Effective network security event monitoring method and system based on dynamic AI

By adopting a dynamic AI-based network security incident monitoring method in the hospital data system, and using the decision tree model to automatically identify and respond to potential threats, the problem of insufficient network security awareness in the hospital data system is solved, and efficient monitoring of network security incidents is achieved and handle.

CN119094224APending Publication Date: 2024-12-06HANGZHOU SHUAN WORKSHOP TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411318726.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-09-20
Publication Date
2024-12-06

AI Technical Summary

Technical Problem

The lack of cybersecurity awareness and professional skills in hospital data systems leads to the inability to timely identify and report potential cybersecurity incidents, and there is a serious risk of information leakage and damage.

Method used

Using a dynamic AI-based method, by collecting network traffic data and system log data from medical information systems, cleaning and standardizing processing, training decision tree models, automatically triggering security response measures based on event severity scores, and displaying them through the visual management interface Processing results.

Benefits of technology

It realizes timely discovery and response to potential network security threats, protects patient data and system security, and improves the monitoring and handling efficiency of network security incidents.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119094224A_ABST
    Figure CN119094224A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of network security monitoring, in particular to an effective network security event monitoring method and system based on dynamic AI, and the method comprises the steps: collecting and storing network flow data and system log data from a medical information system; cleaning and standardizing the network flow data and the system log data; training a decision tree model by adopting access time, request types and source IP reputation score features based on historical network flow data and historical system log data to obtain a monitoring model; automatically triggering corresponding safety response measures according to the severity score of the monitoring model on the event; and displaying a processing result of the network security event by adopting a visual management interface. Therefore, potential network security threats can be found and responded in time, and patient data and system security are protected.
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 monitoring, and in particular to an effective network security event monitoring method and system based on dynamic AI. Background Art

[0002] Cybersecurity incidents refer to a series of security threats such as information leakage, destruction, loss or illegal use caused by security vulnerabilities, attacks, errors or accidents in computer systems, networks, programs or devices. These incidents may cause serious damage to personal privacy, corporate secrets, national interests, and even affect social stability and national security. There are many types of cybersecurity incidents, including computer viruses, phishing, hacker attacks, information leakage, denial of service attacks, etc.

[0003] Cybersecurity incident monitoring of hospital data systems is a critical task that involves protecting patient privacy, maintaining continuity of medical services, and ensuring compliance with legal requirements. Cybersecurity incident monitoring of hospital data systems faces many challenges, such as hospital staff may lack sufficient cybersecurity awareness and professional skills to identify and report potential security incidents in a timely manner. Summary of the invention

[0004] The purpose of the present invention is to provide an effective network security event monitoring method and system based on dynamic AI, aiming to timely discover and respond to potential network security threats and protect patient data and system security.

[0005] To achieve the above-mentioned object, the present invention provides an effective network security event monitoring method based on dynamic AI, including collecting network traffic data and system log data from a medical information system and storing them;

[0006] Clean and standardize network traffic data and system log data;

[0007] Based on historical network traffic data and historical system log data, the decision tree model is trained using access time, request type, and source IP reputation score features to obtain a monitoring model.

[0008] Automatically trigger corresponding security response measures based on the severity score of the event according to the monitoring model;

[0009] A visual management interface is used to display the processing results of network security incidents.

[0010] The specific steps of collecting network traffic data and system log data from the medical information system and storing them include:

[0011] Identifying sources of log files in a medical information system, the sources of log files including operating system logs, network device logs, security device logs, and application logs;

[0012] Determine the source of network traffic data, wherein the source of network traffic data includes switches, routers, firewalls and intrusion detection systems;

[0013] Configure the working parameters of data collection tools according to the format and structure of log and traffic data;

[0014] The collected network traffic data and system log data are transmitted to the central storage system for storage.

[0015] The working parameters include log path, filter conditions and log format analysis.

[0016] The specific steps of cleaning and standardizing the network traffic data and system log data include:

[0017] Identify, process and remove outliers in network traffic data and system log data;

[0018] For categorical data, use one-hot encoding for conversion;

[0019] The numerical data were standardized using the normalization method.

[0020] The specific steps of identifying, processing and deleting abnormal values ​​in network traffic data and system log data include:

[0021] Delete log entries or traffic records that are not relevant to network security event monitoring;

[0022] Remove duplicate records from network traffic data and system log data;

[0023] Delete the missing data.

