A network security assessment method and system based on network risk detection

Through the method based on network risk detection, real-time data is obtained for standardized processing and dimensionality reduction, abnormal data is screened for risk detection, which solves the problem of timely network security assessment in the existing technology, and realizes more efficient security incident analysis and early warning generation , improves network security.

CN118316723BActive Publication Date: 2024-11-19山东慧贝行信息技术有限公司
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410578892.7
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-05-11
Publication Date
2024-11-19
Estimated Expiration
2044-05-11

AI Technical Summary

Technical Problem

The existing network security incident analysis methods are difficult to achieve timeliness, and manually viewing logs is time-consuming and laborious, which cannot effectively meet the needs of network security assessment.

Method used

The network risk detection method is adopted to obtain real-time data, standardized processing, principal component analysis algorithm dimensionality reduction, classified screening of abnormal data, input into the network evaluation model for risk detection, and generate early warning information based on the evaluation results.

Benefits of technology

It improves the accuracy and efficiency of network security assessment, helps enterprises to identify and rectify security incidents in a timely manner, and improves network security.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118316723B_ABST
    Figure CN118316723B_ABST
Patent Text Reader

Abstract

The present invention relates to the technical field of network security, and discloses a network security assessment method and system based on network risk detection. The method comprises the following steps: obtaining real-time data in a network operation process to form a network data set, and performing standardization processing on the data in the network data set; performing data dimension reduction on the standardized network data set by adopting a principal component analysis algorithm to obtain a feature data set; classifying the data in the feature data set to screen out abnormal data to form an abnormal data set, and inputting the abnormal data set into a network assessment model, performing network risk detection through the network assessment model, and outputting a network assessment result; confirming a comprehensive network security risk value based on the network assessment result, and generating corresponding network security early warning information according to the comprehensive network security risk value; the present invention improves the accuracy and efficiency of network security assessment, and enables enterprises to complete rectification in a timely manner for network security analysis, thereby improving the security of the network.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of network security, and in particular to a network security assessment method and system based on network risk detection. Background Art

[0002] With the popularization of the Internet and the acceleration of the digitalization process, network security issues have become increasingly prominent, which has a profound impact on economic development and social stability. Network security incidents are caused by human or hardware defects or failures, which pose a potential hazard to the information system of key protection units in the metropolitan area network, and even affect the normal provision of services by the information system. At present, the network security incident analysis method is mainly aimed at collecting evidence from a certain server or device after a security incident, such as manually checking the logs of the attacked server or device; it is not only time-consuming and labor-intensive, but also difficult to meet the needs of timely network security assessment; therefore, studying a network risk detection-based network security assessment method and system is of great significance to improving network security. Summary of the invention

[0003] The purpose of the present invention is to solve the above problems and to design a network security assessment method and system based on network risk detection.

[0004] A first aspect of the present invention provides a network security assessment method based on network risk detection, the network security assessment method based on network risk detection comprising the following steps:

[0005] Acquire real-time data during network operation to form a network data set, and perform standardization on the data in the network data set;

[0006] The principal component analysis algorithm is used to reduce the dimension of the standardized network data set to obtain the characteristic data set;

[0007] Classifying the data in the feature data set, filtering out abnormal data to form an abnormal data set, inputting the abnormal data set into a network assessment model, performing network risk detection through the network assessment model, and outputting a network assessment result;

[0008] Based on the network assessment result, a comprehensive network security risk value is confirmed, and corresponding network security warning information is generated according to the comprehensive network security risk value.

[0009] Optionally, in a first implementation of the first aspect of the present invention, the principal component analysis algorithm is used to perform data dimension reduction on the standardized network data set to obtain a feature data set, including:

[0010] Obtain a standardized network data set, where the mean of each feature of the standardized network data set is 0 and the variance is 1;

[0011] Calculate the covariance matrix, perform eigenvalue decomposition on the covariance matrix, and obtain the eigenvalues ​​and corresponding eigenvectors;

[0012] Sort the eigenvalues ​​in descending order, and select the eigenvectors corresponding to the first n eigenvalues ​​as the principal components, where n = 1, 2, ... k;

[0013] The standardized network data set is projected onto the selected principal components to obtain the reduced-dimensional data and generate a feature data set.

