Dynamic Security Baseline Modeling Method

Through the dynamic security baseline modeling method, the security baseline model of the industrial Internet system is monitored and updated in real time, and the problem that the static model cannot cope with changing environments is solved, the abnormal detection and response capabilities are improved, and the system's security protection level is enhanced.

CN118784379BActive Publication Date: 2024-11-19NANJING SINOVATIO TECHNOLOGY CO LTD +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411276678.2
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-09-12
Publication Date
2024-11-19
Estimated Expiration
2044-09-12

AI Technical Summary

Technical Problem

The existing static security baseline model cannot effectively respond to the changing network environment and system operating conditions in industrial Internet systems, resulting in lagging identification and response in the face of new threats and abnormal behaviors, reducing the accuracy and reliability of security monitoring.

Method used

Using dynamic security baseline modeling method, the initial security baseline model is constructed by collecting real-time running data, and monitoring the data in real-time compared with it, identifying abnormal data, and dynamically updating the security baseline model to adapt to the security status of the system under different time periods and operating conditions. .

Benefits of technology

It improves the accuracy and response capabilities of abnormal detection, reduces false alarms and underreporting problems in static models, enhances the system's adaptability and protection level to potential security threats, and ensures timely detection and handling of real security threats.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118784379B_ABST
    Figure CN118784379B_ABST
Patent Text Reader

Abstract

The present invention relates to a dynamic security baseline modeling method, including: collecting real-time operation data in an industrial Internet system; constructing an initial security baseline model based on the collected real-time operation data, wherein the initial security baseline model includes system parameters and behavior patterns under normal operating conditions; monitoring the industrial Internet system in real time and collecting real-time monitoring data; comparing the real-time monitoring data with the initial security baseline model to identify abnormal data that deviates from the initial security baseline model; dynamically updating the security baseline model based on the identified abnormal data to generate a new security baseline model; applying the new security baseline model to the security monitoring and protection of the industrial Internet system to detect and respond to potential security threats in real time. The present invention improves the detection and response capabilities of abnormal behaviors in the industrial Internet system and enhances the security and stability of the system by dynamically updating the security baseline model.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of network security technology, and in particular to a dynamic security baseline modeling method. Background Art

[0002] In the existing technology, industrial Internet systems are widely used in manufacturing, energy, transportation and other fields. They connect various devices and systems through the network to achieve intelligent management and control. In order to ensure the safe operation of these systems, the security baseline modeling method is usually adopted to build a baseline model reflecting the normal state of the system by analyzing the normal operation data of the system. This model is used to compare real-time monitoring data to identify potential security threats.

[0003] However, existing security baseline models are usually static and lack a dynamic update mechanism, which makes them unable to effectively cope with the ever-changing network environment and system operating conditions. These static models are often unable to identify and respond to new threats and abnormal behaviors in a timely manner, resulting in a lag in security protection. In addition, static models may also have false positives or false negatives when responding to changes in system status, reducing the accuracy and reliability of security monitoring.

[0004] Therefore, it is necessary to develop a dynamic security baseline modeling method to overcome the above-mentioned defects in the prior art. Summary of the invention

[0005] The present application provides a dynamic security baseline modeling method to improve the accuracy of anomaly detection in industrial Internet systems.

[0006] The dynamic security baseline modeling method provided in this application includes:

[0007] Collecting real-time operation data in the industrial Internet system, including network traffic data, equipment operation data and user behavior data;

[0008] Based on the collected real-time operation data, an initial safety baseline model is constructed, wherein the initial safety baseline model includes system parameters and behavior patterns under normal operating conditions;

[0009] Performing real-time monitoring on the industrial Internet system and collecting real-time monitoring data, wherein the real-time monitoring data is similar to the real-time operation data;

[0010] Comparing the real-time monitoring data with the initial security baseline model to identify abnormal data that deviates from the initial security baseline model;

[0011] Based on the identified abnormal data, dynamically update the security baseline model to generate a new security baseline model, wherein the new security baseline model reflects the security status of the system under different time periods and operating conditions;

[0012] The new security baseline model is applied to the security monitoring and protection of industrial Internet systems to detect and respond to potential security threats in real time.

[0013] Furthermore, the collection of real-time operation data in the industrial Internet system includes:

[0014] Collect network traffic data, including capturing packets and analyzing packet headers and packet contents;

[0015] Collect equipment operation data, including obtaining operation status and operation logs from sensors and controllers of various industrial equipment;

[0016] Collect user behavior data, including recording user login time, operation records, and permission changes.

[0017] Furthermore, the initial security baseline model is constructed based on the collected real-time operation data, including:

[0018] Remove noise and outliers through data cleaning and preprocessing;

[0019] Use cluster analysis to group similar operation modes and extract system parameters under normal operating conditions;

[0020] Use time series analysis to determine the fluctuation range and change trend of system parameters in different time periods;

[0021] Combine the mean and standard deviation of each feature value to build an initial security baseline model.

[0022] Furthermore, the real-time monitoring of the industrial Internet system and the collection of real-time monitoring data include:

[0023] Collect network traffic data regularly and perform content analysis through deep packet inspection;

[0024] Collect equipment operation data regularly and perform preliminary processing and filtering through edge computing;

[0025] Record user behavior data in real time and use behavior analysis models to identify potential abnormal operations.

[0026] Furthermore, the real-time monitoring data is compared with the initial security baseline model to identify abnormal data that deviates from the initial security baseline model, including:

[0027] Use multi-scale analysis methods to decompose real-time monitoring data in frequency and time domains;

[0028] The degree of data deviation is quantified by calculating the Euclidean distance between the real-time monitoring data and the initial safety baseline model;

[0029] Combined with adaptive thresholds, abnormal data that deviates from the initial security baseline model is identified and marked as potential threats.

[0030] Furthermore, dynamically updating the security baseline model based on the identified abnormal data includes:

[0031] Perform multi-scale decomposition on abnormal data and extract abnormal features at different time scales;

[0032] Based on the extracted abnormal features at different time scales, the deviation between the abnormal features and the parameters in the initial security baseline model is calculated according to the following formula 1:

[0033] ,

[0034] in, For the The deviation of the abnormal features; For real-time data eigenvalues; The first The mean of the eigenvalues; is the standard deviation of the 𝑖th eigenvalue in the initial security baseline model; For the The first order differences of the eigenvalues; The first The mean of the first-order differences of the eigenvalues; The first The standard deviation of the first differences of the eigenvalues; For the The second order difference of the eigenvalues; The first The mean of the second-order differences of the eigenvalues; The first The standard deviation of the second-order differences of the eigenvalues; and is the weight coefficient;

[0035] According to the following formula 2, calculate the total deviation :

[0036] ,

[0037] in, For the The weight coefficient of each parameter; For the The deviation of the abnormal features;

[0038] Based on the total deviation, parameter weights and thresholds in the initial security baseline model are adjusted.

[0039] Furthermore, the multi-scale decomposition of the abnormal data to extract abnormal features at different time scales includes:

[0040] Perform wavelet transform on the abnormal data to decompose it into multiple components with different frequencies and time scales;

[0041] According to the following formula 3, the energy characteristics at each time scale after wavelet transformation are calculated: :

[0042] ,

[0043] in, For the Scale in time The wavelet coefficients of is the total number of time points;

[0044] Based on energy characteristics , according to the following formula 4, define the weighted energy difference :

[0045] ,

[0046] in, and are the mean and standard deviation of all scale energy features, respectively; For the The weighting coefficient of each scale;

[0047] Perform singular value decomposition on the wavelet coefficients at each time scale, and extract the nonlinear principal components of the singular values ​​according to the following formula 5: :

[0048] ,

[0049] in, For the The scale of singular values;

[0050] According to the following formula 6, the multi-scale feature vector is calculated :

[0051] ,

[0052] in, is the weight coefficient.

