Power monitoring system distribution network security management active defense system

Through the power monitoring system that integrates security monitoring, intrusion detection, risk assessment and decision support units, the network security management active defense system is solved, and the shortcomings of traditional protection measures are solved in the face of complex attacks, and efficient and proactive defense of the power monitoring system is achieved.

CN120415882APending Publication Date: 2025-08-01STATE GRID SHANDONG ELECTRIC POWER CO
View PDF 0 Cites 7 Cited by

Patent Information

Application Number
CN202510777104.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-06-11
Publication Date
2025-08-01

AI Technical Summary

Technical Problem

Traditional cybersecurity protection measures are difficult to effectively prevent complex and hidden cyber attacks, especially advanced persistent threats, resulting in insufficient security of power monitoring systems.

Method used

An active defense system for network security management of power monitoring systems is designed, including a security monitoring unit, an intrusion detection unit, a risk assessment unit and a decision support response unit. Using network traffic and system log analysis, it optimizes the anti-perturbation direction through dynamic gradient perturbation factors and topological constraints, identifys abnormal behaviors, and provides coping strategies.

Benefits of technology

Real-time collection and analysis of network data is realized, the accuracy and efficiency of network threat detection is improved, and the ability to quickly identify and respond to unknown attacks is enhanced, and the active defense capabilities of the power monitoring system are enhanced.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN120415882A_ABST
    Figure CN120415882A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of network security management, in particular to an active defense system for power monitoring system distribution network security management. The safety monitoring unit is used for collecting and analyzing network data in real time; the intrusion detection unit is used for identifying suspicious activities and attack signs; the risk assessment unit further analyzes the suspicious activities and the attack signs and assesses the influence degree of the suspicious activities and the attack signs on the power grid security; and the decision support response unit provides coping strategy suggestions according to the result of the risk assessment and executes the provided countermeasures. By integrating the safety monitoring unit, the intrusion detection unit, the risk assessment unit and the decision support response unit, real-time collection, analysis and processing of network data are realized. Particularly, the intrusion detection unit adopts an advanced network flow behavior recognition model and a system log behavior recognition model, so that behaviors which are not consistent with a normal communication mode and abnormal operation records in a system log can be effectively recognized.
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 management, and particularly to an active defense system for the network security management of a power monitoring system's distribution network. Background Art

[0002] With the development of information technology, the power system increasingly relies on networked management and automated control. Although this improves efficiency and reliability, it also brings new security threats. As a core component of the power system, the power monitoring system is directly related to the safe and stable operation of the power grid. Therefore, how to effectively prevent network attacks and ensure the security of the power monitoring system has become an urgent problem to be solved. Traditional network security protection measures, such as firewalls and intrusion detection systems, mainly focus on passive defense, that is, responding after an attack occurs. However, in the face of increasingly complex and concealed network attack means, these measures often seem inadequate. Especially in the power monitoring system, attackers may use known vulnerabilities or newly emerging zero-day vulnerabilities to launch attacks, and traditional signature-based detection methods are difficult to detect and prevent such attacks in a timely manner. An advanced persistent threat is a form of network attack that lurks for a long time and is highly targeted. Attackers usually use a variety of technical means to gradually penetrate into the target network, collect information for a long time and carry out sabotage. This type of attack poses a particularly serious threat to the power monitoring system because they can bypass traditional security defenses and cause serious damage to the system. Therefore, an active defense system for the network security management of a power monitoring system's distribution network is designed. Summary of the Invention

[0003] The purpose of the present invention is to provide an active defense system for the network security management of a power monitoring system's distribution network to solve the limitations of traditional security protection measures and the problem of advanced persistent threats mentioned in the above background art.

[0004] To achieve the above object, the present invention aims to provide an active defense system for the network security management of a power monitoring system's distribution network, including: A security monitoring unit for collecting and analyzing network data in real time; An intrusion detection unit for identifying suspicious activities and attack signs, and introducing a dynamic gradient perturbation factor and designing the adversarial perturbation direction of topological constraints for optimization while detecting; A risk assessment unit for further analyzing the suspicious activities and attack signs and evaluating the degree of influence of the suspicious activities and attack signs on the security of the power grid; A decision support and response unit for providing countermeasure suggestions according to the results of the risk assessment and implementing the provided countermeasures.