[0014] Optionally, in a second implementation of the first aspect of the present invention, classifying the data in the feature data set, screening out abnormal data, and forming an abnormal data set includes:

[0015] Collect historical network traffic data, divide the historical network traffic data, and construct a training set with a certain ratio of normal traffic data and abnormal traffic data;

[0016] Set the AUC value of the initial training set for classifier training, calculate the chi-square value and P value of the features in the training set, and obtain the information gain of the correlation between each feature and the label value;

[0017] Sort the features according to the P value to generate a feature subset, and use the classifier to calculate the AUC value of the feature subset to obtain a new data set;

[0018] Compare the classifier result of the new data set with the preset threshold. If it is greater than the threshold, save the new data set and update the threshold until the optimal new data set is found to obtain the Ig-chi hybrid model.

[0019] The feature data set is input into the Ig-chi hybrid model, the normal flow data and the abnormal flow data in the feature data set are classified, and the abnormal data are screened out to form an abnormal data set.

[0020] Optionally, in a third implementation of the first aspect of the present invention, inputting the abnormal data set into a network assessment model, performing network risk detection through the network assessment model, and outputting a network assessment result includes:

[0021] Obtaining the original sample set, preprocessing the original sample set, obtaining the evaluation factor set of the original sample set, obtaining the network historical data risk level of the original sample set, and matching the evaluation factor set of the original sample set with the historical data risk level;

[0022] Set the number of random forest decision trees to T, perform bootstrap resampling on the evaluation factor set of the original sample set and the corresponding historical data risk level, and generate T training data sets;

[0023] Based on the T training data sets, splitting is performed, and parameter optimization is performed using an artificial bee colony to obtain T different base decision trees, and the AUC values ​​corresponding to the T base decision trees are integrated to obtain a network evaluation model;

[0024] Input the abnormal data set and calculate the corresponding evaluation factors, where the evaluation factors include threat type risk value, threat risk level and user asset value;

[0025] The network evaluation model is used to perform classification detection, and the network evaluation result is output through the AUC weighted voting method.

[0026] Optionally, in a fourth implementation of the first aspect of the present invention, splitting is performed based on the T training data sets, and an artificial bee colony is used to perform parameter optimization to obtain T different base decision trees, including:

[0027] Initialize the artificial bee population randomly, calculate the fitness value of the initial nectar source, use hired bees to search for the nectar source, and if the fitness value of the searched nectar source is better than the fitness value of the initial nectar source, update the fitness value of the nectar source;

[0028] The follower bee calculates the selection probability and mines the fitness value of the nectar source found by the employed bee. If the fitness value of the nectar source is not updated during the iteration process after the employed bee and the follower bee have searched the entire space, the corresponding employed bee becomes a scout bee.

[0029] Obtain the fitness value of the abandoned nectar source, generate the fitness value of the offspring nectar source through crossover, mutation, and selection in genetic learning, and find a better nectar source fitness value;

[0030] Compare the fitness value of the new nectar source. If it is better than the original nectar source, replace it. Repeat the operation to determine whether the maximum number of iterations is exceeded. If so, output the final optimization result.

[0031] Optionally, in a fifth implementation of the first aspect of the present invention, the obtaining of the abandoned nectar source fitness value and generating the offspring nectar source fitness value through crossover, mutation, and selection in genetic learning to find a better nectar source fitness value includes:

[0032] Obtain the fitness value of the abandoned nectar source, and generate the fitness value of the offspring nectar source through the crossover operation in genetic learning;

[0033] Perform mutation operation in genetic learning on each dimension of offspring nectar source fitness value according to random mutation probability;

[0034] According to the selection operation in genetic learning, the fitness value of the randomly generated offspring nectar source and the fitness value of the abandoned nectar source are compared. If it is better than the fitness value of the abandoned nectar source, the genetic learning process is terminated. Otherwise, the scout bee honey-seeking process is repeated until a better nectar source fitness value is found.