[0053] Furthermore, adjusting the parameter weights and thresholds in the initial security baseline model based on the total deviation degree includes:

[0054] Compare the total deviation with the preset threshold to determine whether the security baseline model needs to be updated;

[0055] If the total deviation exceeds the preset threshold, the weight coefficient and threshold of each parameter in the initial security baseline model are dynamically adjusted according to the size of the total deviation so that it can adapt to the new abnormal characteristics.

[0056] Furthermore, the application of the new security baseline model to the security monitoring and protection of the industrial Internet system includes:

[0057] Deploy new security baseline models to edge devices and centralized management platforms;

[0058] Detect abnormal behavior by comparing real-time monitoring data with the new security baseline model;

[0059] Automatically generate security reports to record detailed information about anomaly detection, including anomaly type, occurrence time, and treatment measures;

[0060] Trigger security policies in real time, including quarantining suspicious devices, adjusting access permissions, and notifying system administrators.

[0061] Furthermore, dynamically updating the security baseline model based on the identified abnormal data includes:

[0062] Record each abnormal data identified; based on the identified abnormal data, use machine learning algorithms to adjust model parameters to adapt to the newly discovered abnormal patterns.

[0063] The beneficial effects of the technical solution provided by this application include:

[0064] (1) By collecting real-time operating data in the industrial Internet system and building an initial security baseline model, the normal operating status of the system can be accurately reflected. By comparing the real-time monitoring data with the initial security baseline model, abnormal data that deviates from the normal state can be more accurately identified, thereby improving the accuracy of anomaly detection. (2) The security baseline model is dynamically updated to reflect the security status of the system under different time periods and operating conditions. This dynamic adjustment mechanism enables the security baseline model to adapt to emerging threats and changing operating environments in a timely manner, enhancing the system's adaptability and responsiveness to potential security threats. (3) By continuously updating and optimizing the security baseline model, it can accurately capture and reflect system changes and reduce the common false positives and false negatives in static models. In this way, the reliability and effectiveness of the security monitoring system can be greatly improved, ensuring that real security threats are discovered and handled in a timely manner. (4) The updated security baseline model is applied to the security monitoring and protection of the industrial Internet system, which can achieve real-time detection and response to potential security threats. This not only improves the security protection level of the system, but also enables timely intervention in the early stages of the threat to avoid security incidents from causing significant impact on the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 This is a flowchart of a dynamic security baseline modeling method provided in the first embodiment of the present application. DETAILED DESCRIPTION

[0066] Many specific details are described in the following description to facilitate a full understanding of the present application. However, the present application can be implemented in many other ways than those described herein, and those skilled in the art can make similar generalizations without violating the connotation of the present application, so the present application is not limited by the specific implementation disclosed below.

[0067] Embodiment: The first embodiment of the present application provides a dynamic security baseline modeling method. Figure 1 , which is a schematic diagram of the first embodiment of the present application. Figure 1 A dynamic security baseline modeling method is provided in detail for the first embodiment of the present application.

[0068] Step S101: Collect real-time operation data in the industrial Internet system, where the real-time operation data includes network traffic data, equipment operation data, and user behavior data.

[0069] Furthermore, the collection of real-time operation data in the industrial Internet system includes:

[0070] Collect network traffic data, including capturing packets and analyzing packet headers and packet contents;

[0071] Collect equipment operation data, including obtaining operation status and operation logs from sensors and controllers of various industrial equipment;

[0072] Collect user behavior data, including recording user login time, operation records, and permission changes.

[0073] In step S101, collecting real-time operation data in the industrial Internet system is the basis for building a dynamic security baseline model. This step specifically includes the following aspects:

[0074] First, collect data from the network traffic in the industrial Internet system. This includes capturing data packets from various network nodes and device interfaces. The specific method can be to use a distributed data acquisition module, which is arranged on different network nodes of the system, and use network traffic monitoring tools such as Wireshark to capture and record data packet information in real time. The collected data packet information includes not only the header information of the data packet (such as source address, destination address, protocol type, port number, etc.), but also the payload content of the data packet. The capture frequency of the data packet can be adjusted according to the network load and demand of the system to ensure the real-time and comprehensiveness of the data.

[0075] Second, collect equipment operation data. Industrial Internet systems usually contain various sensors, actuators, and controllers, which continuously generate and exchange operation data. Operation data can be obtained directly from the device through interface protocols (such as Modbus, OPC UA, etc.), or by accessing the device logging system to extract the operating status and operation history of the device. These operation data include the start and shutdown time of the equipment, operating parameters (such as temperature, pressure, speed, etc.), alarm information, maintenance records, etc. In order to ensure the real-time nature of the data, edge computing devices can be used to perform preliminary processing and transmission of data on site.

[0076] Third, collect user behavior data. User behavior data refers to various activity records generated by system users during operation. These records include user login time, operation commands, system configuration changes, file access, permission changes, etc. By deploying user behavior monitoring software, these operation logs can be automatically recorded and stored. In addition, behavioral analysis models can be used to analyze user operations and extract key behavioral features.

[0077] In the process of data collection, data preprocessing is required to ensure the integrity and accuracy of the data. This includes removing noise data, filling missing data, and standardizing the data. For example, filters are used to remove noise data in network traffic data, interpolation methods are used to fill missing values ​​in device operation data, and user behavior data is normalized to unify the data format.

[0078] After completing the above data collection and preprocessing, the real-time operation data obtained will include a comprehensive record of network traffic data, device operation data, and user behavior data. These data will provide a basis for subsequent steps to ensure the accuracy and effectiveness of the construction of the initial security baseline model and real-time monitoring.

[0079] Step S102: constructing an initial security baseline model based on the collected real-time operation data, wherein the initial security baseline model includes system parameters and behavior patterns under normal operating conditions.

[0080] In step S102, building an initial security baseline model based on the collected real-time operation data is the key to ensuring that the system can identify normal and abnormal behaviors. First, the collected real-time operation data needs to be preprocessed. This includes denoising, normalizing, and filling missing values. Denoising can use filtering algorithms, such as mean filtering or median filtering, to remove random noise in the data. Normalization can scale the data to the same range for subsequent analysis. Filling missing values ​​can use interpolation or use the average value of the data to ensure data integrity.

[0081] Next, the preprocessed data needs to be statistically analyzed. The purpose of statistical analysis is to extract system parameters and behavior patterns under normal operating conditions. For network traffic data, statistical parameters such as the transmission rate of each data packet, the frequency distribution of packet header information, and the length of the data packet can be calculated. For device operation data, the device's operating time, operation frequency, and state changes can be analyzed. For user behavior data, the number of user logins, the frequency of use of operation commands, and permission changes can be counted.

[0082] After extracting these statistical parameters, cluster analysis methods can be used to group similar operation modes. For example, using the K-means clustering algorithm, network traffic data, device operation data, and user behavior data can be clustered into multiple groups, each representing a typical operation mode. In this way, the main behavior patterns of the system under normal operation can be identified.

[0083] In addition, time series analysis methods can be used to determine the fluctuation range and change trend of system parameters in different time periods. For example, the autoregressive moving average model (ARIMA) is used to model network traffic data and predict its normal range of change in future time periods. Similar methods can also be used to build time series models for device operation data and user behavior data.

