A multi-layer risk monitoring method for computer systems based on digital twin technology

Through digital twin technology, the static and dynamic state maps and virtualized environment models of computer systems are constructed, and the risk analysis is used to use multi-layer attention mechanism models to solve the problems of high false alarm rates and bypass monitoring in the existing technology, and the further development of the new technology is achieved. Fine risk assessment and response enhances the security and business continuity of computer systems.

CN118296604BActive Publication Date: 2024-11-12ZHAOBANG INFORMATION (NANJING) CO LTD +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410727402.5
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-06-06
Publication Date
2024-11-12
Estimated Expiration
2044-06-06

AI Technical Summary

Technical Problem

The existing computer risk monitoring technology has problems such as high false alarm rates and direct attack on the system bypassing monitoring. It is unable to effectively monitor and respond to potential threats in real time, resulting in the risk of data leakage and system paralysis being difficult to prevent.

Method used

Using a multi-layer risk monitoring method based on digital twin technology, by constructing a static and dynamic state knowledge graph and virtualized environment model of computer systems, using a multi-layer attention mechanism model to perform risk analysis on access requests, assess the risk level of data and select Response method, establish a second layer of defense to enhance security.

Benefits of technology

Achieve more sophisticated risk assessment and response, reduce system maintenance time and cost, improve model flexibility and accuracy, enable potential failure points to be detected in advance and quickly diagnose problems, and enhance computer system security and business continuity .

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118296604B_ABST
    Figure CN118296604B_ABST
Patent Text Reader

Abstract

The present invention belongs to the field of computer security technology and relates to a multi-layer risk monitoring method for a computer system based on digital twin technology. The method collects internal data of a computer system, annotates the collected data with entities and relationships using a co-occurrence model method, and constructs a knowledge graph of the static state of the computer system. The state and response of the computer system to external access are established through a finite state machine to construct the dynamic state of the computer system. A twin system based on digital twin technology is constructed through the static and dynamic states of the computer system. An external access response is given by a judgment rule through a risk assessment of external access to the twin system. The method sets two lines of defense to ensure system security. Potential failure points are discovered in advance to reduce system maintenance time and cost. At the same time, risk factors in the data are considered more comprehensively to improve the flexibility and accuracy of the model.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of computer security technology, and in particular to a multi-layer risk monitoring method for a computer system based on digital twin technology. Background Art

[0002] Computer risk monitoring is essential to ensure the security of information systems and data. By monitoring potential security threats and abnormal behaviors, various attacks can be discovered and responded to in advance, thereby preventing risk events such as data leakage, system paralysis or information loss. Timely risk monitoring can help organizations respond to security threats quickly, reduce potential losses and the scope of impact. In addition, continuous risk monitoring helps improve system security, identify and remediate security vulnerabilities, and improve overall risk management efficiency. Through effective computer risk monitoring, organizations can establish a more robust and reliable information security protection system, maintain business continuity and reputation, and ensure the confidentiality of sensitive data, thereby achieving business success and long-term development in today's digital age.

[0003] Existing computer risk monitoring technologies include log analysis, anomaly detection, network traffic analysis, and behavior analysis. Although they can provide real-time monitoring and alarm functions to help quickly respond to potential risk events, some monitoring technologies are post-analysis technologies, and the losses have already occurred. In addition, real-time monitoring has problems such as high false alarm rate and bypassing monitoring to directly attack the system. Summary of the invention

[0004] The purpose of the present invention is to overcome the shortcomings of the prior art and provide a multi-layer risk monitoring method for computer systems based on digital twin technology.

[0005] In order to achieve the purpose of the present invention, the present invention will be implemented by adopting the technical solution described below.

[0006] A multi-layer risk monitoring method for a computer system based on digital twin technology, characterized in that the method comprises the following steps:

[0007] S11. Using the co-occurrence model to annotate entities and relationships on the collected internal data of the real system, construct a knowledge graph of the internal data of the real system, and establish a static state of the real system through the knowledge graph; the method for constructing the knowledge graph includes the following steps:

[0008] S111. Collect documents describing the characteristics and functions of computer systems, including operating systems, network configurations, and hardware components;

[0009] S112, preprocessing the text data including word segmentation, stop word removal, and stem extraction;

[0010] S113, representing the preprocessed document data set as a document-term matrix, where each row represents a document, each column represents a term, and the matrix elements represent the frequency of occurrence of the term in the document;

[0011] S114, performing singular value decomposition on the document-term matrix to decompose it into three matrices: document-topic, singular value diagonal matrix, V^T term-topic;

[0012] S115, based on the result of singular value decomposition, multiplying the document-topic matrix and the word-topic matrix to obtain new document-topic matrix and word-topic matrix, which contain latent semantic information about the characteristics and functions of the computer system;

[0013] S116, using latent semantic vectors to calculate the similarity between documents, using cosine similarity as the measure;

[0014] S117, using the similarity calculation results to achieve computer internal entity induction and topic classification, and construct a knowledge graph;

[0015] S12. Establishing the state and response of the real system to external access through a finite state machine to construct a dynamic state of the real system; the method for constructing the dynamic state includes the following steps:

[0016] Computer system states and state transitions:

[0017] S121, Init state: At this time, the system is in the initial state and there is no access request;

[0018] S122, Add state: Initiate a data add command to the system. At this time, the system state changes from the Init state to the Add state. The system agrees or rejects the data add command.

