Infrastructure system security assessment method, device, electronic device and storage medium

Abstract

The infrastructure system security evaluation method and device, the electronic equipment and the storage medium provided by the embodiment of the application, the method comprises: acquiring an infrastructure undirected graph, the infrastructure undirected graph is generated by the logical connection relationship between a plurality of infrastructures and the infrastructures in an infrastructure system; generating a vulnerability curve corresponding to each node in the infrastructure undirected graph; obtaining a simulated failure probability of each node based on the vulnerability curve of each node; acquiring a node division constraint of the infrastructure system, generating a similarity matrix based on the simulated failure probability of each node, and clustering and dividing the infrastructure system based on the similarity matrix and the node division constraint to obtain a plurality of divided subsystems, each of which includes at most one vulnerable node; and performing security evaluation on each divided subsystem to obtain a security evaluation result of the infrastructure system, which greatly improves the accuracy and foresight of the cascade security evaluation result of the critical information infrastructure.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of system data security processing, and particularly relates to an infrastructure system security evaluation method and device, electronic equipment and a storage medium. BACKGROUND

[0002] In the current era of globalization and informationization, urban critical information infrastructure (such as power, transportation, communication, finance, etc.) has formed a highly interconnected and interdependent complex network system. These infrastructures are the "nerve center" of the stable operation of social economy, and the failure of any link can trigger a cascade failure through close physical connections or logical dependencies, thereby posing a serious threat to the regional network information security of the network system. Therefore, it is important to conduct forward-looking and systematic cascade security risk assessment on critical information infrastructure.

[0003] However, the existing critical information infrastructure security evaluation method is usually based on static topology and focuses on identifying vulnerable components in the topology, thereby performing security evaluation on each vulnerable component to obtain the security evaluation result of the critical information infrastructure. Since the actual infrastructure network is complex, the static topology is difficult to dynamically simulate and quantify the propagation process of failures in complex networks, and the security evaluation for a single vulnerable component lacks the perspective of the overall structural resilience of the system, thereby resulting in low accuracy of the security evaluation result obtained by the existing infrastructure security evaluation method. SUMMARY

[0004] The embodiments of the present application provide an infrastructure system security evaluation method, device, electronic equipment and storage medium, which can improve the accuracy of security evaluation of infrastructure systems.

[0005] To achieve the above-mentioned purpose, the first aspect of the embodiments of the present application provides an infrastructure system security evaluation method, which comprises:

[0006] Obtaining an infrastructure undirected graph generated by a plurality of infrastructures in an infrastructure system and logical connection relationships between the infrastructures;

[0007] Generating a vulnerability curve corresponding to each node in the infrastructure undirected graph;

[0008] Obtaining a simulated failure probability of each node based on the vulnerability curve of each node;

[0009] obtaining node partition constraints of the infrastructure system, generating a similarity matrix based on a simulated failure probability of each node, and performing cluster partitioning on the infrastructure system based on the similarity matrix and the node partition constraints to obtain a plurality of partitioned subsystems, each of the partitioned subsystems including at most one vulnerable node, and at least one of the partitioned subsystems including a plurality of nodes with the same function, the vulnerable node being at least one of the plurality of nodes;

[0010] performing security evaluation on each of the partitioned subsystems to obtain a security evaluation result of the infrastructure system.

[0011] In some embodiments, the generating of the vulnerability curve corresponding to each node in the infrastructure undirected graph comprises:

[0012] determining a failure influence factor of the node according to node attribute information and node connection relationship of each node;

[0013] obtaining a failure pressure parameter of the node based on the failure influence factor;

[0014] obtaining a vulnerability curve of the node based on the failure pressure parameter and a vulnerability function, the vulnerability curve being used to indicate a failure probability of the node under different failure pressures.