[0084] Through the above steps, various statistical parameters and behavior patterns of the system under normal operation are obtained. These parameters and patterns will constitute the initial security baseline model. In order to improve the robustness of the model, the mean and standard deviation of multiple eigenvalues, as well as other statistics such as median and interquartile range, can be introduced. These statistics are combined to form a multidimensional feature space to more comprehensively reflect the normal state of the system.

[0085] Finally, the initial security baseline model needs to be verified. By applying the model to historical data, its ability to identify known normal and abnormal behaviors can be tested. In this way, the accuracy and effectiveness of the model can be evaluated, and necessary adjustments and optimizations can be made. The verified and optimized initial security baseline model will serve as a benchmark for subsequent real-time monitoring and anomaly detection to ensure the security of the system under different time periods and operating conditions.

[0086] Furthermore, the initial security baseline model is constructed based on the collected real-time operation data, including:

[0087] Noise and outliers are removed through data cleaning and preprocessing. Cluster analysis methods are used to group similar operating modes and extract system parameters under normal operating conditions. Time series analysis is used to determine the fluctuation range and change trend of system parameters in different time periods. The initial safety baseline model is constructed by combining the mean and standard deviation of each characteristic value.

[0088] When implementing the dynamic security baseline modeling method, building the initial security baseline model is one of the key steps. This process mainly includes data cleaning and preprocessing, cluster analysis, time series analysis, and model building in combination with eigenvalues. Each step is crucial to ensure that the final baseline model can accurately reflect the normal operating status of the system.

[0089] First of all, data cleaning and preprocessing are basic steps to ensure data quality. In this process, the collected real-time operation data needs to be carefully checked and processed to remove noise and outliers. Noise data refers to random or error data that does not conform to the normal pattern. Filters (such as mean filtering or median filtering) can be used to remove these noises. Outliers refer to abnormal data points that deviate significantly from the normal range. These data can be detected and removed through statistical methods such as standard deviation analysis or IQR (interquartile range) method. After data cleaning, the data needs to be standardized to eliminate the scale differences between different data sources. The commonly used standardization methods are normalization or Z-score standardization.

[0090] Next, cluster analysis methods are used to group similar operating modes. The purpose of this step is to identify and extract characteristic parameters of the system under normal operating conditions. Common clustering algorithms include K-means, DBSCAN (density-based spatial clustering algorithm) and hierarchical clustering. In the implementation process, the K-means algorithm can be selected to cluster the data. First, the number of cluster centers, that is, the K value, is determined, and then the data points are assigned to the closest cluster center according to the distance from each data point to each cluster center, and the iteration is repeated until the clustering result is stable. In this way, a large amount of operating data can be grouped into several categories, each of which represents a typical operating mode, which reflects the operating behavior of the system under normal conditions.

[0091] Subsequently, the time series analysis method is used to determine the fluctuation range and change trend of system parameters in different time periods. Time series analysis is an effective method for processing and analyzing time data. Commonly used time series models include ARIMA (autoregressive integrated moving average model) and exponential smoothing. First, arrange the data in chronological order, analyze its autocorrelation and partial autocorrelation graphs, and select a suitable model for fitting. Through time series analysis, the periodic changes and trends of system parameters within a day, week or month can be identified, and the fluctuation range and change trend of system parameters under normal operating conditions can be determined.

[0092] Finally, the initial security baseline model is constructed by combining the mean and standard deviation of each characteristic value. In the previous steps, the characteristic parameters of the system have been extracted and their time series characteristics have been analyzed. Now, the mean and standard deviation of each characteristic parameter need to be calculated to establish a comprehensive baseline model. The mean reflects the central trend of the characteristic parameter, and the standard deviation describes the fluctuation range of the characteristic parameter. When building the model, these means and standard deviations are combined to form a multidimensional feature space, in which each dimension corresponds to a characteristic parameter and is described by its mean and standard deviation. In this multidimensional feature space, the normal operating state of the system is defined as all characteristic parameters are near their means and fluctuate within a certain standard deviation range.

[0093] Through these detailed steps, an initial security baseline model can be constructed, which accurately reflects the parameters and behavior patterns of the system under normal operation. This baseline model will serve as the basis for real-time monitoring and anomaly detection in subsequent steps, ensuring that the system can identify and respond to potential security threats in a timely manner.

[0094] Step S103: Perform real-time monitoring on the industrial Internet system and collect real-time monitoring data, where the real-time monitoring data is similar to the real-time operation data.

[0095] In step S103, real-time monitoring of the industrial Internet system and collection of real-time monitoring data are key links to ensure the continuous and safe operation of the system. First, real-time monitoring devices need to be deployed at each network node and device interface of the system. These devices include network traffic monitors, device status sensors, and user behavior monitoring software. These devices should be able to collect network traffic data, device operation data, and user behavior data in real time, and transmit these data to the central monitoring system.

[0096] For real-time monitoring of network traffic data, deep packet inspection (DPI) technology can be used to perform detailed analysis of each data packet flowing through the network. DPI not only analyzes the header information of the data packet, such as source address, destination address, protocol type, and port number, but also analyzes the content of the data packet to identify potential threats and abnormal behaviors. The network traffic monitor should set a reasonable sampling rate to ensure the real-time and integrity of the data while avoiding excessive data processing burden.

[0097] For real-time monitoring of equipment operation data, it is necessary to install status sensors and data acquisition modules on each industrial equipment. These sensors and modules should be able to collect the equipment's operating status, operating parameters and fault information in real time, and transmit them to the data concentration system through industrial protocols such as Modbus, OPC UA, etc. The real-time monitoring data includes equipment start and stop time, temperature, pressure, speed and other parameters, as well as equipment alarm and maintenance records.

[0098] Real-time monitoring of user behavior data relies on user behavior monitoring software deployed in the system. The software can record user login time, operation commands, system configuration changes, and permission changes. To ensure the accuracy of the data, the monitoring software should have a high-precision timestamp function and be able to integrate with the system's logging mechanism to synchronize user behavior records.

[0099] In the process of real-time data collection, data preprocessing is also required. This includes data filtering, denoising and normalization. For example, for network traffic data, filtering algorithms can be used to remove network noise, and for device operation data, interpolation methods can be used to fill missing values ​​to ensure data continuity and integrity. User behavior data needs to be formatted to be consistent with other data types.

[0100] The pre-processed real-time monitoring data should be transmitted to the central monitoring system in a timely manner and stored in an efficient database for subsequent analysis and processing. In order to ensure the reliability and security of data transmission, encrypted transmission protocols and data verification technology can be used to prevent data from being tampered with or lost during transmission.

[0101] Through these steps, the real-time monitoring system can continuously collect the operation data of the industrial Internet system to ensure the real-time, complete and accurate data. These real-time monitoring data are similar to the real-time operation data collected previously, and will serve as an important basis for comparison with the initial security baseline model in subsequent steps to help identify abnormal behaviors and potential threats in the system.

[0102] Furthermore, the real-time monitoring of the industrial Internet system and the collection of real-time monitoring data include:

[0103] Collect network traffic data regularly and perform content analysis through deep packet inspection;

[0104] Collect equipment operation data regularly and perform preliminary processing and filtering through edge computing;

[0105] Record user behavior data in real time and use behavior analysis models to identify potential abnormal operations.

[0106] In the dynamic security baseline modeling method described above, real-time monitoring of the industrial Internet system and collection of real-time monitoring data are crucial links. This process includes regular collection of network traffic data, equipment operation data, and real-time recording of user behavior data, and analysis and processing of these data to ensure the security and normal operation of the system.