[0019] S123, Delete state: send a delete data command to the system, then the system state changes from Init state to Delete state, the system agrees or refuses to delete the data command;

[0020] S124, Edit state: a data modification command is sent to the system. At this time, the system state changes from Init state to Edit state, and the system agrees or rejects the data modification command;

[0021] S125, Inq state: a data query command is sent to the system. At this time, the system state changes from Init state to Inq state. The system agrees or rejects the data query command.

[0022] S126, End state: The system completes the external command and the system state changes to End state;

[0023] Among them, states S41-S45 can be converted to each other;

[0024] S13. According to the static state and dynamic state of the real system, a twin system of the real system is constructed using the digital twin technology, wherein the twin system includes a virtualized environment model created by the static state and an access response model created by the dynamic state, wherein the access response model includes a multi-layer attention mechanism model, the input of the multi-layer attention mechanism model is a system log file, configuration information, and monitoring data, and the output is a risk analysis probability value;

[0025] S14. The twin system uses a multi-layer attention mechanism model to perform risk analysis on the data of access requests, evaluates the risk level of the data and the security requirements of the data based on the risk analysis probability value, and selects the data calling method to respond to the access request.

[0026] As a preferred solution of the present invention, the words in S113 include the names of all software and hardware in the computer system.

[0027] As a preferred embodiment of the present invention, the method for constructing the twin system comprises the following steps:

[0028] S31, creating a virtualized environment model of the computer system through a static state of the computer system, including a host, a virtual machine, storage, and network device components;

[0029] S32. Modeling access responses through the dynamic state of a computer system, including a multi-layer attention mechanism model that uses a multi-layer attention mechanism to perform risk analysis on response data.

[0030] As a preferred embodiment of the present invention, the risk analysis method comprises the following steps:

[0031] S41, collecting system log files, configuration information, and monitoring data;

[0032] S42, inputting the data into a multi-layer attention mechanism model, wherein the model includes an input layer, at least two attention layers and an output layer;

[0033] S43. For each feature, each attention layer calculates its importance weight relative to other features through the soft attention mechanism;

[0034] S44, multiply each input feature by the corresponding attention weight to obtain a weighted feature representation;

[0035] S45, summarizing the weighted feature representations to generate new feature representations;

[0036] S46, through the back propagation algorithm, dynamically adjust the attention weight according to the error between the model output and the actual label;

[0037] S47. Finally, a risk assessment probability value is given.

[0038] As a preferred solution of the present invention, the specific method of the twin system responding to the access request is:

[0039] For data with low risk and low data security requirements, the twin system can directly call the data to respond to access requests;

[0040] For data with low risk and high data security requirements, the twin system initiates a call request to the real system. After secondary risk analysis by the multi-layer attention mechanism model, it determines whether to respond to the access request:

[0041] If not, the twin system directly rejects the access request and reports the risk to the computer system, requiring risk processing;

[0042] If yes, the twin system directly calls the database to respond to external access requests.

[0043] The present invention has the following advantages: through detailed analysis of the static and dynamic states of the computer system, a more sophisticated twin system is constructed, which is more in line with the real system; by constructing a twin system, external access does not directly contact the computer system, and the behavior of the system under specific fault conditions can be simulated, helping engineers to diagnose and solve problems more quickly. Through digital twin technology, potential fault points can be discovered in advance, reducing system maintenance time and cost; the risk assessment method based on the multi-layer attention mechanism is used as the second line of defense for the computer system. This method will weight various risk factors and output the corresponding risk assessment results, which can more comprehensively consider the risk factors in the data and improve the flexibility and accuracy of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 It is the overall framework diagram of the present invention;

[0045] Figure 2 A schematic diagram of the process of constructing a knowledge graph of a static state of a computer system of the present invention;

[0046] Figure 3 A schematic diagram of the computer system state and state transition process of the present invention;

[0047] Figure 4 Schematic diagram of the risk assessment process based on the multi-layer attention mechanism of the present invention;

[0048] Figure 5 Schematic diagram of the twin system and response scheme based on digital twin technology of the present invention. DETAILED DESCRIPTION

[0049] The following is a description of the implementation of the present invention by specific embodiments. People familiar with the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. 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 are within the scope of protection of the present invention.