[0005] As a further improvement of this technical solution, the network data includes network traffic and system logs.

[0006] As a further improvement of this technical solution, the intrusion detection unit includes a network traffic module and a system log module; Among them, the network traffic module is used to extract the feature vectors related to network traffic in the network data, analyze the network traffic pattern, find behaviors that do not conform to the normal communication pattern, and establish a network traffic behavior recognition model; The system log module is used to parse the system logs, extract the key feature vectors representing security events, and establish a system log behavior recognition model.

[0007] As a further improvement of this technical solution, establishing the network traffic behavior recognition model includes the following steps: S1.1. Construct a probability density function model linearly combined by K Gaussian distribution components; S1.2. Extract feature vectors and perform clustering analysis through historical network traffic data, and use the expectation-maximization algorithm to iteratively optimize the model parameters; S1.3. Determine the parameters of each Gaussian component, and the parameters include: mixing coefficient, mean vector, and covariance matrix; S1.4. Correct the parameters based on the maximum likelihood criterion to construct a network traffic behavior recognition model, and introduce a dynamic gradient perturbation factor and design a topological constraint to correct the parameters of the adversarial perturbation direction.

[0008] As a further improvement of this technical solution, establishing the system log behavior recognition model includes the following steps: S1.5. Establish an initial classification model based on Bayes' theorem, extract security event feature vectors by parsing system log data, calculate the prior probabilities of each security event category and the conditional probabilities of feature vectors under a given category, and form a basic posterior probability calculation model; S1.6. For the sensitive information features in the logs, construct a feature quantization processing layer, and by introducing a normalization conversion mechanism of feature vectors, encode the sensitive information into a standardized feature space to optimize the conditional probability calculation process; S1.7. Further integrate the topological characteristics of the power network, embed a topological awareness adjustment factor in the conditional probability calculation to form a topology-enhanced probability model, and introduce a device correlation correction function to perform spatial constraints on the probability density calculation.

[0009] As a further improvement of this technical solution, the risk assessment unit includes a risk analysis module and a risk assessment module; Among them, the risk analysis module classifies suspicious activities according to the nature and characteristics of the activities, determines the targets of the suspicious activities, and analyzes the specific means and techniques used by the attacker through intrusion detection methods; The risk assessment module classifies risks into different levels according to the severity and likelihood of the impact, calculates the risk value of each suspicious activity using a risk assessment model; and ranks the suspicious activities according to the risk level and impact degree to determine the activities to be processed first.

[0010] As a further improvement of this technical solution, analyzing the specific means and techniques used by the attacker through the intrusion detection method includes the following steps: S2.1. Identify the real-time network traffic data based on the constructed network traffic behavior recognition model, identify the traffic that does not conform to the conventional pattern, and through the traffic decision algorithm, dynamically learn the subtle features in the network traffic in real time. Parse the real-time system logs based on the system log behavior recognition model to find abnormal operation records; S2.2. Match the detected abnormal activities with the known attack feature library to identify the specific attack means. By analyzing the behavior pattern of the attacker, reconstruct the attack path, understand how the attacker uses the vulnerability to enter the system, and the attacker's target, and represent the attacker's behavior in the system as a graph structure, where the nodes represent different operation behaviors and the edges represent the sequence between the behaviors.

[0011] As a further improvement of this technical solution, in S2.1, dynamically learning the subtle features in the network traffic in real time through the traffic decision algorithm includes the following steps: S2.11. Establish a basic decision function to calculate the similarity between the current traffic feature vector and the historical support vector; S2.12. Generate an initial decision boundary by combining the weight coefficient and class label of the support vector; S2.13. To address the misjudgment problem caused by unknown attack patterns, introduce a dynamic adjustment mechanism into the decision function, and construct an enhanced decision function by superimposing adaptive parameters; S2.14. Continuously optimize the composition of the support vector set through an online learning mechanism, and use a feature clustering algorithm to identify new traffic patterns and generate candidate support vectors.

[0012] As a further improvement of this technical solution, the decision support response unit includes a decision support module and a decision response module; Among them, the decision support module generates specific countermeasure suggestions according to the risk assessment results based on the defense criteria; The decision response module executes corresponding countermeasures by configuring automated response measures according to the generated countermeasures to mitigate or eliminate the threat.