[0107] First of all, regular collection of network traffic data and content analysis through deep packet inspection (DPI) are the basis of real-time monitoring. Regular collection of network traffic data means continuously capturing data packets from the network at set time intervals. This can be achieved by deploying network traffic monitoring devices on network nodes and switches. These devices can regularly capture every data packet passing through the network, including its header information (such as source address, destination address, protocol type, port number) and payload content. Then, deep packet inspection technology is used to perform detailed analysis on the captured data packets. DPI technology not only parses the header information of the data packet, but also goes deep into the payload content of the data packet to identify the specific application protocol and data content. In this way, potential threats in network traffic can be identified, such as malware communication, abnormal protocol usage, and suspicious data transmission behavior.

[0108] Secondly, regular collection of equipment operation data and preliminary processing and filtering through edge computing are important steps to ensure the normal operation of the equipment. Sensors and data acquisition modules are installed on each industrial equipment. These sensors and modules can monitor the operating parameters of the equipment in real time, such as temperature, pressure, speed, etc. Through industrial protocols (such as Modbus, OPC UA, etc.), these data are regularly transmitted to edge computing devices. Edge computing devices are responsible for preliminary processing and filtering of the collected data, including removing noise, calibrating sensor data, and filtering out irrelevant or duplicate data. After preliminary processing, the equipment operation data is more accurate and reliable and can be used for further analysis and monitoring.

[0109] Recording user behavior data in real time and using behavior analysis models to identify potential abnormal operations are important means to ensure system security. Deploy behavior monitoring software on each user terminal and server. The software can record various user operations in real time, including login time, operation commands, file access, and permission changes. Whenever a user performs an operation, the behavior monitoring software will record the operation time, commands used, modified files, and permission changes in detail. These data are transmitted to the central monitoring system in real time for storage and analysis. Using behavior analysis models, potential abnormal operations in user behavior can be identified. For example, if a user frequently logs into the system during non-working hours, or executes an operation command that does not meet their permissions, the behavior analysis model can automatically identify and generate an alarm.

[0110] During the entire monitoring process, all collected data should be uniformly transmitted to the central data processing platform for comprehensive analysis. The data processing platform can correlate and analyze network traffic data, device operation data, and user behavior data to identify complex cross-domain threats. For example, by combining abnormal communications in network traffic with abnormal operations of a certain device and suspicious user behavior, potential security threats in the system can be more accurately identified.

[0111] Step S104: Compare the real-time monitoring data with the initial security baseline model to identify abnormal data that deviates from the initial security baseline model.

[0112] In step S104, the real-time monitoring data is compared with the initial security baseline model to identify abnormal data that deviates from the initial security baseline model, which is an important part of ensuring system security. First, the real-time monitoring data collected in step S103 needs to be aligned and synchronized with the initial security baseline model constructed in step S102. The purpose of alignment is to ensure the consistency of data in the time dimension, which can be achieved through timestamp calibration.

[0113] Next, the difference analysis method is used to compare the parameters and behavior patterns in the real-time monitoring data and the initial security baseline model. For network traffic data, the characteristics such as the data packet transmission rate, the frequency distribution of the packet header information, and the length of the data packet can be compared. Specifically, if the data packet transmission rate in a certain period of time is significantly higher than the normal value in the baseline model, or the frequency distribution of the data packet has an abnormal peak, it can be determined that there may be an abnormality in the network traffic data of this section.

[0114] For equipment operation data, abnormalities can be identified by comparing parameters such as the equipment's operating status, operating frequency, and state changes. For example, a sudden increase or decrease in the operating time of a piece of equipment, a significant increase or decrease in operating frequency, or frequent state changes may all be indicators of abnormalities. These parameters can be compared by calculating the difference between the mean and standard deviation in the real-time data and the baseline model.

[0115] In the comparison of user behavior data, abnormal behavior can be found by analyzing the user's login time, operation commands, and permission changes. For example, if a user frequently logs into the system during non-working hours, or uses operation commands that are inconsistent with their permissions, it can be determined that the user's behavior is abnormal. Similarly, these behaviors can be statistically analyzed to identify deviations from the baseline model.

[0116] After completing the above comparative analysis, the threshold judgment method can be used to identify abnormal data. Specifically, a reasonable threshold range is set based on the statistical parameters (such as mean and standard deviation) in the initial security baseline model. If the real-time monitoring data exceeds this range, it is judged as abnormal. For example, assuming that the mean of a parameter in the baseline model is , the standard deviation is , then the threshold range can be set to ,in Is an adjustment factor. If the real-time data exceeds this range, it is considered abnormal.

[0117] In addition, in order to improve the accuracy of anomaly detection, more advanced detection algorithms can be introduced. For example, machine learning algorithms such as support vector machines (SVM) or deep learning models can be used to compare the feature patterns in real-time data and baseline models to automatically identify complex anomalies. These algorithms can improve the accuracy and robustness of detection through continuous optimization of training sets and test sets.

[0118] Through the above steps, the real-time monitoring data is compared with the initial security baseline model in detail, which can effectively identify abnormal data that deviates from the normal state. These identified abnormal data will provide an important basis for dynamically updating the security baseline model in the subsequent steps, ensuring that the system can respond to and handle potential security threats in a timely manner.

[0119] Furthermore, comparing the real-time monitoring data with the initial security baseline model to identify abnormal data that deviates from the initial security baseline model includes:

[0120] Use multi-scale analysis methods to decompose real-time monitoring data in frequency and time domains;

[0121] The degree of data deviation is quantified by calculating the Euclidean distance between the real-time monitoring data and the initial safety baseline model;

[0122] Combined with adaptive thresholds, abnormal data that deviates from the initial security baseline model is identified and marked as potential threats.

[0123] In the dynamic security baseline modeling method described above, comparing the real-time monitoring data with the initial security baseline model and identifying abnormal data that deviates from the initial security baseline model is an important part of ensuring system security. This process includes three main steps: multi-scale analysis, Euclidean distance calculation, and adaptive threshold determination. Each step needs to be carefully implemented to ensure accurate identification and response to potential threats.

[0124] First, the real-time monitoring data is decomposed in the frequency domain and time domain using a multi-scale analysis method. The purpose of this process is to extract different time and frequency features in the data for a more comprehensive comparison. Multi-scale analysis can use techniques such as wavelet transform to decompose the data into components of different frequencies and time scales. Wavelet transform can decompose the signal into multiple scales, each scale containing information of different frequencies, thereby revealing details and overall trends in the data. For example, for network traffic data, high-frequency short-term fluctuations and low-frequency long-term trends can be decomposed; for equipment operation data, short-term operation changes and long-term operation patterns can be identified; for user behavior data, the characteristics of daily operations and abnormal behaviors can be distinguished.

[0125] Next, the degree of data deviation is quantified by calculating the Euclidean distance between the real-time monitoring data and the initial safety baseline model. The Euclidean distance is a standard method for measuring the distance between two points in a multidimensional space, and can be used to quantify the difference between the monitoring data and the baseline model. Specifically, the corresponding parameters in the real-time monitoring data and the baseline model are represented as multidimensional vectors, and then the Euclidean distance between the two vectors is calculated. , the calculation formula is:

[0126] ,

[0127] in, It is real-time monitoring data in The value of the dimension, The baseline model is The value of the dimension, is the total number of pending data. By calculating the Euclidean distance at each time point, we can get the degree of deviation of the real-time monitoring data from the baseline model. The larger the distance value, the more obvious the deviation of the data from the baseline model.