[0015] In some embodiments, the obtaining of the simulated failure probability of each node based on the vulnerability curve of each node comprises:

[0016] determining an initial failure node from the plurality of nodes, and performing multiple failure propagation simulations on the plurality of nodes in the infrastructure system based on the vulnerability curve of each node, the initial failure node, and a plurality of failure propagation scenarios to obtain a simulated failure probability of each node under each of the failure propagation scenarios;

[0017] In each failure propagation simulation, setting the initial failure node to a failure state, generating an initial propagation queue based on the initial failure node, determining a simulated failure probability of a neighbor node of each failure node in the initial propagation queue based on the simulated failure probability of the neighbor node of the failure node, updating the initial propagation queue based on the simulated failure probability of the neighbor node, and determining a simulated failure probability of a neighbor node of the failure node based on the updated initial propagation queue until the initial propagation queue is unchanged.

[0018] In some embodiments, the determining of the simulated failure probability of the neighbor node of the failure node and the updating of the initial propagation queue based on the simulated failure probability of the neighbor node comprise:

[0019] When the neighbor node of the faulty node is in a normal state, an invalidation influence value of the neighbor node is obtained, and an invalidation pressure of the neighbor node is determined based on the invalidation influence value;

[0020] Based on the invalidation pressure of the neighbor node and the fragility curve, an invalidation probability of the neighbor node is obtained, and an update state of the neighbor node is determined based on a Bernoulli trial and the invalidation probability;

[0021] When the update state of the neighbor node is an invalidation state, the neighbor node is added to the initial propagation queue;

[0022] When the neighbor node of the faulty node is in an invalidation state, the neighbor node is added to the initial propagation queue.

[0023] In some embodiments, the generating a similarity matrix based on the simulated failure probability of each node comprises:

[0024] Based on the square of the difference between the simulated failure probability of each two nodes, divided by the square of a control parameter, a node probability difference value is obtained;

[0025] Based on the negative value of the node probability difference value, and after exponential processing, a similarity between each two nodes is obtained;

[0026] Based on all the similarities, the similarity matrix is obtained.

[0027] In some embodiments, the clustering division of the infrastructure system based on the similarity matrix and the node division constraint obtains a plurality of division subsystems, comprising:

[0028] Based on the node division constraint, the similarity matrix is updated to obtain a constraint similarity matrix;

[0029] Based on the difference between the node degree matrix and the constraint similarity matrix, a graph Laplacian matrix is obtained;

[0030] A plurality of eigenvectors are obtained by solving the graph Laplacian matrix, and a predetermined number of eigenvectors are selected as an embedding matrix according to the ordering of a plurality of eigenvectors, the embedding matrix corresponding to an embedding space;

[0031] Using a clustering algorithm, clustering processing is performed in the embedding space to obtain a clustering identifier of each node;

[0032] Based on the clustering identifier, a plurality of nodes are divided to obtain a plurality of division subsystems.

[0033] In some embodiments, the node partition constraint comprises a same partition constraint and a fragile independent partition constraint, and the constrained updating of the similarity matrix based on the node partition constraint comprises:

[0034] updating, in the similarity matrix, a first target element indicated in the same partition constraint to a first numerical value;

[0035] updating, in the similarity matrix, a second target element indicated in the fragile independent partition constraint to a second numerical value;

[0036] obtaining the constrained similarity matrix based on the updated similarity matrix.

[0037] In some embodiments, the security assessment of each of the partition subsystems comprises:

[0038] obtaining node information of all nodes in each of the partition subsystems and node association information between the nodes, and determining a sub-security assessment result of each of the partition subsystems based on all the node information and the node association information;

[0039] obtaining system dependency relationships between the partition subsystems, and determining a propagation influence relationship between the partition subsystems based on the system dependency relationships;

[0040] performing a cascading security assessment of all the partition subsystems based on the propagation influence relationship and all the sub-security assessment results, to obtain the security assessment result of the infrastructure system.