[0013] As a further improvement of this technical solution, the defense criterion based on the risk assessment result includes the following steps: S3.1. Define the rules for threat types under the risk level, and clarify the countermeasures to be taken in each case; S3.2. Build a rule engine for matching and executing corresponding countermeasures according to predefined rules; S3.3. Receive the risk assessment result from the risk assessment unit, and determine the potential threats and risk levels in the current network environment; S3.4. According to the risk assessment result, match the predefined rules and generate specific countermeasure suggestions.

[0014] Compared with the prior art, the beneficial effects of the present invention are: 1. In this active defense system for power monitoring system distribution network network security management, through the integration of a security monitoring unit, an intrusion detection unit, a risk assessment unit, and a decision support response unit, the real-time collection, analysis, and processing of network data are realized. In particular, the intrusion detection unit adopts advanced network traffic behavior recognition models and system log behavior recognition models, which can effectively identify behaviors that do not conform to the normal communication mode and abnormal operation records in the system logs. By continuously optimizing these models, the system can not only more accurately identify the real abnormal behaviors hidden in normal traffic but also reduce the false alarm rate, thereby improving the accuracy and efficiency of network threat detection.

[0015] 2. In this active defense system for power monitoring system distribution network network security management, through the decision support response unit, countermeasure suggestions are provided according to the result of risk assessment, and corresponding countermeasures can be automatically executed, such as automatically isolating infected hosts and automatically updating firewall rules. This intelligent threat detection and automated response mechanism enable the system to respond to potential threats in the shortest time, reducing or eliminating the impact of threats. In addition, by continuously optimizing the traffic decision algorithm, the system can better cope with unknown attacks, ensuring that even in the face of new or variant attacks, it can quickly identify and take countermeasures, greatly enhancing the active defense ability of the power monitoring system distribution network network security management. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is the overall flowchart of the present invention; Reference Signs: 1. Security monitoring unit; 2. Intrusion detection unit; 3. Risk assessment unit; 4. Decision support response unit. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0017] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0018] Please refer to Figure 1 as shown, this embodiment provides an active defense system for power monitoring system distribution network network security management, including: The security monitoring unit 1 is used to collect and analyze network data in real time; In this embodiment, the network data includes network traffic (information such as source address, destination address, port number, protocol type, etc. of incoming and outgoing data packets) and system logs (these logs record system operations, login attempts, error messages, etc., which are very important for detecting abnormal behaviors).