[0128] Finally, combined with the adaptive threshold, abnormal data that deviates from the initial security baseline model is identified and marked as a potential threat. The adaptive threshold is a threshold that is dynamically adjusted according to the system's operating conditions and historical data, which can improve the accuracy and sensitivity of anomaly detection. In specific implementation, you can first set the initial threshold range, for example, set a static threshold range based on the mean and standard deviation in the baseline model. Then, by real-time monitoring of the data change trend and the statistical distribution of the Euclidean distance, dynamically adjust the threshold range to adapt it to the current operating status.

[0129] During real-time detection, if the Euclidean distance of a data point exceeds the current adaptive threshold, the data point is marked as abnormal data. Data marked as abnormal will be further analyzed and processed, recorded and alerted as potential threats. This process can be achieved by writing automated detection scripts to ensure that the system can identify and respond to abnormal situations in a timely manner.

[0130] Through these detailed steps, real-time monitoring data can be effectively compared with the initial security baseline model, abnormal data that deviates from the baseline can be accurately identified, and potential threats can be marked and handled in a timely manner.

[0131] Step S105: Based on the identified abnormal data, dynamically update the security baseline model to generate a new security baseline model, wherein the new security baseline model reflects the security status of the system under different time periods and operating conditions.

[0132] In step S105, dynamically updating the security baseline model based on the identified abnormal data to generate a new security baseline model is a key step to ensure that the system can continuously reflect the latest security status. First, after the abnormal data is identified, it is necessary to analyze the data in detail to determine its characteristics and patterns. These characteristics and patterns will help understand the root cause of the anomaly and guide how to adjust the baseline model.

[0133] The analysis of abnormal data can be conducted from multiple perspectives. For network traffic data, the analysis should include checking the source, destination, transmission protocol, and content characteristics of the abnormal data packets. For device operation data, the analysis should include checking the changes in the device's operating parameters during the abnormal period, such as temperature, pressure, and operating time. For user behavior data, the analysis should include checking the user's operation records, login time, and command execution during the abnormal period.

[0134] After the detailed analysis is completed, these abnormal data can be used to adjust the parameters of the initial security baseline model. The adjustment process includes updating the statistical parameters in the model, such as the mean, standard deviation, and threshold range, to adapt to the new operating environment and operating conditions. For example, if it is found that the traffic of a certain network node frequently exceeds the normal range in a specific time period, the threshold of the node traffic can be appropriately increased to make the model more sensitive to this change.

[0135] In addition, the dynamic update of the security baseline model also needs to consider the adaptability of the system. Adaptive algorithms such as Kalman filtering or adaptive mean algorithms can be introduced to enable the model to update its parameters in real time. These algorithms automatically adjust the model parameters through the continuous input of real-time data to maintain the accuracy and reliability of the model.

[0136] In order to achieve dynamic updates, an automated model update mechanism can be designed. This mechanism should include a data acquisition module, a data analysis module, and a model update module. The data acquisition module is responsible for real-time monitoring and data collection, the data analysis module is responsible for preprocessing and anomaly analysis of the data, and the model update module is responsible for adjusting the parameters of the baseline model according to the analysis results.

[0137] In the model update process, the data acquisition module first transmits real-time data to the data analysis module. The data analysis module preprocesses the data, removes noise and irrelevant information, and extracts abnormal features. These features are then input into the model update module. The model update module automatically adjusts the model parameters and updates the mean, standard deviation, and threshold range based on the comparison results between the abnormal features and the initial baseline model.

[0138] The updated security baseline model should be verified to ensure its effectiveness. Verification methods can include applying the new baseline model to historical data to check its ability to identify known anomalies, and applying the model to a simulated environment to observe its response to newly emerged anomalies. If the model can accurately identify anomalies and reduce false positives and false negatives, it indicates that the model update is successful.

[0139] Through the above steps, the new security baseline model can reflect the latest security status of the system under different time periods and operating conditions, ensuring that the industrial Internet system can detect and respond to potential security threats in a timely manner. This dynamic update process is a continuous cycle. As the system operating environment changes and new threats emerge, the baseline model is continuously adjusted and optimized to maintain the security and stability of the system.

[0140] In the dynamic security baseline modeling method, dynamically updating the security baseline model based on the identified abnormal data is an important step to improve the accuracy and responsiveness of the model. This process includes multi-scale decomposition of abnormal data, extracting abnormal features at different time scales, and calculating the deviation and total deviation based on these features to dynamically adjust the model parameters.

[0141] First, the abnormal data is decomposed at multiple scales to extract the characteristics of the data at different time scales. Multiscale decomposition can be achieved through wavelet transform, which can decompose the signal into different frequency components and reveal the characteristics of the data at each time scale. In specific operations, the abnormal data is subjected to wavelet transform to obtain decomposition results at different scales. These decomposition results include high-frequency parts (indicating short-term changes) and low-frequency parts (indicating long-term trends). Through these decomposition results, the characteristics of abnormal data can be analyzed more comprehensively.

[0142] After extracting the abnormal features at different time scales, the deviation between each abnormal feature and each parameter in the initial security baseline model is calculated according to Formula 1. Formula 1 is as follows:

[0143] ,

[0144] In Formula 1:

[0145] For the The deviation of abnormal features,

[0146] For real-time data eigenvalues, which are the original eigenvalues ​​extracted from real-time monitoring data.

[0147] The first The mean of the characteristic values, which is the average value obtained by statistical analysis of normal operating data.

[0148] The first The standard deviation of the eigenvalues, which is a statistic that represents the fluctuation range of the eigenvalues.

[0149] For the The first-order difference of the eigenvalues, that is, the difference between the current eigenvalue and the eigenvalue at the previous moment,

[0150] The first The mean of the first-order differences of the eigenvalues, which is the average value obtained by statistical analysis of the first-order differences of the normal operation data.

[0151] The first The standard deviation of the first-order differences of the eigenvalues, which is a statistic that represents the fluctuation range of the first-order differences,

[0152] For the The second-order difference of the eigenvalues, that is, the rate of change between the current eigenvalue and the eigenvalues ​​at the previous two moments,

[0153] The first The mean of the second-order differences of the eigenvalues, which is the average value obtained by statistical analysis of the second-order differences of the normal operation data.

[0154] The first The standard deviation of the second-order differences of the eigenvalues, which is a statistic that represents the fluctuation range of the second-order differences,

[0155] and is the weight coefficient, which is used to adjust the influence of the first-order difference and the second-order difference in calculating the deviation degree.

[0156] By calculating the deviation of each eigenvalue , which can quantify the difference between each feature value and the baseline model. Then, the total deviation is calculated according to Formula 2 :

[0157] ,

[0158] In Formula 2:

[0159] is the total deviation, which is the weighted sum of all feature deviations.

[0160] For the The weight coefficients of the parameters are used to adjust the influence of different features in the total deviation calculation.

[0161] For the The deviation of the abnormal features is calculated by formula 1.

[0162] is the total number of features,

[0163] By calculating the total deviation , the overall degree of abnormality can be evaluated. If the total deviation exceeds a preset threshold, it indicates that the current system state deviates significantly from the normal operating state and the baseline model needs to be updated.

[0164] Based on the total deviation, the parameter weights and thresholds in the initial security baseline model are adjusted. Specifically, by analyzing the total deviation, it is possible to identify which features contribute most to the anomaly, and then dynamically adjust the parameter weights and thresholds in the model based on this information to make the model more sensitive and accurate in reflecting the current system operation status.

[0165] Through these detailed steps, the dynamic update of the initial security baseline model can be achieved so that it can effectively reflect the security status of the system under different time periods and operating conditions.