[0024] The specific steps of training the decision tree model based on the historical network traffic data and the historical system log data using the access time, request type, and source IP reputation score features to obtain the monitoring model include:

[0025] Based on known security incidents in historical data, mark each data sample as a security incident.

[0026] Extract access time, request type, and source IP reputation score from data samples as features;

[0027] Divide the data samples into training set, validation set and test set;

[0028] The monitoring model is trained using the decision tree algorithm using the training set.

[0029] Among them, the security incidents include intrusion attempts, data leakage and abnormal login.

[0030] The specific steps of automatically triggering corresponding security response measures based on the severity score of the event according to the monitoring model include:

[0031] Set different scoring thresholds based on operational requirements;

[0032] Predefine a group of security response measures and set a specific score range for each response measure, including blocking malicious IP, isolating infected terminals, sending alerts to security analysts, and recording event details;

[0033] When the severity score of an event exceeds the preset threshold, the decision logic module automatically triggers the corresponding response measures;

[0034] Records operational details of all automatically triggered response measures.

[0035] In a second aspect, the present invention also provides an effective network security incident monitoring system based on dynamic AI, including a data collection module, a standardization module, a monitoring model generation module, a processing module and a visualization module:

[0036] The data collection module is used to collect network traffic data and system log data from the medical information system and store them;

[0037] The standardization module is used to clean and standardize network traffic data and system log data;

[0038] The monitoring model generation module is used to train a decision tree model based on historical network traffic data and historical system log data using access time, request type, and source IP reputation score features to obtain a monitoring model;

[0039] The processing module is used to automatically trigger corresponding security response measures according to the severity score of the event by the monitoring model;

[0040] The visualization module is used to display the processing results of network security events using a visualization management interface.

[0041] An effective network security incident monitoring method and system based on dynamic AI of the present invention first collects network traffic data and system log data from a medical information system and stores them in a secure database. These data include information such as access time, request type, source IP address, and the reputation score of the source IP, which reflects the historical behavior and credibility of the source IP. Next, the collected network traffic data and system log data are cleaned and standardized. The cleaning process includes removing duplicate records, filling missing values, processing outliers, etc. to ensure the integrity and accuracy of the data. Standardization processing includes converting the data into a format suitable for machine learning algorithms, such as converting timestamps to hours of the day, and standardizing or normalizing numerical data. Based on the historical network traffic data and historical system log data after cleaning and standardization, the decision tree model is trained using a decision tree algorithm using features such as access time, request type, and source IP reputation score. The decision tree model can accurately predict network security incidents by learning normal and abnormal network behavior patterns. The trained decision tree model is deployed in the production environment of the medical information system for real-time monitoring of network security incidents. According to the severity score of the incident by the monitoring model, the system can automatically trigger corresponding security response measures. These response measures may include blocking malicious IPs, isolating infected terminals, sending alerts to security analysts, recording event details, etc., to ensure that network security incidents are handled promptly and effectively. In order to better monitor and analyze network security incidents, a visual management interface is used to display the processing results of network security incidents. The visual management interface intuitively displays the decision-making process of the decision tree model, the distribution and processing status of security incidents, enabling security analysts to quickly understand the network security status and take corresponding measures. Through this effective network security incident monitoring method based on dynamic AI, medical information systems can promptly detect and respond to potential network security threats and protect patient data and system security. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0043] Figure 1 It is a flowchart of an effective network security event monitoring method based on dynamic AI according to the first embodiment of the present invention.

[0044] Figure 2 It is a flowchart of the first embodiment of the present invention for collecting and storing network traffic data and system log data from a medical information system.

[0045] Figure 3 It is a flow chart of cleaning and standardizing network traffic data and system log data according to the first embodiment of the present invention.

[0046] Figure 4 It is a flowchart of the first embodiment of the present invention for identifying, processing and deleting abnormal values ​​in network traffic data and system log data.

[0047] Figure 5 This is a flowchart of the first embodiment of the present invention, which trains a decision tree model based on historical network traffic data and historical system log data using access time, request type, and source IP reputation score features to obtain a monitoring model.

[0048] Figure 6 This is a flow chart of automatically triggering corresponding security response measures based on the severity score of an event according to the monitoring model in the first embodiment of the present invention. DETAILED DESCRIPTION

[0049] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, and should not be construed as limiting the present invention.