[0019] The intrusion detection unit 2 is used to identify suspicious activities and signs of attacks; In this embodiment, the intrusion detection unit 2 includes a network traffic module and a system log module; Among them, the network traffic module is used to extract the feature vectors related to network traffic in the network data, analyze the network traffic pattern, find behaviors that do not conform to the normal communication pattern, such as abnormal large amounts of data transmission, connection attempts to unknown ports, etc., and establish a network traffic behavior recognition model; Further, establishing a network traffic behavior recognition model includes the following steps: S1.1. Construct a probability density function model linearly combined by K Gaussian distribution components. The formula of the probability density function model is equal to the weighted sum of each Gaussian component, and each Gaussian component represents a potential network traffic behavior pattern; S1.2. Extract feature vectors and perform clustering analysis through historical network traffic data, and use the expectation maximization algorithm to iteratively optimize the model parameters. Among them, the formula of the expectation maximization algorithm (EM algorithm) involves two steps: the E step (expectation step) and the M step (maximization step). It estimates the parameters in the statistical model through iterative optimization. In the E step, the expected value of the latent variable is calculated, and the expectation of the logarithmic likelihood function of the data is completed based on the current parameter estimation; in the M step, the parameter value that maximizes the expected logarithmic likelihood function is searched, and these two steps are alternated until convergence; S1.3. Determine the parameters of each Gaussian component. The parameters include: mixing coefficient, mean vector, and covariance matrix, so that the overall probability distribution of the model fits the normal network traffic characteristics to the greatest extent; S1.4. Based on the maximum likelihood criterion, the model parameters or their application methods are corrected and enhanced by combining the following adaptive mechanisms to construct a network traffic behavior recognition model (the core of the maximum likelihood criterion is to find the parameter values that maximize the likelihood function of the observed data), enabling the model to adaptively learn the characteristics of normal traffic patterns in different network communication scenarios, and introducing a dynamic gradient perturbation factor and designing a perturbation influence correction parameter for topological constraints; by using the real-time operating state parameters of nodes (communication frequency, load rate, topological connectivity), a dynamic gradient perturbation factor is generated, enabling the network traffic behavior recognition model to automatically adjust the sensitivity to abnormal traffic according to the current operating state of the power grid. For example, the probability density threshold for determining abnormalities in the GMM model is dynamically adjusted by this factor, reducing false alarms by lowering the detection threshold during peak load periods and improving detection accuracy during low load periods to capture hidden attacks; traditional anomaly detection models are vulnerable to adversarial samples generated by attackers through gradient backpropagation. By introducing dynamic perturbations related to device states, the sensitivity change trend of the model to inputs is affected, destroying the predictability of the model response gradient by attackers and making it difficult for attackers to construct effective adversarial traffic; Extract the power network topology feature vector through the Graph Convolutional Network (GCN), and encode the physical connection relationships of substations, feeders, and terminal devices into the interpretation or subsequent judgment of the GMM model output. For example, if the traffic of a certain branch is abnormal but there is no associated change in its upstream and downstream nodes, it is inclined to be judged as a false alarm; if the abnormal traffic propagates along the topological path, it is identified as a potential attack chain. The sensitivity of anomaly determination is related to topological features, avoiding misjudgment of abnormal traffic that does not conform to the power communication path (cross-region direct access to protection devices) by the model. Attackers often use the dependencies between power devices to carry out cascading attacks (penetrating the master station through distribution automation terminals). The influence of topological constraints enables the model to capture cross-device and cross-layer collaborative abnormal behaviors, triggering a high-level alarm when abnormal logs are detected in multiple adjacent nodes simultaneously; The detailed steps for introducing a dynamic gradient perturbation factor and designing a perturbation influence correction parameter for topological constraints are as follows: Using the three-dimensional features of the communication frequency, load rate, and topological connectivity of nodes, a dynamic gradient perturbation factor is generated through an attention weight allocation mechanism: according to the real-time operating state parameter set of each node (including the three-dimensional features of the communication frequency, load rate, and topological connectivity of the node), the dynamic gradient perturbation factor is calculated using the sigmoid function and a trainable parameter matrix: , where As a regulatory factor, its magnitude reflects the adjustment requirement of the current node state for anomaly detection sensitivity; represents the Sigmoid activation function; is a trainable weight matrix, The feature vector representing the set of real-time operating state parameters of a single node is the bias term.

[0020] Then, the topological structure of the power grid is modeled as a graph structure, and the topological feature vector of this graph structure is extracted through a graph convolutional network. The influence direction of the topological constraint is designed, and a multi-layer perceptron (MLP) is used as the topological feature modulator to reflect the correlation between the interpretation or threshold adjustment of the output probability of the GMM model and the topological features; The dynamic gradient perturbation factor provides the dynamic adjustment ability based on the node operating state. Designing the influence direction of the topological constraint adds the constraint on the global topological structure on this basis, ensuring that the anomaly determination considers both the local state of the node and conforms to the overall characteristics of the power grid; the dynamic gradient perturbation factor adjusts the detection strategy according to the time series change of the device state, and the topological constraint correction parameter ensures that the anomaly determination conforms to the physical structure of the power grid. When a distributed denial of service (DDoS) attack occurs in a certain area, the dynamic perturbation factor automatically improves the detection sensitivity according to the attack traffic intensity, and at the same time the topological constraint module verifies whether the attack path conforms to the feeder partition logic to quickly locate the controlled terminal; Among them, the network traffic behavior recognition model is ; Among them, represents the probability density of the feature vector and is used to model the behavior of normal network traffic; represents the number of components in the Gaussian mixture model; represents the weight of the th Gaussian component; represents the mean vector of the

[0021] The system log module is used to parse the system log, extract the key feature vectors representing security events, such as failed login attempts, abnormal command executions, etc., and establish a system log behavior recognition model.