[0050] As an embodiment of the present invention, Figure 1 , Figure 2 and Figure 3 As shown, a multi-layer risk monitoring method for a computer system based on digital twin technology comprises the following steps:

[0051] S11. Using the co-occurrence model to annotate entities and relationships on the collected internal data of the real system, construct a knowledge graph of the internal data of the real system, and establish a static state of the real system through the knowledge graph; the method for constructing the knowledge graph includes the following steps:

[0052] S111. Collect documents describing the characteristics and functions of computer systems, including operating systems, network configurations, and hardware components;

[0053] S112, preprocessing the text data including word segmentation, stop word removal, and stem extraction;

[0054] S113, representing the preprocessed document data set as a document-term matrix, where each row represents a document, each column represents a term, and the matrix elements represent the frequency of occurrence of the term in the document;

[0055] S114, performing singular value decomposition on the document-term matrix to decompose it into three matrices: document-topic, singular value diagonal matrix, V^T term-topic;

[0056] S115, based on the result of singular value decomposition, multiplying the document-topic matrix and the word-topic matrix to obtain new document-topic matrix and word-topic matrix, which contain latent semantic information about the characteristics and functions of the computer system;

[0057] S116, using latent semantic vectors to calculate the similarity between documents, using cosine similarity as the measure;

[0058] S117, using the similarity calculation results to achieve computer internal entity induction and topic classification, and construct a knowledge graph;

[0059] S12. Establishing the state and response of the real system to external access through a finite state machine to construct a dynamic state of the real system; the method for constructing the dynamic state includes the following steps:

[0060] Computer system states and state transitions:

[0061] S121, Init state: At this time, the system is in the initial state, without external access;

[0062] S122, Add state: Initiate a data add command to the system. At this time, the system state changes from the Init state to the Add state. The system agrees or rejects the data add command.

[0063] S123, Delete state: send a delete data command to the system, then the system state changes from Init state to Delete state, the system agrees or refuses to delete the data command;

[0064] S124, Edit state: a data modification command is sent to the system. At this time, the system state changes from Init state to Edit state, and the system agrees or rejects the data modification command;

[0065] S125, Inq state: a data query command is sent to the system. At this time, the system state changes from Init state to Inq state. The system agrees or rejects the data query command.

[0066] S126, End state: The system completes the external command and the system state changes to End state;

[0067] Among them, states S41-S45 can be converted to each other;

[0068] S13. According to the static state and dynamic state of the real system, a twin system of the real system is constructed using the digital twin technology, wherein the twin system includes a virtualized environment model created by the static state and an access response model created by the dynamic state, wherein the access response model includes a multi-layer attention mechanism model for performing risk analysis on access data;

[0069] S14. The twin system uses a multi-layer attention mechanism model to perform risk analysis on the data of the access request, evaluates the risk level of the data and the security requirements of the data based on the risk analysis probability value, and selects a data call method to respond to the access request.

[0070] As an embodiment of the present invention, Figure 4 As shown, the risk analysis method comprises the following steps:

[0071] S41, collecting system log files, configuration information, and monitoring data;

[0072] S42, inputting the data into a multi-layer attention mechanism model, wherein the model includes an input layer, at least two attention layers and an output layer;

[0073] S43. For each feature, each attention layer calculates its importance weight relative to other features through the soft attention mechanism;

[0074] S44, multiply each input feature by the corresponding attention weight to obtain a weighted feature representation;

[0075] S45, summarizing the weighted feature representations to generate new feature representations;

[0076] S46, through the back propagation algorithm, dynamically adjust the attention weight according to the error between the model output and the actual label;

[0077] S47. Finally, a risk assessment probability value is given.

[0078] As an embodiment of the present invention, Figure 5 As shown, the twin system uses a multi-layer attention mechanism model to perform risk analysis on the data of the access request, evaluates the risk level of the data and the security requirements of the data according to the risk analysis probability value output by the multi-layer attention mechanism model, and selects a data call method to respond to the access request; the data includes low-risk data with low data security requirements and low-risk data with high data security requirements, wherein:

[0079] For data with low risk and low data security requirements, the twin system can directly call the data to respond to the access;

[0080] For data with low risk and high data security requirements, the twin system initiates a call request to the real system. After secondary risk analysis by the multi-layer attention mechanism model, it determines whether to respond to the access request:

[0081] If not, the twin system directly rejects the access request and reports the risk to the computer system, requiring risk processing;

[0082] If yes, the twin system directly calls the database to respond to external access requests.

[0083] As an embodiment of the present invention, the words in S113 include the names of all software and hardware in the computer system.

[0084] As an embodiment of the present invention, an attack is given to a certain hardware of the system to see how the twin system reacts to the attack. If there is a reaction, it means that there is no vulnerability. Otherwise, the vulnerability can be improved on the real system.

