A server cluster anomaly detection method based on isolation forest
By adopting isolated forest architecture and dynamic permission management in the server cluster, the problems of insufficient isolation and poor dynamic adaptability in the multi-tenant environment are solved, and high security and flexibility are achieved, ensuring that the system can respond to new threats in a timely manner and meet different tenants. need.
Patent Information
- Application Number
- CN202411207512.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-30
- Publication Date
- 2024-11-26
- Estimated Expiration
- 2044-08-30
AI Technical Summary
The existing server cluster security solution has insufficient isolation, poor dynamic adaptability and insufficient performance optimization in a multi-tenant environment, making it difficult to effectively deal with new threats, especially in terms of permission management, resulting in excessive limitation of system resources or inability to respond in a timely manner. New attacks.
Using an isolated forest-based method, the server cluster is divided into multiple virtual areas. Each virtual area is an independent isolated forest, and multiple isolation levels are divided. Risk assessment and permission adjustment are carried out through a dynamic permission management system, combining security audits and logs. Record the system, dynamically adjust resource allocation, and create security monitoring models to identify threat patterns and abnormal activities.
Improve system security and flexibility, ensure that security policies can adapt to changing threat environments in a timely manner, achieve fine-grained isolation of different applications, services and tenants, ensure that permissions always match actual needs, and enhance system initiative and Adaptability to meet the personalized needs of different tenants.
Smart Images

Figure CN118740519B_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the fields of computer science and network technology, and in particular relates to a server cluster anomaly detection method based on isolation forest. Background Art
[0002] In the current server cluster environment, security has always been a topic of continuous concern. Although traditional security measures, such as firewalls, intrusion detection systems, and access control lists, can provide protection to a certain extent, they are difficult to cope with increasingly complex network attacks and evolving malware. Especially in a multi-tenant environment, the resource sharing of server clusters increases the attack surface of the system, making traditional security strategies more challenging. Therefore, a more advanced and comprehensive solution is urgently needed to improve the security of server clusters. At present, there have been some attempts to improve the security of server clusters, as follows:
[0003] First, containerized security solutions encapsulate different applications into independent containers to achieve a certain degree of isolation. However, these solutions may be insufficient in dynamic permission management and adaptability to multi-tenant environments, and lack dynamic assessment of the overall risk of the system.
[0004] Second, virtualization isolation solutions use virtualization technology to divide server resources and achieve isolation of multiple virtual environments. However, these solutions may not be flexible enough in terms of fine-grained isolation and dynamic permission management, and fail to effectively solve security challenges in multi-tenant environments.
[0005] Third, network security-based solutions attempt to protect server clusters through network security measures, such as intrusion detection systems and firewalls. However, these solutions may have performance bottlenecks in multi-tenant environments and may be slow to respond to new threats. Since they may introduce large performance overhead, they may affect the system's response speed and overall performance for server clusters that require high performance.
[0006] Although the above solutions have improved the security of server clusters to a certain extent, they have some shortcomings in terms of fine isolation, dynamic adaptability and performance optimization. Most existing solutions lack dynamism in permission management and often use static permission configuration. Such methods are difficult to cope with real-time changing network threats and may lead to excessive restrictions on system resources or inability to respond to new attacks in a timely manner.
[0007] Therefore, how to improve the existing server cluster anomaly detection method to enhance its adaptability, improve flexibility and the speed of response to new threats is a technical problem that needs to be solved urgently. Summary of the invention
[0008] The purpose of the present invention is to provide a server cluster anomaly detection method based on isolation forest, which improves the existing server cluster anomaly detection method to enhance its adaptability, improve flexibility and response speed to new threats.