[0035] Optionally, in a sixth implementation of the first aspect of the present invention, the determining the network security comprehensive risk value based on the network assessment result, and generating corresponding network security warning information according to the network security comprehensive risk value includes:

[0036] Obtaining the network assessment result, and determining the threat type risk value, threat risk level and user asset value according to the network assessment result;

[0037] Determining the probability of occurrence of a safe time according to the threat type risk value and the threat risk level;

[0038] Determine the degree of harm of the security incident according to the threat risk level and the value of the user's assets;

[0039] Using the probability of occurrence of the security event and the degree of harm of the security incident, the network operation status is scored to determine the comprehensive risk value of network security;

[0040] Determine a network risk level based on the network security comprehensive risk value, wherein the network risk level includes levels 1-5;

[0041] When the network risk level is level 1, the network operation is of low risk; when the network risk level is level 2, the network operation is of medium risk; when the network risk level is level 3, the network operation is of high risk; when the network risk level is level 4, the network operation is of high risk; when the network risk level is level 5, the network operation is of extremely high risk.

[0042] The second aspect of the present invention provides a network security assessment system based on network risk detection, the network security assessment system based on network risk detection includes a standardization processing module, a data dimension reduction module, a network assessment module and an early warning generation module, wherein:

[0043] A standardization processing module is used to obtain real-time data during network operation, form a network data set, and perform standardization processing on the data in the network data set;

[0044] The data dimension reduction module is used to perform data dimension reduction on the standardized network data set using the principal component analysis algorithm to obtain a feature data set;

[0045] A network assessment module is used to classify the data in the feature data set, filter out the data with abnormalities, form an abnormal data set, input the abnormal data set into a network assessment model, perform network risk detection through the network assessment model, and output a network assessment result;

[0046] The warning generation module is used to confirm the comprehensive network security risk value based on the network assessment result, and generate corresponding network security warning information according to the comprehensive network security risk value.

[0047] Optionally, in a first implementation of the second aspect of the present invention, the data dimension reduction module includes an acquisition submodule, an eigenvalue decomposition submodule, a sorting submodule and a selection submodule, wherein:

[0048] The acquisition submodule is used to obtain the standardized network data set, wherein the mean value of each feature of the standardized network data set is 0 and the variance is 1;

[0049] The eigenvalue decomposition submodule is used to calculate the covariance matrix, perform eigenvalue decomposition on the covariance matrix, and obtain the eigenvalues ​​and corresponding eigenvectors;

[0050] The sorting submodule is used to sort the eigenvalues ​​from large to small, and select the eigenvectors corresponding to the first n eigenvalues ​​as the principal components, where n = 1, 2, ... k;

[0051] The selection submodule is used to project the standardized network data set onto the selected principal components to obtain the reduced-dimensional data and generate a feature data set.

[0052] Optionally, in a second implementation of the second aspect of the present invention, the network evaluation module includes a corresponding submodule, a resampling submodule, a splitting submodule, a calculation submodule and a classification detection submodule, wherein:

[0053] The corresponding submodule is used to obtain the original sample set, pre-process the original sample set, obtain the evaluation factor set of the original sample set, obtain the network historical data risk level of the original sample set, and correspond the evaluation factor set of the original sample set with the historical data risk level;

[0054] The resampling submodule is used to set the number of random forest decision trees T, perform bootstrap resampling on the evaluation factor set of the original sample set and the corresponding historical data risk level, and generate T training data sets;

[0055] A splitting submodule is used to split based on the T training data sets, and use artificial bee colonies to optimize parameters to obtain T different base decision trees, integrate the AUC values ​​corresponding to the T base decision trees, and combine them to obtain a network evaluation model;

[0056] The calculation submodule is used to input the abnormal data set and calculate the corresponding evaluation factors, where the evaluation factors include the threat type risk value, the threat risk level and the user asset value;

[0057] The classification detection submodule is used to perform classification detection using the network evaluation model and output the network evaluation result through the AUC weighted voting method.

[0058] In the technical solution provided by the present invention, real-time data in the network operation process is acquired to form a network data set, and the data in the network data set is standardized; the principal component analysis algorithm is used to reduce the dimension of the standardized network data set to obtain a feature data set; the data in the feature data set is classified, and the abnormal data is screened out to form an abnormal data set, and the abnormal data set is input into a network evaluation model, and network risk detection is performed through the network evaluation model to output a network evaluation result; based on the network evaluation result, a comprehensive network security risk value is confirmed, and corresponding network security early warning information is generated according to the comprehensive network security risk value; the present invention improves the accuracy and efficiency of network security evaluation, analyzes network security incidents, helps enterprises complete rectification in a timely manner, and improves network security. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the following detailed description of the preferred embodiment.The drawings are only for the purpose of illustrating the preferred embodiments and are not to be construed as limiting the invention.