[0022] Furthermore, establishing a system log behavior recognition model includes the following steps: S1.5. Establish an initial classification model based on Bayes' theorem, extract security event feature vectors by parsing system log data, and calculate the prior probability of each security event category and the conditional probability of the feature vector given the category to form a basic posterior probability calculation model to achieve the preliminary classification of log events; S1.6. For the sensitive information features in the log, construct a feature quantization processing layer, and by introducing the feature vector The normalization conversion mechanism encodes sensitive information into a standardized feature space, optimizes the feature representation used in the subsequent conditional probability calculation process, and enhances the model's ability to analyze sensitive data features (a large amount of sensitive information is contained in the power monitoring system, such as user data, operation instructions, system configurations, etc. By optimizing the system log behavior model, log entries containing sensitive information can be more accurately identified and filtered to prevent these information from being illegally accessed or leaked; a large number of normal operation records may be contained in the system logs, and these records sometimes resemble attack behaviors, resulting in false alarms. By optimizing the model, normal operations and abnormal behaviors can be more accurately distinguished, reducing the false alarm rate; by quantifying sensitive information, it can be more precisely identified which log entries contain sensitive data, and corresponding protection measures can be taken. Quantifying sensitive information helps reduce false alarms and missed alarms caused by misjudgment and improves the accuracy of anomaly detection). S1.7. Further integrate the power network topology characteristics and embed a topology-aware adjustment factor in the conditional probability calculation This factor is generated by extracting structural features such as the current node's topological connectivity and the status of neighboring devices through a graph neural network to form a topology-enhanced probability model where the conditional probability is modeled using a multi-dimensional Gaussian distribution, and its covariance matrix is iteratively updated through a dynamic optimization algorithm guided by topological features, and a device correlation correction function is introduced to impose spatial constraints on the probability density calculation, enabling the model to perceive the topological dependence relationship between devices (the topological structure of the power monitoring system is complex, and the relationships between different nodes and devices are different. Considering the topological structure can more accurately identify abnormal behaviors on specific nodes, and according to the topological structure, monitoring resources can be reasonably allocated to focus on monitoring key nodes and important devices, improving the overall system security; combining the topological structure can better understand the context information of log records and improve the accuracy of anomaly detection).

[0023] ; wherein represents the sum of the numerator terms of all categories to ensure probability normalization; represents an index variable, indicating traversing all possible security event categories ; ; The specific expression of ; wherein represents a topological structure factor, reflecting the position and connection relationship of the node in the network Represents the node importance factor, reflecting the importance degree of the node. Represents the data sensitivity factor, reflecting the sensitivity degree of the log data. Is the topological structure weight factor. Is the node importance weight factor. Is the data sensitivity weight factor.

[0024] Topological structure factor Is to calculate a topological weight according to the position and connection relationship of the nodes in the network: , Represents the node To the central node Distance; Node importance factor Is to assign a weight according to the importance degree of the node: , Represents the node Importance score, which is comprehensively calculated by factors such as the function, traffic, and user access frequency of the node; Data sensitivity factor Is to assign a weight according to the sensitivity degree of the log data: , Represents the log data Sensitivity score, which is comprehensively calculated by factors such as keyword matching, data type, and data volume, Represents the maximum sensitivity score among all log data; Combining the above factors, we get Specific expression: ; Among them, Represents the marginal probability of the system log data ; The eigenvector Contains the comprehensive observation data of system logs and sensitive information; Represents the comprehensive observation data containing system logs and sensitive information Dimension; Represents the base of the natural logarithm; Represents different normal behavior patterns Mean value.