[0009] In order to solve the above technical problems, the technical solution adopted by the present invention is as follows:
[0010] A server cluster anomaly detection method based on isolation forest includes the following steps:
[0011] S1: Create a dynamic isolation forest model: Use virtual machines or containerization to divide the server cluster into multiple virtual areas, each of which is an independent isolation forest;
[0012] S2: Divide into multiple isolation levels: In each isolation forest, divide into multiple isolation levels, each isolation level corresponds to different applications, services or tenants in the server cluster;
[0013] S3: Create a dynamic permission management system for risk assessment: Through real-time data sources and threat intelligence, the dynamic permission management system assesses the risks of the current network environment;
[0014] S4: Adjust the permissions of each isolation level based on the risk assessment results;
[0015] S5: Record various key events in the server cluster through security auditing and logging systems, and generate periodic reports;
[0016] S6: Dynamically adjust the allocation of computing resources, storage resources, and network resources according to the actual load requirements of different tenants in a multi-tenant environment, establish its internal network structure and communication rules, and isolate the data of each tenant in a multi-tenant environment;
[0017] S7: Create a security monitoring model to identify threat patterns and abnormal activities based on system logs, network traffic, and user behavior.
[0018] Preferably, step S1 includes the following specific process:
[0019] S11: Divide the server cluster into multiple virtual areas. Multiple virtual areas coexist on physical hardware and are isolated from each other.
[0020] S12: Establish an independent resource pool for each virtual area, wherein the resource pool includes computing resources, storage resources and network resources. The computing resources include CPU and memory, the storage resources include disk space, and the network resources include bandwidth and IP address.
[0021] Preferably, in step S2, the following specific processes are included:
[0022] S21: In each isolation forest, multiple isolation levels are divided to create network isolation between different isolation levels. Direct communication between isolation levels is restricted through virtual networks or network isolation.
[0023] S22: Create resource isolation within each isolation level to enable exclusive use of specified resources; create data isolation to isolate data in different isolation levels from each other, including databases and file systems; create process isolation to establish isolation between applications in different isolation levels;
[0024] Data isolation is achieved through data encryption and access control.
[0025] Preferably, step S3 includes the following specific process:
[0026] S31: Monitor network threats, user behavior and identity authentication information in real time, and conduct real-time risk assessment based on network threats, user behavior and identity authentication information;
[0027] S32: Automatically adjust the permissions of each isolation level based on the identity authentication results and risk assessment results, where the risk assessment results include high risk, medium risk, and low risk;
[0028] S33: Dynamically adjust the permissions of each isolation level based on changes in the network and user behavior.
[0029] Preferably, in step S32, when the risk assessment result of a certain isolation level is high risk, the authority of the isolation level will be reduced; when the risk assessment result of a certain isolation level is medium risk, the authority of the isolation level will remain unchanged; when the risk assessment result of a certain isolation level is low risk, the authority of the isolation level will be increased according to the specified requirements;
[0030] In step S33, the process of adjusting the permissions of each isolation level according to changes in the network and user behavior is to reduce the permissions of the corresponding isolation level when there is an increase in threats in the network or the risk of user behavior is greater.
[0031] Preferably, step S5 includes the following specific process:
[0032] S51: Record key events that occur in the server cluster, including permission changes, user logins, system configuration changes, and abnormal events;
[0033] S52: Add a unique identifier for each key event, where the identifier is a string;
[0034] S53: Analyze the permission change, user login, system configuration change, and abnormal events to determine whether there are risks or abnormal key events. If so, execute step S54;
[0035] S54: Tracing back based on the identifier of the key event with risk or abnormality, and obtaining the time of occurrence of the key event and related information;
[0036] Abnormal events include potential attacks and malware activity.
[0037] Preferably, in step S7, the security monitoring model is a convolutional neural network model, which uses the server's historical system logs, network traffic, user behavior and risk data as the security monitoring data basis, trains the security monitoring model with various historical data, and then uses real-time system logs, network traffic, and user behavior as input to output threat patterns and risk data of abnormal activities.
[0038] Preferably, an automated security policy update mechanism is also provided to collect threat intelligence sources in big data in real time and obtain the latest threat intelligence information. The threat intelligence includes known attack patterns, malicious IP addresses, and vulnerability information. The security policy is updated in real time based on the latest threat intelligence information. At the same time, after the security policy is updated, each change to the security policy is recorded and the historical version is maintained.