[0060] Figure 1 A schematic diagram of a first embodiment of a network security assessment method based on network risk detection provided by an embodiment of the present invention;

[0061] Figure 2 A schematic diagram of a second embodiment of a network security assessment method based on network risk detection provided by an embodiment of the present invention;

[0062] Figure 3 A schematic diagram of a third embodiment of a network security assessment method based on network risk detection provided by an embodiment of the present invention;

[0063] Figure 4 A schematic diagram of the structure of a network security assessment system based on network risk detection provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0064] The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, device, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0065] For ease of understanding, the specific process of the embodiment of the present invention is described below. Figure 1 A schematic diagram of a first embodiment of a network security assessment method based on network risk detection provided by an embodiment of the present invention, the method specifically comprises the following steps:

[0066] Step 101: Acquire real-time data during network operation to form a network data set, and perform standardization on the data in the network data set;

[0067] Step 102: using a principal component analysis algorithm to perform data dimension reduction on the standardized network data set to obtain a feature data set;

[0068] In this embodiment, a standardized network data set is obtained, wherein the mean of each feature of the standardized network data set is 0 and the variance is 1; the covariance matrix is ​​calculated, and the eigenvalue decomposition of the covariance matrix is ​​performed to obtain eigenvalues ​​and corresponding eigenvectors; the eigenvalues ​​are sorted in order from large to small, and the eigenvectors corresponding to the first n eigenvalues ​​are selected as principal components, where n = 1, 2, ...k; the standardized network data set is projected onto the selected principal component to obtain the reduced-dimensional data, thereby generating a feature data set.

[0069] Step 103: classify the data in the feature data set, filter out the abnormal data, form an abnormal data set, input the abnormal data set into the network assessment model, perform network risk detection through the network assessment model, and output the network assessment result;

[0070] Step 104: Based on the network assessment results, confirm the comprehensive network security risk value, and generate corresponding network security warning information according to the comprehensive network security risk value.

[0071] In this embodiment, a network assessment result is obtained, and based on the network assessment result, a threat type risk value, a threat risk level, and a user asset value are determined; based on the threat type risk value and the threat risk level, a probability of a safe event is determined; based on the threat risk level and the user asset value, a degree of harm of a security incident is determined; using the probability of a safe event and the degree of harm of a security incident, the network operation status is scored to determine a comprehensive network security risk value; based on the comprehensive network security risk value, a network risk level is determined, wherein the network risk level includes levels 1-5; when the network risk level is level 1, the network operation status is a relatively low risk; when the network risk level is level 2, the network operation status is a medium risk; when the network risk level is level 3, the network operation status is a relatively high risk; when the network risk level is level 4, the network operation status is a high risk; when the network risk level is level 5, the network operation status is an extremely high risk.

[0072] When the network risk level is level 1, it may be some common network security incidents, such as minor system vulnerabilities, common phishing attacks, etc. These risks may pose certain threats to individual users or small networks, but will not have a significant impact on the entire network system; when the network risk level is level 2, it may be some more complex network attacks, such as malware propagation, targeted phishing attacks, etc. These risks may pose threats to some network systems or important data, and corresponding security protection measures need to be taken; when the network risk level is level 3, it may be attacks on key information systems, such as paralysis of key business systems, leakage of important data, etc. These risks may have a serious impact on organizational operations and information assets, and require high vigilance and strengthening of security protection measures; when the network risk level is level 4, it may be security threats involving important information systems or key infrastructure. These risks pose a serious threat to the organization and require the highest level of security protection measures; when the network risk level is level 5, it may be the highest level of network risk, which may involve security threats to core confidential systems and may have catastrophic consequences, requiring the most stringent and comprehensive security protection measures.

[0073] In an embodiment of the present invention, real-time data in a network operation process is acquired to form a network data set, and the data in the network data set is standardized; a principal component analysis algorithm is used to perform data dimension reduction on the standardized network data set to obtain a feature data set; data in the feature data set is classified, and abnormal data is screened out to form an abnormal data set, and the abnormal data set is input into a network assessment model, and network risk detection is performed through the network assessment model, and a network assessment result is output; based on the network assessment result, a comprehensive network security risk value is confirmed, and corresponding network security early warning information is generated according to the comprehensive network security risk value; the present invention improves the accuracy and efficiency of network security assessment, analyzes network security incidents, helps enterprises complete rectification in a timely manner, and improves network security.