[0166] Furthermore, the multi-scale decomposition of the abnormal data to extract abnormal features at different time scales includes:

[0167] Perform wavelet transform on the abnormal data to decompose it into multiple components with different frequencies and time scales;

[0168] According to the following formula 3, the energy characteristics at each time scale after wavelet transformation are calculated: :

[0169] ,

[0170] in, For the Scale in time The wavelet coefficients of is the total number of time points;

[0171] Based on energy characteristics , according to the following formula 4, define the weighted energy difference ,

[0172] ,

[0173] in, and are the mean and standard deviation of all scale energy features, respectively; For the The weighting coefficient of each scale;

[0174] Perform singular value decomposition on the wavelet coefficients at each time scale, and extract the nonlinear principal components of the singular values ​​according to the following formula 5: :

[0175] ,

[0176] in, For the The scale of singular values;

[0177] According to the following formula 6, the multi-scale feature vector is calculated :

[0178] ,

[0179] in, is the weight coefficient.

[0180] First, the abnormal data is subjected to wavelet transform to decompose it into multiple components with different frequencies and time scales. Wavelet transform is a time-frequency analysis method that can decompose the signal into multiple scales with different resolutions to reveal the local and global characteristics of the signal. , the wavelet transform produces a set of wavelet coefficients , these coefficients represent the signal at this scale and time point The strength on.

[0181] Next, the energy characteristics of the wavelet transform at each time scale are calculated according to Formula 3: ,

[0182] ,

[0183] In this formula:

[0184] For the Energy characteristics at each time scale,

[0185] For the Scale in time The wavelet coefficients of are obtained through wavelet transform.

[0186] is the total number of time points, that is, the number of sampling points on the time axis after the signal is decomposed.

[0187] Energy characteristics Indicates that the signal is The total energy at a scale reflects the feature strength at that scale. After calculating the energy features, the weighted energy difference is defined based on these energy features. , which is calculated as Formula 4:

[0188] ,

[0189] In this formula,

[0190] For the The weighted energy difference of the scales,

[0191] For the The energy characteristics of each scale are calculated by formula 3.

[0192] and are the mean and standard deviation of all scale energy features, respectively. These values ​​are obtained by Statistical analysis revealed that

[0193] For the The weighting coefficient of each scale is used to adjust the weight of different scales in the energy difference calculation.

[0194] Weighted Energy Difference By normalizing the energy features, the magnitude differences between different scales are eliminated, and the weighting coefficients Emphasize or weaken the impact of specific dimensions.

[0195] After that, singular value decomposition (SVD) is performed on the wavelet coefficients at each time scale to extract the nonlinear principal components of the singular values. Singular value decomposition is a matrix decomposition method that decomposes the wavelet coefficient matrix into the product of three matrices, which contain eigenvalues ​​and eigenvectors. According to Formula 5, the nonlinear principal components of the singular values ​​are extracted.

[0196] ,

[0197] In this formula:

[0198] For the The nonlinear principal components of the scales,

[0199] For the The scale of Singular values ​​are obtained by performing SVD on the wavelet coefficient matrix.

[0200] is the number of singular values ​​selected, that is, only the first The largest singular values ​​represent the main features.

[0201] is a nonlinear transformation of singular values, which is used to emphasize the importance of larger singular values ​​to features.

[0202] Nonlinear principal components It combines singular values ​​and their nonlinear transformation to reflect the main feature information at different scales.

[0203] Finally, the multi-scale feature vector is calculated according to Formula 6 :

[0204] ,

[0205] In this formula,

[0206] For the The multi-scale feature vector of scales.

[0207] For the The weighted energy difference of each scale is calculated by formula 4.

[0208] For the The nonlinear principal components of the scales are calculated by formula 5.

[0209] is the weight coefficient, which is used to adjust the contribution of weighted energy difference and nonlinear principal component in the eigenvector.

[0210] Multi-scale feature vector The nonlinear principal components of energy features and singular values ​​are combined to provide a comprehensive indicator for describing the abnormal characteristics at each scale.

[0211] Through these detailed steps, the abnormal data is decomposed at multiple scales and the abnormal features at different time scales are extracted, which can effectively improve the ability to identify abnormal behaviors.

[0212] Furthermore, adjusting the parameter weights and thresholds in the initial security baseline model based on the total deviation degree includes:

[0213] Compare the total deviation with the preset threshold to determine whether the security baseline model needs to be updated;

[0214] If the total deviation exceeds the preset threshold, the weight coefficient and threshold of each parameter in the initial security baseline model are dynamically adjusted according to the size of the total deviation so that it can adapt to the new abnormal characteristics.

[0215] In the dynamic security baseline modeling method, adjusting the parameter weights and thresholds in the initial security baseline model based on the total deviation is a key step to ensure that the model can adapt to new abnormal characteristics and continue to operate effectively. In specific implementation, this process includes comparing the total deviation with the preset threshold and dynamically adjusting the model parameters when necessary.

[0216] First, the total deviation is calculated It is an indicator that measures the difference between the current monitoring data and the initial security baseline model. This total deviation is compared with the preset threshold to determine whether the security baseline model needs to be updated. The preset threshold is determined based on the normal operation status of the system and historical data analysis, which can reflect the maximum allowable deviation of the system under normal circumstances. If it is less than or equal to this threshold, it means that the current monitoring data is consistent with the initial baseline model and there is no need to update the model.

[0217] However, if the total deviation If the preset threshold is exceeded, it means that there is a significant deviation between the current monitoring data and the baseline model, indicating that the system may have undergone changes or new abnormal characteristics have emerged. In this case, the initial security baseline model needs to be updated. The update process includes dynamically adjusting the weight coefficients and thresholds of each parameter in the model so that the model can more accurately reflect the current system status and newly emerging characteristics.

[0218] Specifically, the adjustment process can be carried out in the following steps:

[0219] 1. Analyze the size and source of the total deviation: First, we need to analyze the total deviation Perform a detailed analysis to determine which features contribute the most to the total deviation. , the parameters that need the most adjustment can be identified;

[0220] 2. Adjust the weight coefficient: Dynamically adjust the weight coefficient of each feature in the initial security baseline model according to the size of the total deviation and the contribution of each feature , weight coefficient Determines the influence of each feature in the total deviation calculation. For features with greater contributions, their weight coefficients can be increased to make the model pay more attention to the changes of these features; for features with smaller contributions, their weight coefficients can be appropriately reduced;

[0221] 3. Adjust the threshold: When adjusting the weight coefficient, the threshold in the model also needs to be updated. The threshold adjustment can be made based on the trend of the total deviation and historical data. The new threshold should be able to adapt to the changes in the current system while avoiding too many false positives and false negatives. The specific adjustment method can include recalculating a suitable threshold based on the distribution of the total deviation in the recent period of time;

[0222] 4. Verification and optimization: After adjusting the weight coefficients and thresholds, the new security baseline model needs to be verified and optimized. This can be done by applying the new model to historical data or a simulated environment to check its detection effect on known anomalies and normal situations, ensuring that the adjusted model can accurately identify new abnormal features while maintaining a good description of the normal operating state. Through these detailed steps, the initial security baseline model can be dynamically adjusted based on the total deviation degree so that it can adapt to changes in the system and new abnormal features. This process ensures the continued effectiveness and accuracy of the model, enabling it to provide reliable security monitoring and early warning in practical applications.

[0223] Furthermore, dynamically updating the security baseline model based on the identified abnormal data includes:

[0224] Record each abnormal data identified; based on the identified abnormal data, use machine learning algorithms to adjust model parameters to adapt to the newly discovered abnormal patterns.