[0039] The beneficial effects of the present invention include:
[0040] The server cluster anomaly detection method based on isolation forest provided by the present invention creates a dynamic isolation forest model, divides the server cluster into multiple virtual areas, each virtual area is an independent isolation forest; in each isolation forest, divides multiple isolation levels; evaluates the risk of the current network environment; adjusts the authority of each isolation level based on the risk assessment result; records various key events in the server cluster through a security audit and log recording system, and generates periodic reports; dynamically adjusts resource allocation according to actual load requirements, establishes its internal network structure and communication rules, and isolates the data of each tenant in a multi-tenant environment; and creates a security monitoring model to identify threat patterns and abnormal activities.
[0041] First, the system security and flexibility are improved by isolating the forest architecture.
[0042] Second, real-time threat intelligence integration, security monitoring models and dynamic security policy generation ensure that security policies can adapt to the ever-changing threat environment in a timely manner, making the system proactive and adaptive.
[0043] Third, fine-grained isolation and multi-level isolation levels enable fine-grained isolation of different applications, services, and tenants, ensuring that each entity runs in a relatively independent environment.
[0044] Fourth, through real-time risk assessment, user authentication and dynamic adjustment of permissions, ensure that the permissions within each isolation level always match actual needs.
[0045] Fifth, by recording detailed information such as key events, permission changes, user activities, etc., the overall visibility of the system operation status is improved.
[0046] Sixth, through flexible isolation strategies, tenant-customized permission management, and flexible resource allocation measures, ensure that the system can meet the personalized needs of different tenants. BRIEF DESCRIPTION OF THE DRAWINGS
[0047] Figure 1 The figure is a flow chart of the server cluster anomaly detection method based on isolation forest of the present invention.
[0048] Figure 2 This is a schematic diagram of the isolation forest structure of the present invention.
[0049] Figure 3 It is a schematic diagram of the containerization principle of the present invention. DETAILED DESCRIPTION
[0050] The following is combined with Figure 1~Figure 3 The present invention is further described in detail:
[0051] See also Figure 1 , a server cluster anomaly detection method based on isolation forest, comprising the following steps:
[0052] S1: Create a dynamic isolation forest model: Use virtual machines or containerization to divide the server cluster into multiple virtual areas, each of which is an independent isolation forest;
[0053] S2: Divide into multiple isolation levels: In each isolation forest, divide into multiple isolation levels, each isolation level corresponds to different applications, services or tenants in the server cluster;
[0054] S3: Create a dynamic permission management system for risk assessment: Through real-time data sources and threat intelligence, the dynamic permission management system assesses the risks of the current network environment;
[0055] S4: Adjust the permissions of each isolation level based on the risk assessment results;
[0056] S5: Record various key events in the server cluster through security auditing and logging systems, and generate periodic reports;
[0057] S6: Dynamically adjust the allocation of computing resources, storage resources, and network resources according to the actual load requirements of different tenants in a multi-tenant environment, establish its internal network structure and communication rules, and isolate the data of each tenant in a multi-tenant environment;
[0058] S7: Create a security monitoring model to identify threat patterns and abnormal activities based on system logs, network traffic, and user behavior.
[0059] The isolation of existing server cluster security solutions is not sophisticated enough, which leads to interference between applications in a multi-tenant environment, increasing the risk of system attacks and poor flexibility. In terms of permission management, there is a lack of dynamism, and static permission configuration is often used, resulting in excessive restrictions on system resources or inability to respond to new attacks in a timely manner. If a network security solution is introduced, there may be a large performance overhead, which may affect the system's response speed and overall performance for server clusters that require high performance.
[0060] This application first divides the server cluster into multiple virtual areas, and divides multiple isolation levels downstream of each virtual area. Each isolation level corresponds to different applications, services or tenants in the server cluster, and establishes a more refined isolation forest level to avoid interference between applications in a multi-tenant environment and reduce the risk of system attacks. Based on the risk assessment results, the permissions of each isolation level are adjusted to minimize the system performance overhead, ensuring that the overall performance of the server cluster is not affected while improving security.