[0050] First embodiment

[0051] See also Figure 1 to Figure 6 The present invention provides an effective network security incident monitoring method based on dynamic AI, comprising:

[0052] S101 collects and stores network traffic data and system log data from the medical information system;

[0053] The specific steps include:

[0054] S201 identifies the source of log files in the medical information system, wherein the source of log files includes operating system logs, network device logs, security device logs, and application logs;

[0055] Identify and locate system logs from servers, workstations and other operating systems (such as Windows Server, Linux, etc.). These logs record key information such as system operation status, error information, security events, etc. Network device logs include logs of switches and routers. These logs record the activities, configuration changes, errors and warning information of network devices, which are essential for tracking abnormal network behavior. Security device logs cover logs of security components such as firewalls, intrusion detection systems (IDS), and anti-virus software. These logs provide direct evidence of network intrusions and malicious activities. Application logs involve logs of various applications used in medical information systems (such as electronic medical record systems, image archiving and communication systems, etc.), which record detailed information when the application is running, which helps to discover security issues at the application layer.

[0056] S202 determines the source of network traffic data, where the source of network traffic data includes switches, routers, firewalls, and intrusion detection systems;

[0057] Switches monitor internal network communications, capture data packets, and analyze traffic patterns. Routers record traffic information in and out of the network, which helps to understand the overall network traffic status and external access patterns. Firewalls collect filter rule matching logs and identify blocked suspicious traffic, which is the first line of defense for network boundary security. Intrusion detection systems (IDS) are specifically designed to identify and record potential malicious network behavior and provide more in-depth traffic analysis, including attack attempts, abnormal traffic, etc.

[0058] S203 configures working parameters of the data collection tool according to the format and structure of the log and traffic data;

[0059] The working parameters include log path, filtering conditions and log format analysis.

[0060] Log path specifies the log storage location of each data source, and configures data collection tools to automatically capture. Filter conditions set logical rules to filter out security-related or important log information, reduce the collection of irrelevant data, and improve efficiency. Log format parsing configures parsing rules according to the structure and format of different logs (such as Syslog, CSV, JSON, etc.) to ensure that the collected data can be correctly understood and processed.

[0061] S204 transmits the collected network traffic data and system log data to the central storage system for storage.

[0062] The carefully screened and formatted network traffic data and system log data are sent to the central storage system through a secure transmission protocol (such as HTTPS, SFTP). The system must have high availability and security to ensure that the data is not tampered with or lost. In the central storage system, the data is organized according to the preset architecture to facilitate subsequent query, analysis and processing.

[0063] S102 cleans and standardizes network traffic data and system log data;

[0064] The specific steps include:

[0065] S301 identifies and processes abnormal values ​​in network traffic data and system log data and deletes them;

[0066] The specific steps include:

[0067] S401 deletes log entries or traffic records that are not related to network security event monitoring;

[0068] Based on predefined security-related keywords or event types (such as failed logins, permission changes, data leakage attempts, etc.), filter out log entries directly related to network security monitoring and exclude non-critical information such as daily operations and system maintenance.

[0069] Analyze network traffic data, remove background noise (such as normal heartbeat packets) and known harmless traffic, and focus on traffic patterns that may contain security threats.

[0070] S402 deletes duplicate records in the network traffic data and the system log data;

[0071] Use hashing technology or timestamp comparison to identify and delete identical or highly similar records in network traffic data, reducing the burden of data processing. By comparing log IDs, timestamps, and content, duplicate log entries are eliminated to ensure the uniqueness and accuracy of the data set.

[0072] S403 deletes the missing data.

[0073] For missing data in key fields, adopt appropriate strategies, such as using the average or median of historical data to fill in numerical data, or setting default values ​​based on business logic to fill in categorical data. For missing non-key fields, consider directly deleting them or using model prediction to fill in.

[0074] S302 converts the categorical data using one-hot encoding;

[0075] When processing network traffic data and categorical data in system logs, using One-Hot Encoding is a common and effective strategy. This step specifically includes the following aspects:

[0076] First, identify all categorical variables from log and traffic data, such as user role, access type, event code, etc. These variables usually have a finite and discrete set of values.

[0077] For each categorical variable, create a new feature column, assign 1 to each possible value of the variable, and assign 0 to the remaining columns. In this way, each category is converted into a binary vector, ensuring mutual exclusivity between categories and making it easier for the model to understand.