[0074] See also Figure 2 , a schematic diagram of a second embodiment of a network security assessment method based on network risk detection provided by an embodiment of the present invention, the method comprising:

[0075] Step 201: collect historical network traffic data, divide the historical network traffic data, and construct a training set with a certain proportion of normal traffic data and abnormal traffic data;

[0076] Step 202: Set the AUC value of the initial training set for classifier training, calculate the chi-square value and P value of the features in the training set, and obtain the information gain of the correlation between each feature and the label value;

[0077] Step 203: Sort the features according to the P value to generate a feature subset, and use the classifier to calculate the AUC value of the feature subset to obtain a new data set;

[0078] Step 204: compare the classifier result of the new data set with a preset threshold. If it is greater than the threshold, save the new data set and update the threshold until the optimal new data set is found to obtain an Ig-chi hybrid model.

[0079] Step 205: Input the feature data set into the Ig-chi hybrid model, classify the normal traffic data and the abnormal traffic data in the feature data set, filter out the abnormal data, and form an abnormal data set.

[0080] See also Figure 3 , a schematic diagram of a third embodiment of a network security assessment method based on network risk detection provided by an embodiment of the present invention, the method comprising:

[0081] Step 301: obtain an original sample set, pre-process the original sample set, obtain an evaluation factor set of the original sample set, obtain the network historical data risk level of the original sample set, and correspond the evaluation factor set of the original sample set to the historical data risk level;

[0082] Step 302: Set the number of random forest decision trees to T, perform bootstrap resampling on the evaluation factor set of the original sample set and the corresponding historical data risk level, and generate T training data sets;

[0083] Step 303: split based on T training data sets, and use artificial bee colonies to optimize parameters to obtain T different base decision trees, integrate them according to the AUC values ​​corresponding to the T base decision trees, and combine them to obtain a network evaluation model;

[0084] In this embodiment, the artificial bee population is randomly initialized, the fitness value of the initial nectar source is calculated, and the employed bees are used to search for the nectar source. If the fitness value of the searched nectar source is better than the fitness value of the initial nectar source, the nectar source fitness value is updated; the follower bees calculate the selection probability to mine the fitness value of the nectar source searched by the employed bees. If the fitness value of the nectar source is not updated during the iteration process after the employed bees and the follower bees have searched the entire space, the corresponding employed bees become scout bees; the abandoned nectar source fitness value is obtained, and the offspring nectar source fitness value is generated through crossover, mutation, and selection in genetic learning to find a better nectar source fitness value; the new nectar source fitness value is compared, and if it is better than the original nectar source fitness value, it is replaced, and the operation is repeated to determine whether the maximum number of iterations is exceeded. If so, the final optimization result is output.

[0085] In this embodiment, the fitness value of the abandoned nectar source is obtained, and the fitness value of the offspring nectar source is generated through the crossover operation in genetic learning; the mutation operation in genetic learning is performed on each dimension of the offspring nectar source fitness value according to the random mutation probability; the randomly initially generated offspring nectar source fitness value and the abandoned nectar source fitness value are compared according to the selection operation in genetic learning. If it is better than the abandoned nectar source fitness value, the genetic learning process is terminated, otherwise the scout bee honey-seeking process is repeated until a better nectar source fitness value is found.

[0086] Step 304: Input the abnormal data set and calculate the corresponding evaluation factor, where the evaluation factor includes the threat type risk value, the threat risk level and the user asset value;

[0087] Step 305: Use the network evaluation model to perform classification detection and output the network evaluation result through the AUC weighted voting method.

[0088] See also Figure 4 , a schematic diagram of the structure of a network security assessment system based on network risk detection provided by an embodiment of the present invention, the system includes a standardization processing module, a data dimension reduction module, a network assessment module and an early warning generation module, wherein,

[0089] The standardization processing module 401 is used to obtain real-time data during network operation, form a network data set, and perform standardization processing on the data in the network data set;

[0090] The data dimension reduction module 402 is used to perform data dimension reduction on the standardized network data set by using a principal component analysis algorithm to obtain a feature data set;

[0091] The network assessment module 403 is used to classify the data in the feature data set, filter out the abnormal data, form an abnormal data set, input the abnormal data set into the network assessment model, perform network risk detection through the network assessment model, and output the network assessment result;

[0092] The warning generation module 404 is used to confirm the network security comprehensive risk value based on the network assessment result, and generate corresponding network security warning information according to the network security comprehensive risk value.