[0061] Through real-time data sources and threat intelligence, the dynamic permission management system assesses the risks of the current network environment, and dynamically adjusts the permissions of each isolated forest based on real-time risk assessment and identity authentication to better respond to changing network threats and reduce the burden on system performance. In a multi-tenant environment, according to the actual load requirements of different tenants, dynamically adjust the allocation of computing resources, storage resources, and network resources, establish its internal network structure and communication rules, and isolate the data of each tenant in a multi-tenant environment. Provide flexible solutions to meet the special needs of different tenants and improve the system's adaptability to multi-tenant environments.
[0062] The principle of containerization technology is shown in the attached Figure 3As shown in the figure, servers generally do not deploy only one service application, but mostly deploy multiple service applications. However, since these services are server resources such as CPU, memory, hard disk, and network IO in public servers, there will inevitably be resource contention and service impact. Therefore, containerization technology packages everything needed by the application together. The container includes all the code, virtualizing multiple environments or containers on a physical machine. By providing a way to create and enter containers, the operating system allows the application to run like an independent machine, but can share many underlying resources. While providing isolation, it also saves costs by sharing these resources.
[0063] In the above scheme, step S1 includes the following specific processes:
[0064] S11: Divide the server cluster into multiple virtual areas. Multiple virtual areas coexist on physical hardware and are isolated from each other, forming a secure containerized environment.
[0065] S12: Establish an independent resource pool for each virtual area, the resource pool includes computing resources, storage resources and network resources, the computing resources include CPU and memory, the storage resources include disk space, and the network resources include bandwidth and IP addresses, to ensure that the applications in each isolated forest can run in a controlled environment without being affected by other isolated forests.
[0066] In step S2, the following specific processes are included:
[0067] S21: In each isolation forest, multiple isolation levels are divided to create network isolation between different isolation levels. For the structure of the isolation forest, see Figure 2 As shown. By limiting direct communication between isolation levels through virtual networks or network isolation, applications running in the same forest can be more finely isolated, effectively preventing resource conflicts and lateral movement attacks;
[0068] S22: Create resource isolation within each isolation level to ensure exclusive use of designated resources, and ensure that each application or service can exclusively use certain computing resources, storage resources, and network resources. Create data isolation to isolate data within different isolation levels from each other, and ensure that data within each isolation level is isolated from each other. Data includes databases and file systems. Data isolation is achieved through data encryption and access control. Create process isolation to establish isolation between applications within different isolation levels, and ensure that the processes of each application or service run in an independent operating environment.
[0069] Step S3 includes the following specific processes:
[0070] S31: Real-time monitoring of network threats, user behavior and authentication information, and real-time risk assessment based on network threats, user behavior and authentication information. The core of dynamic permission management is user authentication. The system ensures accurate verification of the identity of each user to ensure that the permissions assigned to them are legal and appropriate. Authentication can be based on multi-factor authentication, including username and password, biometric features, hardware tokens, etc.
[0071] S32: Automatically adjust the permissions of each isolation level based on the identity authentication results and risk assessment results, where the risk assessment results include high risk, medium risk, and low risk;
[0072] S33: Dynamically adjust the permissions of each isolation level based on changes in the network and user behavior to ensure that the system can respond promptly to changing threats and user activities, thereby improving the security of the entire system.
[0073] In step S32, when the risk assessment result of a certain isolation level is high risk, the permissions of the isolation level will be reduced to reduce the potential attack surface; when the risk is low, the system may increase the permissions to ensure business smoothness. When the risk assessment result of a certain isolation level is medium risk, the permissions of the isolation level will remain unchanged; when the risk assessment result of a certain isolation level is low risk, the permissions of the isolation level will be increased according to the specified requirements;
[0074] In step S33, the process of adjusting the permissions of each isolation level according to changes in the network and user behavior is to reduce the permissions of the corresponding isolation level when there is an increase in threats in the network or the risk of user behavior is greater.