[0078] One-hot encoding may significantly increase the dimensionality of the feature space, especially when the categorical variables have many unique values. Therefore, in practical applications, it may be necessary to consider dimensionality reduction techniques such as principal component analysis (PCA), or encode only important categorical variables to balance computational cost and model performance.

[0079] S303 normalizes the numerical data using a normalization method.

[0080] Normalization of numerical data is an important step to ensure the stability and efficiency of the model training process. Its main purpose is to eliminate the dimension and scale differences between different features. Common normalization methods include minimum-maximum scaling, Z-score normalization, etc.

[0081] S103 trains a decision tree model based on historical network traffic data and historical system log data using access time, request type, and source IP reputation score features to obtain a monitoring model;

[0082] The specific steps include:

[0083] S501 marks each data sample as a security incident based on known security incidents in historical data;

[0084] The security incidents include intrusion attempts, data leaks and abnormal logins.

[0085] First, review historical data to identify confirmed security incidents, such as intrusion attempts (such as DDoS attacks, SQL injections), data leaks (illegal transmission of sensitive information), abnormal logins (unauthorized access attempts), etc.

[0086] Label each data sample to indicate whether it is associated with a known security incident. This can be done manually or automatically using preset rules and algorithms to ensure that there are clear target labels when training the model.

[0087] S502 extracts access time, request type, and source IP reputation score from the data sample as features;

[0088] Extract access time features from data samples and analyze the abnormality of activities within a specific time period, such as abnormal late-night access, differences in access patterns between weekdays and non-workdays, etc.

[0089] Identify different types of network requests (e.g., GET, POST, HEAD, etc.) and assess their role in security incidents. Certain types of requests may be more commonly associated with malicious activity.

[0090] Integrate external IP reputation databases or internally accumulated reputation scores to evaluate the trustworthiness of the source IP address that initiates the request. Low-reputation IPs may indicate potential attack behavior.

[0091] S503 divides the data samples into a training set, a validation set and a test set;

[0092] The sorted, labeled data set is randomly divided into training set, validation set and test set. The general ratio can be set to 70% for training, 15% for verifying model performance, and 15% for the final test of the generalization ability of the model.

[0093] Special attention is paid to maintaining a balanced ratio of positive and negative samples (i.e., safety events and non-safety events) when dividing the data set to avoid the model being biased towards the majority class and affecting the prediction accuracy.

[0094] S504 uses the training set to train the monitoring model using a decision tree algorithm.

[0095] Based on the CART decision tree algorithm, the first step is to select features to determine which features are most critical for predicting cybersecurity events. This is achieved through feature importance assessment, correlation analysis, or model-based feature selection methods. Then, according to the specific needs of the medical information system and the limitations of computing resources, a limit on the tree depth is set to prevent the model from overfitting.

[0096] When training the monitoring model, you also need to choose an appropriate pruning strategy, such as pre-pruning and post-pruning. Pre-pruning is to regularly evaluate the split gain of the node during the tree construction process. If the split gain is lower than a certain threshold, the split is stopped. Post-pruning is to check each non-leaf node from top to bottom after the decision tree is built. If the performance of the pruned model is not lower than the original model, the node and its child nodes are pruned.

[0097] Use the validation set to tune the model and evaluate the model performance through cross-validation and other techniques. Cross-validation is a statistical method that divides the data set into multiple subsets, selects one subset as the validation set each time, and the remaining subsets as the training set, repeating multiple times to evaluate the stability and performance of the model. Through cross-validation, the performance of the model on different data sets can be more accurately evaluated and the best model parameters can be selected.

[0098] Adjust the model parameters until the best performance indicators (such as accuracy, recall, F1 score, etc.) are achieved. These performance indicators reflect the prediction accuracy and recall of the model, as well as the balance between the two. By adjusting parameters such as tree depth limit and pruning strategy, the best model configuration can be found to achieve the highest performance indicators on the validation set.

[0099] Use the test set to evaluate the performance of the model on unseen data to ensure that the model not only performs well on the training data, but also has good generalization capabilities. The test set is a data set independent of the training set and validation set, which is used to evaluate the real performance of the model. Through test set evaluation, the performance of the model in actual applications can be verified and it can be ensured that it can accurately predict unseen network security events.

[0100] Through the above steps, an effective network security incident monitoring model can be trained based on the CART decision tree algorithm. This model can accurately predict network security incidents in practical applications and trigger corresponding security response measures in a timely manner to protect the security of medical information systems.