[0025] The risk assessment unit 3 further analyzes the suspicious activities and signs of attacks, and evaluates the impact degree of the suspicious activities and signs of attacks on the grid security; In this embodiment, the risk assessment unit 3 includes a risk analysis module and a risk assessment module; Among them, the risk analysis module classifies suspicious activities according to the nature and characteristics of the activities, such as scanning, injection, denial of service (DoS) attacks, etc., determines the targets of the suspicious activities, such as specific servers, network devices or applications, analyzes the specific means and techniques used by the attacker through intrusion detection methods, such as SQL injection, buffer overflow, etc.; analyzes the direct impact of the suspicious activities on the system and network, such as data leakage, system crash, service interruption, etc., and considers the indirect impacts that the suspicious activities may cause, such as business interruption, economic loss, reputation damage, etc.; Further, analyzing the specific means and techniques used by the attacker through intrusion detection methods includes the following steps: S2.1. Initially identify the real-time network traffic data based on the constructed network traffic behavior recognition model (i.e., the GMM model described in S1.1 - S1.4), identify the traffic that does not conform to the conventional pattern, and combine with the traffic decision algorithm to dynamically learn the subtle features in the network traffic in real time for more accurate identification of normal traffic and abnormal traffic (this traffic decision algorithm can be used as a supplementary enhancement to the GMM model or an independent parallel / serial decision module. This algorithm can adaptively adjust the recognition threshold for traffic features, not only paying attention to conventional features such as the magnitude and port of the traffic, but also being able to mine the semantic information hidden in the traffic data, such as special identifiers in specific data packet headers or potential association patterns between fields. These semantic information may imply new attack means or abnormal behaviors), such as extremely high data transmission volume, connection attempts to unknown ports, etc., parse the real-time system logs based on the system log behavior recognition model, and find out abnormal operation records, such as frequently failed login attempts, abnormal command executions, etc. (Compare the currently monitored network data with the normal behavior model to identify suspicious activities and attack signs. If it is found that the network traffic shows abnormal peaks or valleys (compared with the normal traffic model), for example, a large amount of traffic to a specific external network IP appears during non-business peak hours, it may be regarded as suspicious. For system logs, if there is an abnormal frequency of a certain operation (beyond the normal operation frequency model), such as frequent database query operations, it may be a sign of an attack); Further, through the traffic decision algorithm (the traffic decision algorithm can effectively detect network attacks by distinguishing normal data from abnormal data in network data and can refine the results of the initial GMM identification), dynamically learn the subtle features in the network traffic in real time, including the following steps: S2.11. Establish a basic decision function : , through the kernel function Calculate the similarity between the current traffic feature vector and the historical support vector . Indicates the total number of historical support vectors; S2.12. Combine the weight coefficients of the support vectors and the class labels to generate an initial decision boundary; S2.13. For the misjudgment problem caused by unknown attack patterns, introduce a dynamic adjustment mechanism in the decision function, and construct an enhanced decision function by superimposing adaptive parameters ; , (The power monitoring system involves the security of national critical infrastructure and faces increasingly complex and concealed attack means. Traditional detection methods based on known attack characteristics are difficult to cope with new or variant attacks. By optimizing the traffic decision algorithm, the detection ability of the system for unknown attacks can be improved, ensuring that even when encountering never-before-seen attack patterns, it can quickly identify and take countermeasures. The optimized algorithm can better combine with the decision support response unit to achieve intelligent threat detection and automated response. When an unknown attack is detected, the system can quickly generate countermeasures and execute them automatically, such as isolating infected hosts, updating firewall rules, etc., so as to mitigate or eliminate threats in the shortest time); S2.14. Continuously optimize the composition of the support vector set through an online learning mechanism , use the feature clustering algorithm to identify new traffic patterns and generate candidate support vectors. When the distribution entropy of the similarity in the continuous time window exceeds the threshold f, trigger the incremental update of the support vector set, so as to achieve progressive learning of unknown attack characteristics.

[0026] Among them, is used to judge whether the feature vector is abnormal; represents the bias term.

[0027] S2.2. Match the detected abnormal activities with the known attack feature library to identify specific attack means, such as SQL injection (specifically including: SQL injection attacks usually contain specific SQL keywords and syntax structures, such as SELECT, UPDATE, DROP; in HTTP requests, the URL parameters contain the above SQL keywords or patterns; the length of the input data is abnormally long and may contain complex SQL queries; requests containing SQL keywords are sent frequently in a short time), cross-site scripting (specifically including: containing <script>、所述风险评估模块根据影响的严重程度和可能性,将风险分为不同的等级,如低风险、中风险、高风险,使用风险评估模型(DREAD模型,DREAD模型通过五个维度来评估安全漏洞或可疑活动的风险值,每个维度都有一个评分范围(0到10),这五个维度分别是损害、可重复性、可利用性、受影响用户和可发现性)计算每个可疑活动的风险值;并根据风险等级和影响程度,对可疑活动进行优先级排序,确定首先处理的活动。