[0075] Step S5 includes the following specific processes:
[0076] S51: Record key events that occur in the server cluster, including permission changes, user logins, system configuration changes, and abnormal events, where abnormal events include potential attacks and malware activities;
[0077] S52: Add a unique identifier for each key event, where the identifier is a string;
[0078] S53: Analyze the permission change, user login, system configuration change, and abnormal events to determine whether there are risks or abnormal key events. If so, execute step S54;
[0079] S54: Based on the identifiers of key events with risks or anomalies, trace back to obtain the time of occurrence of the key events and related information. Abnormal event records provide early warnings of potential threats and help with timely response and investigation.
[0080] For each change in permission management, the system will generate a corresponding permission change record. The record includes information such as the permission status before and after the change, the reason for the change, and the identity of the user who performed the change. User activities in the server cluster will be recorded in detail, including login and logout time, resources accessed, operations performed, etc. For configuration changes of the server cluster, such as network configuration and system parameter modifications, corresponding records will be generated to ensure that system administrators can track changes in server cluster configuration and prevent unauthorized modifications.
[0081] In step S7, the security monitoring model is a convolutional neural network model, which uses the server's historical system logs, network traffic, user behavior and risk data as the basis for the security monitoring data, trains the security monitoring model through various historical data, and then uses real-time system logs, network traffic, and user behavior as input to output threat patterns and risk data of abnormal activities. The security monitoring model can extract features based on the server's historical system logs, network traffic, user behavior and risk data, can be trained to obtain the relationship between system logs, network traffic, user behavior and risks, and can identify security risks through real-time system logs, network traffic, and user behavior.
[0082] An automated security policy update mechanism is also set up to collect threat intelligence sources from big data in real time and obtain the latest threat intelligence information. The threat intelligence includes known attack patterns, malicious IP addresses, and vulnerability information. The latest threat intelligence information triggers real-time updates to security policies. At the same time, after the security policy is updated, each change to the security policy is recorded and historical versions are maintained.
[0083] In summary, the server cluster anomaly detection method based on isolation forest provided by the present invention creates a dynamic isolation forest model, divides the server cluster into multiple virtual areas, each virtual area is an independent isolation forest; in each isolation forest, divides multiple isolation levels; evaluates the risks of the current network environment; adjusts the permissions of each isolation level based on the risk assessment results; records various key events in the server cluster through a security audit and logging system, and generates periodic reports; dynamically adjusts resource allocation according to actual load requirements, establishes its internal network structure and communication rules, and isolates the data of each tenant in a multi-tenant environment; creates a security monitoring model to identify threat patterns and abnormal activities.
[0084] The isolation forest architecture improves system security and flexibility. Real-time threat intelligence integration, security monitoring models, and dynamic security policy generation ensure that security policies can adapt to the changing threat environment in a timely manner, making the system proactive and adaptive. Fine-grained isolation and multi-level isolation levels achieve fine-grained isolation of different applications, services, and tenants, ensuring that each entity runs in a relatively independent environment. Real-time risk assessment, user authentication, and dynamic adjustment of permissions ensure that permissions within each isolation level always match actual needs. By recording key events, permission changes, user activities, and other detailed information, the overall visibility of the system's operating status is improved. Flexible isolation strategies, tenant-customized permission management, and resource elasticity allocation measures ensure that the system can meet the personalized needs of different tenants.
Claims
1. A server cluster anomaly detection method based on isolation forest, characterized in that: The following steps are involved: S1: Create a dynamic isolation forest model: Use virtual machines or containerization to divide the server cluster into multiple virtual areas, each of which is an independent isolation forest; S2: Divide into multiple isolation levels: In each isolation forest, divide into multiple isolation levels, each isolation level corresponds to different applications, services or tenants in the server cluster; S3: Create a dynamic permission management system for risk assessment: Through real-time data sources and threat intelligence, the dynamic permission management system assesses the risks of the current network environment; S4: Adjust the permissions of each isolation level based on the risk assessment results; S5: Record various key events in the server cluster through security auditing and logging systems, and generate periodic reports; S6: Dynamically adjust the allocation of computing resources, storage resources, and network resources according to the actual load requirements of different tenants in a multi-tenant environment, establish its internal network structure and communication rules, and isolate the data of each tenant in a multi-tenant environment; S7: Create a security monitoring model to identify threat patterns and abnormal activities based on system logs, network traffic, and user behavior.