[0225] In the dynamic security baseline modeling method, dynamically updating the security baseline model based on the identified abnormal data is an important step to ensure that the model can adapt to the changing security environment and newly discovered abnormal patterns. This process includes recording each identified abnormal data and using machine learning algorithms to adjust the model parameters so that the model can effectively reflect the new abnormal characteristics.

[0226] First, when abnormal data is identified during system operation, it is necessary to record these abnormal data in detail. The content of the record should include the specific characteristic value of the abnormal data, the time when the abnormality occurred, the type of abnormality, and related context information (for example, the operating state of the system when the abnormality occurred, user behavior, etc.). These recorded data should be saved in a structured database for subsequent analysis and processing.

[0227] The abnormal data identified each time not only provides basic data for dynamically updating the model, but also helps to understand the abnormal behavior patterns of the system in different situations. The recorded abnormal data should be organized and classified for better analysis. For example, the abnormal data can be classified according to different feature types, occurrence frequency and severity to identify the most representative abnormal patterns.

[0228] After accumulating enough abnormal data, the initial security baseline model can be adjusted using machine learning algorithms. By learning from data, machine learning algorithms can identify complex patterns and relationships and automatically adjust model parameters to improve the accuracy and robustness of the model.

[0229] First, choose an appropriate machine learning algorithm. Common machine learning algorithms include support vector machines (SVM), random forests, neural networks, etc. The choice of algorithm depends on the characteristics of the data and actual needs. For multi-dimensional feature data, support vector machines and neural networks usually have better performance.

[0230] Next, the recorded abnormal data is divided into a training set and a test set. The training set is used to train the machine learning model to learn to identify abnormal patterns; the test set is used to verify the performance of the model to ensure that it can work effectively on unseen data.

[0231] In the training phase, the abnormal data is first preprocessed, including data cleaning, normalization, and feature extraction. Then, the preprocessed data is input into the machine learning algorithm for model training. The training process is an iterative optimization process. The machine learning algorithm continuously adjusts the model parameters to minimize the prediction error and improve the accuracy of identifying abnormal data.

[0232] After the model training is completed, the model is verified using the test set. During the verification process, the performance of the model is checked by evaluating the model's accuracy, recall, F1 score and other indicators. If the model performs well on the test set, it means that the trained model parameters can effectively identify abnormal patterns.

[0233] Finally, the trained machine learning model parameters are applied to the initial security baseline model. Specific operations include updating the weight coefficients, thresholds, and other related parameters in the model so that the model can adapt to newly discovered abnormal patterns. This process is dynamic. As more abnormal data is accumulated and identified, the machine learning algorithm can regularly retrain and update the model parameters to ensure that the model always reflects the latest system status and security threats.

[0234] Through these detailed steps, the security baseline model can be dynamically updated using machine learning algorithms based on the identified abnormal data, so that it can adapt to newly discovered abnormal patterns.

[0235] Step S106: Apply the new security baseline model to the security monitoring and protection of the industrial Internet system to detect and respond to potential security threats in real time. In step S106, the new security baseline model is applied to the security monitoring and protection of the industrial Internet system to detect and respond to potential security threats in real time. This is the last and critical step to ensure the safe operation of the system. First, the updated security baseline model needs to be deployed to each monitoring node and central control platform of the system. The deployment process should include synchronizing the model parameters and behavior patterns to all relevant devices to ensure that all monitoring points in the entire system use the latest baseline model. During the deployment process, automated tools can be used to distribute the model update package to each monitoring node. These tools should have an efficient distribution mechanism to ensure that the model covers the entire network in the shortest time. At the same time, the security of model deployment must also be considered. Through encrypted transmission and verification mechanisms, the model data is prevented from being tampered with or lost during transmission. Once the new security baseline model is successfully deployed, the system enters the real-time monitoring stage. Each monitoring node will continue to monitor network traffic, device operation, and user behavior based on the latest baseline model. Whenever the monitored data is compared with the baseline model, if the data deviates from the normal range, the system will immediately generate an alert. This alert will not only be recorded in the log system, but also notify the system administrator in real time so that they can take corresponding measures quickly.

[0236] In addition, in order to enhance the ability to respond to potential threats, the system can also preset some automatic response strategies. These strategies can automatically execute corresponding protective measures according to the type and severity of the threat. For example, for minor anomalies, an alarm can be triggered and a log can be recorded; for serious anomalies, the affected device or node can be automatically isolated to prevent the threat from spreading. Automatic response strategies should be flexible and configurable to adapt to different security requirements and operating environments.

[0237] After receiving the alarm, the system administrator can view detailed abnormal data and analysis reports through the central control platform. The report should include information such as the specific characteristics of the abnormal data, the degree of deviation, the time of occurrence, and the scope of impact. Based on this information, the administrator can further analyze and confirm the root cause of the abnormality and take appropriate remedial measures.

[0238] In order to ensure the effectiveness and reliability of the new model, the system should also conduct regular security assessments and performance tests. The assessment content includes the model's detection accuracy of anomalies, false positive and false negative rates, system response time, etc. Through these assessments, problems in the model can be discovered and resolved in a timely manner, further optimizing the system's security performance.

[0239] In addition, the system should have self-learning and self-adaptive capabilities. By continuously collecting and analyzing new operating data, the system can automatically adjust and optimize the security baseline model so that it always adapts to the current operating environment and security requirements. This self-adaptive capability can be achieved by introducing machine learning and artificial intelligence technologies to automatically adjust model parameters and update strategies.

[0240] Through the above detailed steps, applying the updated security baseline model to the security monitoring and protection of the industrial Internet system can effectively improve the security protection level of the system, achieve real-time detection and rapid response to potential threats, and ensure the continuous and stable operation of the system.

[0241] Furthermore, the application of the new security baseline model to the security monitoring and protection of the industrial Internet system includes: deploying the new security baseline model to edge devices and centralized management platforms; detecting abnormal behavior by comparing real-time monitoring data with the new security baseline model; automatically generating security reports to record detailed information on anomaly detection, including anomaly type, occurrence time, and handling measures; and triggering security policies in real time, including isolating suspicious devices, adjusting access rights, and notifying system administrators.

[0242] In the dynamic security baseline modeling method described, applying the new security baseline model to the security monitoring and protection of the industrial Internet system is an important step to ensure that the system can detect and respond to potential threats in real time. This process involves deploying the new model into the system and using the model for real-time monitoring and protection operations.

[0243] First, deploy the new security baseline model to edge devices and the centralized management platform. Edge devices are usually located at the edge of the network, close to the data source, and can process and analyze data in real time. The centralized management platform is a centralized control center responsible for coordinating and managing the security monitoring of the entire system. To achieve deployment, the parameters and rules of the new model can be uploaded to the edge devices and the centralized management platform, enabling them to use the latest baseline model for data analysis and anomaly detection. The deployment process needs to ensure the correct transmission and configuration of the model parameters to prevent any possible errors or security vulnerabilities.

[0244] Next, abnormal behavior is detected by comparing the real-time monitoring data with the new security baseline model. After receiving the real-time data, the edge devices and centralized management platform compare it with the parameters and rules in the new model to identify behaviors that deviate from the normal baseline. For example, if abnormal patterns appear in network traffic, device operation, or user behavior, these anomalies will be detected immediately. The real-time comparison process requires efficient computing power and reliable data processing algorithms to ensure timely and accurate identification of anomalies.