[0093] In this embodiment, the data dimension reduction module includes an acquisition submodule, an eigenvalue decomposition submodule, a sorting submodule and a selection submodule, wherein:

[0094] The acquisition submodule is used to obtain the standardized network data set, wherein the mean value of each feature of the standardized network data set is 0 and the variance is 1;

[0095] The eigenvalue decomposition submodule is used to calculate the covariance matrix, perform eigenvalue decomposition on the covariance matrix, and obtain the eigenvalues ​​and corresponding eigenvectors;

[0096] The sorting submodule is used to sort the eigenvalues ​​from large to small, and select the eigenvectors corresponding to the first n eigenvalues ​​as the principal components, where n = 1, 2, ... k;

[0097] The selection submodule is used to project the standardized network data set onto the selected principal components to obtain the reduced-dimensional data and generate a feature data set.

[0098] In this embodiment, the network evaluation module includes a corresponding submodule, a resampling submodule, a splitting submodule, a calculation submodule and a classification detection submodule, wherein:

[0099] The corresponding submodule is used to obtain the original sample set, pre-process the original sample set, obtain the evaluation factor set of the original sample set, obtain the network historical data risk level of the original sample set, and correspond the evaluation factor set of the original sample set with the historical data risk level;

[0100] The resampling submodule is used to set the number of random forest decision trees T, perform bootstrap resampling on the evaluation factor set of the original sample set and the corresponding historical data risk level, and generate T training data sets;

[0101] The splitting submodule is used to split based on T training data sets, and use artificial bee colonies to optimize parameters to obtain T different base decision trees, integrate the AUC values ​​corresponding to the T base decision trees, and combine them to obtain a network evaluation model;

[0102] The calculation submodule is used to input the abnormal data set and calculate the corresponding evaluation factors, where the evaluation factors include the threat type risk value, the threat risk level and the user asset value;

[0103] The classification detection submodule is used to perform classification detection using the network evaluation model and output the network evaluation results through the AUC weighted voting method.

[0104] Through the implementation of the above scheme, the system includes a standardization processing module, a data dimension reduction module, a network evaluation module and an early warning generation module; the present invention improves the accuracy and efficiency of network security evaluation, analyzes network security incidents, helps enterprises complete rectification in a timely manner, and improves network security.

