of the device and the attack targets of different abnormal events to obtain the correlation index between the abnormal event and the business; The second calculation unit evaluates the business data loss risk index based on the abnormal coefficient of each device on the switching node and the correlation index between the abnormal event and the business; The correlation analysis unit includes: The communication frequency subunit obtains the communication frequency of each abnormal event and service on the device; The protocol correlation unit determines the correlation between the attack target of each abnormal event on the device and the business communication transmission protocol, and determines the correlation between the abnormal event and the business communication transmission protocol; The permission matching sub-unit marks whether the initiating interface of each abnormal event meets the permission to access business communication resources, and determines the matching degree of the access permission between the abnormal event and the business; The third calculation unit calculates the correlation index between the abnormal event and the business through a normalized scoring function based on the communication frequency between each abnormal event on the device and the business, the correlation between the communication transmission protocol between the abnormal event and the business, and the matching degree of access rights between the abnormal event and the business.
Citation Information
Cited By
Business question-answering system and method based on big data and artificial intelligence
CN119415663A