[0101] S104 automatically triggers corresponding security response measures based on the severity score of the event according to the monitoring model;

[0102] The specific steps include:

[0103] S601 sets different scoring thresholds according to operational requirements;

[0104] Based on historical event analysis, business impact assessment and security policy requirements, specific scoring thresholds are set for different levels of security events. For example, the minor event threshold, medium event threshold and major event threshold correspond to different levels of urgency and potential harm.

[0105] S602 pre-defines a security response measure group and sets a corresponding specific scoring range for each response measure, wherein the security response measure group includes blocking malicious IP, isolating infected terminals, sending alerts to security analysts, and recording event details;

[0106] According to the severity level of the incident, a series of security response measure groups are pre-defined, and each group of measures targets incidents within a different scoring range.

[0107] Block malicious IPs: Automatically implement network access control for access from known malicious or risky IP addresses to prevent further communication.

[0108] Isolate infected endpoints: Identify and isolate endpoints suspected of being infected or controlled by malware to prevent the spread of threats.

[0109] Send alerts to security analysts: For high-scoring events, the security operations center or designated security analysts are automatically notified for in-depth investigation and intervention.

[0110] Record incident details: Regardless of the response measures, the time, source IP, destination, severity score and actions taken should be recorded in detail for audit and subsequent analysis.

[0111] S603: When the severity score of the event exceeds a preset threshold, the decision logic module automatically triggers corresponding response measures;

[0112] Develop a decision logic module that monitors the severity score of events in real time and immediately triggers corresponding security response measures once the score exceeds a preset threshold. Rapid response is achieved through API calls or integration with security automation platforms.

[0113] S604 records the operation details of all automatically triggered response measures.

[0114] For each automatically triggered security response measure, the system should record in detail the time of response triggering, the triggered scoring threshold, the specific measures taken, the results of the measures, and any related system feedback. These records are crucial for evaluating the response effect, optimizing threshold settings, and subsequent event analysis.

[0115] S105 uses a visual management interface to display the processing results of network security events.

[0116] Design a central panel to display the overall status of the current system security, including the number of active security events, classification statistics (such as intrusion attempts, data leaks, abnormal logins, etc.), and the overall score or color-coded indication of the security situation (such as green for safe, yellow for warning, and red for danger). Then create a timeline view to display the occurrence of network security events, the triggering of response measures, and their results in chronological order. Each event is represented by a different icon or color. Click on any event to view its detailed information, including the triggered score, measures taken, response time, and processing status. Allow users to directly generate security incident reports through the interface, including event details, response process, impact assessment, and follow-up recommendations, and support users to filter and customize report content as needed.

[0117] Second embodiment

[0118] On the basis of the first embodiment, the present invention also provides an effective network security event monitoring system based on dynamic AI, including a data collection module, a standardization module, a monitoring model generation module, a processing module and a visualization module: the data collection module is used to collect and store network traffic data and system log data from the medical information system; the standardization module is used to clean and standardize the network traffic data and system log data; the monitoring model generation module is used to train a decision tree model based on historical network traffic data and historical system log data using access time, request type, and source IP reputation scoring features to obtain a monitoring model; the processing module is used to automatically trigger corresponding security response measures based on the severity score of the event according to the monitoring model; the visualization module is used to display the processing results of network security events using a visual management interface.

[0119] In this embodiment, the data collection module captures network traffic data and system log information from the medical information system in real time and continuously. These data include but are not limited to data packets entering and leaving the network, connection attempts, data transmission rates, and system operation records, etc., which are collected and stored in order to lay the foundation for subsequent analysis. The standardization module removes invalid, duplicate and abnormal data, and converts log information from different sources and formats into a unified standard to facilitate subsequent analysis and processing. This process includes timestamp standardization, field mapping, data type conversion, etc. to ensure the accuracy and comparability of the data. The monitoring model generation module deeply mines historical network traffic data and system log data, and the model selects key features such as access time mode, abnormal detection of request type, and source IP reputation score as the basis for judgment. These features comprehensively reflect the normal and abnormal modes of network activities. The model can learn potential threat patterns after training, so as to instantly identify suspicious behaviors in new data streams. Once the monitoring model finds a potential network security incident, the processing module immediately starts the automated response mechanism. Based on the severity score of the incident, the module automatically classifies the incident level and triggers the preset security response measures. This may include but is not limited to blocking malicious IPs, isolating infected terminals, sending alarms to notify security teams, and even launching emergency response plans to ensure that threats are contained in a timely manner and reduce potential losses. The visualization module intuitively displays the real-time processing progress and results of network security incidents through a highly interactive management interface. The interface not only shows the geographical location, time series, and type distribution of the incident, but also intuitively reflects the changing trend of the security situation and the effectiveness evaluation of response measures through charts. In summary, this dynamic AI-based network security incident monitoring system, through its comprehensive data processing links and intelligent response strategies, has built a solid security barrier for the medical information system and effectively improved the early warning and response capabilities to network security threats.