[0028] 决策支持响应单元4根据风险评估的结果,提供应对策略建议,并执行提供的对策;本实施例中,决策支持响应单元4包括决策支持模块和决策响应模块;其中,所述决策支持模块通过防御准则根据风险评估结果,生成具体的应对策略建议;进一步的,通过防御准则根据风险评估结果,包括以下步骤:S3.1、定义风险等级下的威胁类型的规则,明确每种情况下采取的应对措施,定义规则的结构,包括条件部分(if)和动作部分(then),如:规则一:如果检测到低风险的SQL注入攻击,记录日志并通知管理员,检查并修复相关输入验证逻辑;规则二:如果检测到高风险的远程代码执行攻击,立即隔离受影响的主机,更新相关软件或补丁,修复漏洞,通知安全团队进行全面检查;并集成外部威胁情报源,实时更新规则库;S3.2、构建规则引擎,用于根据预定义的规则匹配和执行相应的应对措施,加载所有预定义的规则,并配置规则引擎的输入和输出接口,确保规则引擎能够处理多个规则,并按优先级顺序执行,设置规则引擎的输入和输出接口,确保规则引擎能够动态更新规则,并支持多维度数据的综合评估;S3.3、从风险评估单元3接收风险评估结果,确定当前网络环境中的潜在威胁和风险等级,建立与风险评估单元的数据接口,确保实时接收风险评估结果,并收集多维度数据,进行全面的风险评估;S3.4、根据风险评估结果,匹配预定义的规则,生成具体的应对策略建议,将风险评估结果中的威胁类型和风险等级与规则引擎中的条件部分进行匹配,使用综合评估模型对不同维度的风险进行加权评估,如果有多个规则符合条件,选择优先级最高的规则,并根据最新的威胁情报和系统状态动态更新规则;所述决策响应模块根据生成的应对策略,通过配置自动化响应措施(如自动隔离受感染主机(使用网络隔离工具(如iptables、firewalld)配置防火墙规则,禁止受感染主机的网络访问)、自动更新防火墙规则(使用防火墙管理工具(如iptables、firewalld)动态更新防火墙规则,阻止特定IP地址或端口的访问)等)执行相应的对策,以减轻或消除威胁。

[0029] 以上显示和描述了本发明的基本原理、主要特征和本发明的优点。本行业的技术人员应该了解,本发明不受上述实施例的限制,上述实施例和说明书中描述的仅为本发明的优选例,并不用来限制本发明,在不脱离本发明精神和范围的前提下,本发明还会有各种变化和改进,这些变化和改进都落入要求保护的本发明范围内。< / script>

Claims

1. An active defense system for the security management of the distribution network in a power monitoring system, characterized in that, Including: A security monitoring unit (1) for collecting and analyzing network data in real time; An intrusion detection unit (2) for identifying suspicious activities and signs of attacks, and introducing a dynamic gradient perturbation factor and designing an adversarial perturbation direction with topological constraints for optimization during detection; A risk assessment unit (3) for further analyzing the suspicious activities and signs of attacks and evaluating the degree of impact of the suspicious activities and signs of attacks on the power grid security; A decision support response unit (4) for providing countermeasure suggestions based on the results of risk assessment and executing the provided countermeasures.

2. The active defense system for power distribution network security management of the power monitoring system according to claim 1, characterized in that The network data includes network traffic and system logs.

3. The power monitoring system distribution network network security management active defense system according to claim 2, characterized in that, The intrusion detection unit (2) includes a network traffic module and a system log module; Wherein, the network traffic module is used to extract feature vectors related to network traffic in the network data, analyze the network traffic pattern, find behaviors inconsistent with the normal communication pattern, and establish a network traffic behavior recognition model; The system log module is used to parse the system logs, extract key feature vectors representing security events, and establish a system log behavior recognition model.

4. The active defense system for power distribution network security management of the power monitoring system according to claim 3, characterized in that, The establishment of the network traffic behavior recognition model includes the following steps: S1.

1. Construct a probability density function model linearly combined by K Gaussian distribution components; S1.

2. Extract feature vectors and perform clustering analysis through historical network traffic data, and use the expectation maximization algorithm to iteratively optimize the model parameters; S1.

3. Determine the parameters of each Gaussian component, and the parameters include: mixing coefficient, mean vector and covariance matrix; S1.

4. Construct a network traffic behavior recognition model by correcting the parameters based on the maximum likelihood criterion, and introduce a dynamic gradient perturbation factor and design an adversarial perturbation direction with topological constraints to correct the parameters.

5. The active defense system for power distribution network security management of the power monitoring system according to claim 4, characterized in that, The establishment of the system log behavior recognition model includes the following steps: S1.