[0085] As an embodiment of the present invention, the system operation is simulated to check whether the system is operating normally. For example, if data is stored on the twin system, failure indicates that there is a fault in the storage device, and the storage device with the problem is called a fault point.

Claims

1. A multi-layer risk monitoring method for computer systems based on digital twin technology. It is characterized in that The method comprises the following steps: S11. Using the co-occurrence model to annotate entities and relationships on the collected internal data of the real system, construct a knowledge graph of the internal data of the real system, and establish a static state of the real system through the knowledge graph; the method for constructing the knowledge graph includes the following steps: S111. Collect documents describing the characteristics and functions of computer systems, including operating systems, network configurations, and hardware components; S112, preprocessing the text data including word segmentation, stop word removal, and stem extraction; S113, representing the preprocessed document data set as a document-term matrix, where each row represents a document, each column represents a term, and the matrix elements represent the frequency of occurrence of the term in the document; S114, performing singular value decomposition on the document-term matrix to decompose it into three matrices: document-topic, singular value diagonal matrix, V^T term-topic; S115, based on the result of singular value decomposition, multiplying the document-topic matrix and the word-topic matrix to obtain new document-topic matrix and word-topic matrix, which contain latent semantic information about the characteristics and functions of the computer system; S116, using latent semantic vectors to calculate the similarity between documents, using cosine similarity as the measure; S117, using the similarity calculation results to achieve computer internal entity induction and topic classification, and construct a knowledge graph; S12. Establishing the state and response of the real system to external access through a finite state machine to construct a dynamic state of the real system; the method for constructing the dynamic state includes the following steps: Computer system states and state transitions: S121, Init state: At this time, the system is in the initial state and there is no access request; S122, Add state: Initiate a data add command to the system. At this time, the system state changes from Init state to Add state. The system agrees or rejects the data add command. S123, Del state: Initiate a data deletion command to the system. At this time, the system state changes from Init state to Del state. The system agrees or refuses to delete the data command. S124, Edit state: Initiate a data modification command to the system. At this time, the system state changes from Init state to Edit state, and the system agrees or rejects the data modification command; S125, Inq state: Initiate a data query command to the system. At this time, the system state changes from Init state to Inq state. The system agrees or rejects the data query command. S126, End state: The system completes the external command and the system state changes to End state; Among them, states S41-S45 can be converted to each other; S13. According to the static state and dynamic state of the real system, a twin system of the real system is constructed using the digital twin technology, wherein the twin system includes a virtualized environment model created by the static state and an access response model created by the dynamic state, wherein the access response model includes a multi-layer attention mechanism model, the input of the multi-layer attention mechanism model is a system log file, configuration information, and monitoring data, and the output is a risk analysis probability value; S14. The twin system uses a multi-layer attention mechanism model to perform risk analysis on the data of access requests, evaluates the risk level of the data and the security requirements of the data based on the risk analysis probability value, and selects the data calling method to respond to the access request.

2. According to claim 1, a computer system multi-layer risk monitoring method based on digital twin technology, It is characterized in that The words in S113 include the names of all software and hardware in the computer system.

3. According to claim 1, a computer system multi-layer risk monitoring method based on digital twin technology, It is characterized in that The twin system construction method comprises the following steps: S31, creating a virtualized environment model of the computer system through a static state of the computer system, including a host, a virtual machine, storage, and network device components; S32. Creating an access response model through the dynamic state of a computer system, including a multi-layer attention mechanism model that uses a multi-layer attention mechanism to perform risk analysis on access data.

4. According to the multi-layer risk monitoring method of computer system based on digital twin technology according to claim 1 or 3, It is characterized in that The risk analysis method comprises the following steps: S41, collecting system log files, configuration information, and monitoring data; S42, inputting the data into a multi-layer attention mechanism model, wherein the model includes an input layer, at least two attention layers and an output layer; S43. For each feature, each attention layer calculates its importance weight relative to other features through the soft attention mechanism; S44, multiply each input feature by the corresponding attention weight to obtain a weighted feature representation; S45, summarizing the weighted feature representations to generate new feature representations; S46, through the back propagation algorithm, dynamically adjust the attention weight according to the error between the model output and the actual label; S47. Finally, a risk assessment probability value is given.

5. According to claim 1, a computer system multi-layer risk monitoring method based on digital twin technology, It is characterized in that The specific method of the twin system responding to the access request: For data with low risk and low data security requirements, the twin system can directly call the data to respond to access requests; For data with low risk and high data security requirements, the twin system initiates a call request to the real system. After secondary risk analysis by the multi-layer attention mechanism model, it determines whether to respond to the access request: If not, the twin system directly rejects the access request and reports the risk to the computer system, requiring risk processing; If yes, the twin system directly calls the database to respond to external access requests.