[0120] What is disclosed above is only a preferred embodiment of the present invention, and it certainly cannot be used to limit the scope of rights of the present invention. Ordinary technicians in this field can understand that all or part of the processes of the above embodiment and equivalent changes made according to the claims of the present invention still fall within the scope of the invention.

Claims

1. An effective network security incident monitoring method based on dynamic AI , It is characterized in that Includes: collecting and storing network traffic data and system log data from medical information systems; Clean and standardize network traffic data and system log data; Based on historical network traffic data and historical system log data, the decision tree model is trained using access time, request type, and source IP reputation score features to obtain a monitoring model. Automatically trigger corresponding security response measures based on the severity score of the event according to the monitoring model; A visual management interface is used to display the processing results of network security incidents.

2. An effective network security incident monitoring method based on dynamic AI as claimed in claim 1, It is characterized in that The specific steps of collecting network traffic data and system log data from the medical information system and storing them include: Identifying sources of log files in a medical information system, the sources of log files including operating system logs, network device logs, security device logs, and application logs; Determine the source of network traffic data, wherein the source of network traffic data includes switches, routers, firewalls and intrusion detection systems; Configure the working parameters of data collection tools according to the format and structure of log and traffic data; The collected network traffic data and system log data are transmitted to the central storage system for storage.

3. An effective network security incident monitoring method based on dynamic AI as claimed in claim 2, It is characterized in that The working parameters include log path, filtering conditions and log format analysis.

4. An effective network security incident monitoring method based on dynamic AI as claimed in claim 3, It is characterized in that The specific steps of cleaning and standardizing the network traffic data and system log data include: Identify, process and remove outliers in network traffic data and system log data; For categorical data, use one-hot encoding for conversion; The numerical data were standardized using the normalization method.

5. An effective network security incident monitoring method based on dynamic AI as claimed in claim 4, It is characterized in that The specific steps of identifying and processing outliers in network traffic data and system log data and deleting them include: Delete log entries or traffic records that are not relevant to network security event monitoring; Remove duplicate records from network traffic data and system log data; Delete the missing data.

6. An effective network security incident monitoring method based on dynamic AI as claimed in claim 5, It is characterized in that The specific steps of training the decision tree model based on historical network traffic data and historical system log data using access time, request type, and source IP reputation score features to obtain the monitoring model include: Based on known security incidents in historical data, mark each data sample as a security incident. Extract access time, request type, and source IP reputation score from data samples as features; Divide the data samples into training set, validation set and test set; The monitoring model is trained using the decision tree algorithm using the training set.

7. An effective network security incident monitoring method based on dynamic AI as claimed in claim 6, It is characterized in that The security incidents include intrusion attempts, data leaks and abnormal logins.

8. An effective network security incident monitoring method based on dynamic AI as claimed in claim 7, It is characterized in that The specific steps of automatically triggering corresponding security response measures based on the severity score of the event according to the monitoring model include: Set different scoring thresholds based on operational requirements; Predefine a group of security response measures and set a specific score range for each response measure, including blocking malicious IP, isolating infected terminals, sending alerts to security analysts, and recording event details; When the severity score of an event exceeds the preset threshold, the decision logic module automatically triggers the corresponding response measures; Records operational details of all automatically triggered response measures.

9. An effective network security incident monitoring system based on dynamic AI, It is characterized in that It includes data collection module, standardization module, monitoring model generation module, processing module and visualization module: The data collection module is used to collect network traffic data and system log data from the medical information system and store them; The standardization module is used to clean and standardize network traffic data and system log data; The monitoring model generation module is used to train a decision tree model based on historical network traffic data and historical system log data using access time, request type, and source IP reputation score features to obtain a monitoring model; The processing module is used to automatically trigger corresponding security response measures according to the severity score of the event by the monitoring model; The visualization module is used to display the processing results of network security events using a visualization management interface.