5. Establish an initial classification model based on Bayes' theorem, extract security event feature vectors by parsing system log data, calculate the prior probabilities of each security event category and the conditional probabilities of the feature vectors under a given category, and form a basic posterior probability calculation model; S1.

6. For the sensitive information features in the logs, construct a feature quantization processing layer, and by introducing a normalization conversion mechanism of the feature vectors, encode the sensitive information into a standardized feature space to optimize the conditional probability calculation process; S1.

7. Further integrate the power network topology characteristics, embed a topology-aware adjustment factor in the conditional probability calculation to form a topology-enhanced probability model, and introduce a device correlation correction function to perform spatial constraints on the probability density calculation.

6. The active defense system for power distribution network security management of the power monitoring system according to claim 5, characterized in that, The risk assessment unit (3) includes a risk analysis module and a risk assessment module; Wherein, the risk analysis module classifies the suspicious activities according to the nature and characteristics of the activities, determines the targets of the suspicious activities, and analyzes the specific means and technologies used by the attacker through intrusion detection methods; The risk assessment module classifies risks into different levels according to the severity and likelihood of impact, calculates the risk value of each suspicious activity using a risk assessment model, and prioritizes the suspicious activities according to the risk level and impact degree to determine the activities to be processed first.

7. The active defense system for power distribution network security management of the power monitoring system according to claim 6, characterized in that, The specific means and techniques used by the attacker are analyzed through the intrusion detection method, including the following steps: S2.

1. Identify the real-time network traffic data based on the constructed network traffic behavior recognition model, identify the traffic that does not conform to the conventional pattern, and dynamically learn the subtle features in the network traffic in real time through the traffic decision algorithm. Parse the real-time system logs based on the system log behavior recognition model to find abnormal operation records. S2.

2. Match the detected abnormal activities with the known attack feature library to identify the specific attack means. Reconstruct the attack path by analyzing the attacker's behavior pattern, understand how the attacker uses the vulnerability to enter the system and the attacker's target, and represent the attacker's behavior in the system as a graph structure, where the nodes represent different operation behaviors and the edges represent the sequence between the behaviors.

8. The active defense system for power distribution network security management of the power monitoring system according to claim 7, characterized in that, In S2.1, the subtle features in the network traffic are dynamically learned in real time through the traffic decision algorithm, including the following steps: S2.

11. Establish a basic decision function to calculate the similarity between the current traffic feature vector and the historical support vector. S2.

12. Generate an initial decision boundary by combining the weight coefficient and class label of the support vector. S2.

13. To address the misjudgment problem caused by unknown attack patterns, introduce a dynamic adjustment mechanism into the decision function and construct an enhanced decision function by superimposing adaptive parameters. S2.

14. Continuously optimize the composition of the support vector set through the online learning mechanism, and use the feature clustering algorithm to identify new traffic patterns and generate candidate support vectors.

9. The active defense system for power distribution network security management of the power monitoring system according to claim 8, characterized in that, The decision support response unit (4) includes a decision support module and a decision response module. Among them, the decision support module generates specific countermeasure suggestions according to the risk assessment result based on the defense criterion. The decision response module executes corresponding countermeasures by configuring automated response measures according to the generated countermeasures to mitigate or eliminate the threat.

10. The active defense system for power distribution network security management of the power monitoring system according to claim 9, characterized in that, The generation of specific countermeasure suggestions according to the risk assessment result based on the defense criterion includes the following steps: S3.

1. Define the rules for threat types under the risk level, and clarify the countermeasures to be taken in each case. S3.

2. Build a rule engine for matching and executing corresponding countermeasures according to the predefined rules. S3.

3. Receive the risk assessment result from the risk assessment unit (3) to determine the potential threats and risk levels in the current network environment. S3.

4. According to the risk assessment result, match the predefined rules to generate specific countermeasure suggestions.

Citation Information

Cited By

  • System network security evaluation system and method based on artificial intelligence

    CN120768661A

  • Cloud control system highly hidden attack detection method based on hierarchical state machine behavior modeling

    CN120880792A

  • A cloud control system high concealment attack detection method based on hierarchical state machine behavior modeling

    CN120880792B

  • Information security detection method and device for informatization equipment

    CN120915540A

  • Intrusion detection system of enterprise data information management platform

    CN121077812A