[0245] After detecting abnormal behavior, the system automatically generates security reports. These reports record the detected abnormal information in detail, including the type of abnormality (such as network attacks, equipment failures, or abnormal user behavior), the time of occurrence, and the treatment measures taken by the system. Security reports can help system administrators quickly understand the security status of the system and provide a basis for subsequent security analysis and audits. The report generation process needs to be able to extract relevant information from the detection system and organize it into a format that is easy to understand and analyze.

[0246] At the same time, the system triggers security policies in real time to respond to detected abnormal behaviors. Specific security policies include isolating suspicious devices, adjusting access rights, and notifying system administrators. Isolating suspicious devices can be achieved by disconnecting the device from the network or limiting its network activities to prevent the spread of potential threats. Adjusting access rights can include reducing the permission level of suspicious users and limiting their access to key system resources. Notifying system administrators reminds them to pay attention to and handle detected security events by sending alarms or notifications.

[0247] Through these detailed steps, the new security baseline model can be effectively applied in the industrial Internet system. The system can not only detect and respond to potential security threats in real time, but also generate detailed security reports and execute automated protective measures.

[0248] Embodiment 2: A second embodiment of the present application provides an electronic device, the electronic device comprising: a processor;

[0249] The memory is used to store a program, and when the program is read and executed by the processor, the dynamic security baseline modeling method provided in the first embodiment of the present application is executed.

[0250] Embodiment 3: The third embodiment of the present application provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the dynamic security baseline modeling method provided in the first embodiment of the present application is executed.

[0251] Although the present application is disclosed as above in the form of a preferred embodiment, it is not intended to limit the present application. Any technical personnel in this field may make possible changes and modifications without departing from the spirit and scope of the present application. Therefore, the scope of protection of the present application shall be based on the scope defined by the claims of the present application.

Claims

1. Dynamic security baseline modeling method, characterized by: The method comprises the following steps: S101. Collect real-time operation data in the industrial Internet system, where the real-time operation data includes network traffic data, equipment operation data, and user behavior data; S102, constructing an initial security baseline model based on the collected real-time operation data, wherein the initial security baseline model includes system parameters and behavior patterns under normal operating conditions; S103, performing real-time monitoring on the industrial Internet system and collecting real-time monitoring data, where the real-time monitoring data is similar to the real-time operation data; S104, comparing the real-time monitoring data with the initial security baseline model, and identifying abnormal data that deviates from the initial security baseline model; S105. Based on the identified abnormal data, dynamically update the security baseline model to generate a new security baseline model, wherein the new security baseline model reflects the security status of the system under different time periods and operating conditions; S106. Applying the new security baseline model to security monitoring and protection of industrial Internet systems to detect and respond to potential security threats in real time; In S103, the real-time monitoring of the industrial Internet system and the collection of real-time monitoring data include: Collect network traffic data regularly and perform content analysis through deep packet inspection; Collect equipment operation data regularly and perform preliminary processing and filtering through edge computing; Record user behavior data in real time and use behavior analysis models to identify potential abnormal operations; Among them, in S104, comparing the real-time monitoring data with the initial security baseline model to identify abnormal data that deviates from the initial security baseline model includes: Use multi-scale analysis methods to decompose real-time monitoring data in frequency and time domains; The degree of data deviation is quantified by calculating the Euclidean distance between the real-time monitoring data and the initial safety baseline model; Combined with adaptive thresholds, abnormal data that deviates from the initial security baseline model is identified and marked as potential threats; In S105, dynamically updating the security baseline model based on the identified abnormal data includes: Perform multi-scale decomposition on abnormal data and extract abnormal features at different time scales; Based on the extracted abnormal features at different time scales, the deviation between the abnormal features and the parameters in the initial security baseline model is calculated according to the following formula 1: Among them, D i is the deviation of the ith abnormal feature; X i is the i-th eigenvalue in real-time data; μ i,baseline is the mean of the i-th eigenvalue in the initial security baseline model; σ i,bseline is the standard deviation of the i-th eigenvalue in the initial security baseline model; ΔX i is the first-order difference of the i-th eigenvalue; Δμ i,baseline is the mean of the first-order difference of the i-th eigenvalue in the initial security baseline model; Δσ i,baseline is the standard deviation of the first-order difference of the i-th eigenvalue in the initial security baseline model; Δ 2 X i is the second-order difference of the i-th eigenvalue; Δ 2 μ i,baseline is the mean of the second-order difference of the i-th eigenvalue in the initial security baseline model; Δ 2 σ i,baseline is the standard deviation of the second-order difference of the i-th eigenvalue in the initial security baseline model; α and β are weight coefficients; According to the following formula 2, calculate the total deviation D: Among them, w i is the weight coefficient of the i-th parameter; D i is the deviation degree of the i-th abnormal feature; Based on the total deviation, the parameter weights and thresholds in the initial security baseline model are adjusted. The multi-scale decomposition of abnormal data to extract abnormal features at different time scales includes: Perform wavelet transform on the abnormal data to decompose it into multiple components with different frequencies and time scales; According to the following formula 3, the energy characteristic E at each time scale after wavelet transform is calculated: i : Among them, W i,t is the wavelet coefficient of the i-th scale at time t; T is the total number of time points; Based on the energy feature E i , according to the following formula 4, define the weighted energy difference WE i : Among them, E mean and E std are the mean and standard deviation of all scale energy features respectively; w i is the weighting coefficient of the i-th scale; Perform singular value decomposition on the wavelet coefficients at each time scale, and extract the nonlinear principal component NPC of the singular value according to the following formula 5: i : Among them, σ i,j is the jth singular value of the i-th scale; According to the following formula 6, the multi-scale feature vector MSF is calculated i : MSF i =WE i +c·NPC i (6), Among them, γ is the weight coefficient, The adjusting the parameter weights and thresholds in the initial security baseline model based on the total deviation degree includes: Compare the total deviation with the preset threshold to determine whether the security baseline model needs to be updated; If the total deviation exceeds the preset threshold, the weight coefficient and threshold of each parameter in the initial security baseline model are dynamically adjusted according to the size of the total deviation so that it can adapt to the new abnormal characteristics.

2. The dynamic security baseline modeling method according to claim 1, characterized in that: In S101, the collecting of real-time operation data in the industrial Internet system includes: Collect network traffic data, including capturing packets and analyzing packet headers and packet contents; Collect equipment operation data, including obtaining operation status and operation logs from sensors and controllers of various industrial equipment; Collect user behavior data, including recording user login time, operation records, and permission changes.

3. The dynamic security baseline modeling method according to claim 1, characterized in that: In S102, the initial security baseline model is constructed based on the collected real-time operation data, including: Remove noise and outliers through data cleaning and preprocessing; Use cluster analysis to group similar operation modes and extract system parameters under normal operating conditions; Use time series analysis to determine the fluctuation range and change trend of system parameters in different time periods; Combine the mean and standard deviation of each feature value to build an initial security baseline model.

4. The dynamic security baseline modeling method according to claim 1, characterized in that: In S106, applying the new security baseline model to security monitoring and protection of the industrial Internet system includes: Deploy new security baseline models to edge devices and centralized management platforms; Detect abnormal behavior by comparing real-time monitoring data with the new security baseline model; Automatically generate security reports to record detailed information about anomaly detection, including anomaly type, occurrence time, and treatment measures; Trigger security policies in real time, including quarantining suspicious devices, adjusting access permissions, and notifying system administrators.

5. The dynamic security baseline modeling method according to claim 1, characterized in that: The dynamically updating the security baseline model based on the identified abnormal data includes: Record each abnormal data identified; Based on the identified abnormal data, machine learning algorithms are used to adjust the model parameters to adapt them to the newly discovered abnormal patterns.