2. According to the method of claim 1, the method is characterized in that: Step S1 includes the following specific processes: S11: Divide the server cluster into multiple virtual areas. Multiple virtual areas coexist on physical hardware and are isolated from each other. S12: Establish an independent resource pool for each virtual area, wherein the resource pool includes computing resources, storage resources and network resources. The computing resources include CPU and memory, the storage resources include disk space, and the network resources include bandwidth and IP address.
3. The server cluster anomaly detection method based on isolation forest according to claim 1 is characterized in that: In step S2, the following specific processes are included: S21: In each isolation forest, multiple isolation levels are divided to create network isolation between different isolation levels. Direct communication between isolation levels is restricted through virtual networks or network isolation. S22: Create resource isolation within each isolation level to enable exclusive use of specified resources; create data isolation to isolate data in different isolation levels from each other, including databases and file systems; create process isolation to establish isolation between applications in different isolation levels; Data isolation is achieved through data encryption and access control.
4. The server cluster anomaly detection method based on isolation forest according to claim 1 is characterized in that: Step S3 includes the following specific processes: S31: Monitor network threats, user behavior and identity authentication information in real time, and conduct real-time risk assessment based on network threats, user behavior and identity authentication information; S32: Automatically adjust the permissions of each isolation level based on the identity authentication results and risk assessment results, where the risk assessment results include high risk, medium risk, and low risk; S33: Dynamically adjust the permissions of each isolation level based on changes in the network and user behavior.
5. The server cluster anomaly detection method based on isolation forest according to claim 4 is characterized in that: In step S32, when the risk assessment result of a certain isolation level is high risk, the permissions of the isolation level will be reduced; when the risk assessment result of a certain isolation level is medium risk, the permissions of the isolation level will remain unchanged; when the risk assessment result of a certain isolation level is low risk, the permissions of the isolation level will be increased according to the specified requirements; In step S33, the process of adjusting the permissions of each isolation level according to changes in the network and user behavior is to reduce the permissions of the corresponding isolation level when there is an increase in threats in the network or the risk of user behavior is greater.
6. The server cluster anomaly detection method based on isolation forest according to claim 1 is characterized in that: Step S5 includes the following specific processes: S51: Record key events that occur in the server cluster, including permission changes, user logins, system configuration changes, and abnormal events; S52: Add a unique identifier for each key event, where the identifier is a string; S53: Analyze the permission change, user login, system configuration change, and abnormal events to determine whether there are risks or abnormal key events. If so, execute step S54; S54: Tracing back based on the identifier of the key event with risk or abnormality, and obtaining the time of occurrence of the key event and related information; Abnormal events include potential attacks and malware activity.
7. The server cluster anomaly detection method based on isolation forest according to claim 1 is characterized in that: In step S7, the security monitoring model is a convolutional neural network model. The server's historical system logs, network traffic, user behavior and risk data are used as the security monitoring data basis. The security monitoring model is trained with various historical data, and then real-time system logs, network traffic, and user behavior are used as input to output threat patterns and risk data of abnormal activities.
8. The server cluster anomaly detection method based on isolation forest according to claim 1 is characterized in that: An automated security policy update mechanism is also set up to collect threat intelligence sources from big data in real time and obtain the latest threat intelligence information. The threat intelligence includes known attack patterns, malicious IP addresses, and vulnerability information. The latest threat intelligence information triggers real-time updates to security policies. At the same time, after the security policy is updated, each change to the security policy is recorded and historical versions are maintained.