[0105] 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 by the above embodiments. The above embodiments and descriptions are only preferred examples of the present invention and are not intended to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A network security assessment system based on network risk detection, characterized in that: The network risk detection-based network security assessment system includes a standardization processing module, a data dimension reduction module, a network assessment module and an early warning generation module, wherein: A standardization processing module is used to obtain real-time data during network operation, form a network data set, and perform standardization processing on the data in the network data set; The data dimension reduction module is used to perform data dimension reduction on the standardized network data set using the principal component analysis algorithm to obtain a feature data set; A network assessment module is used to classify the data in the feature data set, filter out the data with abnormalities, form an abnormal data set, input the abnormal data set into a network assessment model, perform network risk detection through the network assessment model, and output a network assessment result; An early warning generation module is used to determine the network security comprehensive risk value based on the network assessment result, and generate corresponding network security early warning information according to the network security comprehensive risk value; The data dimension reduction module includes an acquisition submodule, an eigenvalue decomposition submodule, a sorting submodule and a selection submodule, wherein: The acquisition submodule is used to obtain the standardized network data set, wherein the mean value of each feature of the standardized network data set is 0 and the variance is 1; The eigenvalue decomposition submodule is used to calculate the covariance matrix, perform eigenvalue decomposition on the covariance matrix, and obtain the eigenvalues ​​and corresponding eigenvectors; The sorting submodule is used to sort the eigenvalues ​​from large to small, and select the eigenvectors corresponding to the first n eigenvalues ​​as the principal components, where n = 1, 2, ... k; A selection submodule is used to project the standardized network data set onto the selected principal components to obtain the data after dimensionality reduction and generate a feature data set; The network evaluation module includes a corresponding submodule, a resampling submodule, a splitting submodule, a calculation submodule and a classification detection submodule, wherein: The corresponding submodule is used to obtain the original sample set, pre-process the original sample set, obtain the evaluation factor set of the original sample set, obtain the network historical data risk level of the original sample set, and correspond the evaluation factor set of the original sample set with the historical data risk level; The resampling submodule is used to set the number of random forest decision trees T, perform bootstrap resampling on the evaluation factor set of the original sample set and the corresponding historical data risk level, and generate T training data sets; A splitting submodule is used to split based on the T training data sets, and use an artificial bee colony to optimize parameters to obtain T different base decision trees, and integrate the AUC values ​​corresponding to the T base decision trees to obtain a network evaluation model; The calculation submodule is used to input the abnormal data set and calculate the corresponding evaluation factors, where the evaluation factors include the threat type risk value, the threat risk level and the user asset value; A classification detection submodule, used to perform classification detection using the network evaluation model and output the network evaluation result by AUC weighted voting method; The step of classifying the data in the feature data set and filtering out abnormal data to form an abnormal data set includes: Collect historical network traffic data, divide the historical network traffic data, and construct a training set with a certain ratio of normal traffic data and abnormal traffic data; Set the AUC value of the initial training set for classifier training, calculate the chi-square value and P value of the features in the training set, and obtain the information gain of the correlation between each feature and the label value; Sort the features according to the P value to generate a feature subset, and use the classifier to calculate the AUC value of the feature subset to obtain a new data set; Compare the classifier result of the new data set with the preset threshold. If it is greater than the threshold, save the new data set and update the threshold until the optimal new data set is found to obtain the Ig-chi hybrid model. Inputting the feature data set into the Ig-chi hybrid model, classifying the normal flow data and the abnormal flow data in the feature data set, screening out the abnormal data, and forming an abnormal data set; Based on the T training data sets, splitting is performed and parameter optimization is performed using an artificial bee colony to obtain T different base decision trees, including: Initialize the artificial bee population randomly, calculate the fitness value of the initial nectar source, use hired bees to search for the nectar source, and if the fitness value of the searched nectar source is better than the fitness value of the initial nectar source, update the fitness value of the nectar source; The follower bee calculates the selection probability and mines the fitness value of the nectar source found by the employed bee. If the fitness value of the nectar source is not updated during the iteration process after the employed bee and the follower bee have searched the entire space, the corresponding employed bee becomes a scout bee. Obtain the fitness value of the abandoned nectar source, generate the fitness value of the offspring nectar source through crossover, mutation, and selection in genetic learning, and find a better nectar source fitness value; Compare the fitness value of the new nectar source. If it is better than the original nectar source, replace it. Repeat the operation to determine whether the maximum number of iterations is exceeded. If so, output the final optimization result. The method of obtaining the abandoned nectar source fitness value and generating the offspring nectar source fitness value through crossover, mutation, and selection in genetic learning to find a better nectar source fitness value includes: Obtain the fitness value of the abandoned nectar source, and generate the fitness value of the offspring nectar source through the crossover operation in genetic learning; Perform mutation operation in genetic learning on each dimension of offspring nectar source fitness value according to random mutation probability; According to the selection operation in genetic learning, the fitness value of the randomly generated offspring nectar source and the fitness value of the abandoned nectar source are compared. If the fitness value is better than the abandoned nectar source, the genetic learning process is terminated. Otherwise, the scout bee honey-seeking process is repeated until a better nectar source fitness value is found. The method of determining a network security comprehensive risk value based on the network assessment result and generating corresponding network security warning information according to the network security comprehensive risk value includes: Obtaining the network assessment result, and determining the threat type risk value, threat risk level and user asset value according to the network assessment result; Determining the probability of occurrence of a safe time according to the threat type risk value and the threat risk level; Determine the degree of harm of the security incident according to the threat risk level and the value of the user's assets; Using the probability of occurrence of the security event and the degree of harm of the security incident, the network operation status is scored to determine the comprehensive risk value of network security; Determine a network risk level based on the network security comprehensive risk value, wherein the network risk level includes levels 1-5; When the network risk level is level 1, the network operation is of low risk; when the network risk level is level 2, the network operation is of medium risk; when the network risk level is level 3, the network operation is of high risk; when the network risk level is level 4, the network operation is of high risk; when the network risk level is level 5, the network operation is of extremely high risk.