[0041] To achieve the above object, a second aspect of the embodiments of the present application provides an infrastructure system security assessment device, which comprises:

[0042] an undirected graph obtaining module configured to obtain an infrastructure undirected graph generated by logical connection relationships between a plurality of infrastructures in an infrastructure system and the infrastructures;

[0043] a curve generating module configured to generate a vulnerability curve corresponding to each node in the infrastructure undirected graph;

[0044] a fault probability simulation module configured to obtain a simulated fault probability of each node based on the vulnerability curve of each node;

[0045] a subsystem partitioning module, configured to obtain node partitioning constraints of the infrastructure system, generate a similarity matrix based on the simulated failure probability of each node, and cluster the infrastructure system based on the similarity matrix and the node partitioning constraints to obtain a plurality of partitioned subsystems, each of which includes at most one vulnerable node, and at least one of which includes multiple nodes with the same function, wherein the vulnerable node is at least one of the multiple nodes;

[0046] The security assessment module is used to perform a security assessment on each of the divided subsystems to obtain a security assessment result of the infrastructure system.

[0047] To achieve the above-mentioned purpose, the third aspect of an embodiment of the present application proposes an electronic device, which includes a memory and a processor, the memory storing a computer program, and the processor implementing the infrastructure system security assessment method as described in the first aspect when executing the computer program.

[0048] To achieve the above-mentioned purpose, the fourth aspect of an embodiment of the present application proposes a storage medium, which is a computer-readable storage medium and stores a computer program. When the computer program is executed by a processor, it implements the infrastructure system security assessment method described in the first aspect above.

[0049] The infrastructure system safety evaluation method, device, electronic equipment and storage medium provided by the embodiment of the application, the method comprises the following steps: firstly, an infrastructure undirected graph is acquired, the infrastructure undirected graph is generated by the logical connection relationship between a plurality of infrastructures and the infrastructures in an infrastructure system; then, a vulnerability curve corresponding to each node in the infrastructure undirected graph is generated; next, based on the vulnerability curve of each node, a simulated failure probability of each node is obtained; thereafter, a node division constraint of the infrastructure system is acquired, a similarity matrix is generated based on the simulated failure probability of each node, and the infrastructure system is clustered and divided based on the similarity matrix and the node division constraint, so that a plurality of divided subsystems are obtained, each divided subsystem comprises at most one vulnerable node, and at least one divided subsystem comprises a plurality of nodes with the same function, and the vulnerable node is at least one of the plurality of nodes; finally, safety evaluation is performed on each divided subsystem, and a safety evaluation result of the infrastructure system is obtained. The embodiment of the application can dynamically quantify the real failure probability of the node in the complex network by constructing the vulnerability curve for each infrastructure node by using the infrastructure undirected graph and simulating the failure propagation based thereon, and overcomes the defect that the traditional static topology analysis cannot accurately reflect the cascading failure process. Then, by introducing the node division constraint, the identified vulnerable nodes are forcibly dispersed into different subsystems in the clustering and dividing stage, the "pre-isolation" and structural optimization of the system risk are realized, the potential high-risk propagation path is blocked, the structural resilience of the entire system is significantly enhanced, then the local evaluation of each divided subsystem after optimization is combined with the global dependence relationship for comprehensive evaluation, and a complete closed loop from dynamic simulation to structural optimization and then to hierarchical evaluation is formed, so that the accuracy and foresight of the cascading safety evaluation result of the key information infrastructure are greatly improved.

[0050] Other features and advantages of the present application will be set forth in the following description, and in part will be apparent from the description, or can be learned by practice of the present application. The objects and other advantages of the present application will be realized and attained by the structure particularly pointed out in the written description and claims hereof as well as the appended drawings. BRIEF DESCRIPTION OF DRAWINGS

[0051] Figure 1 is a flowchart of an infrastructure system safety evaluation method provided by an embodiment of the application.

[0052] Figure 2 is Figure 1 is a flowchart of step 102 in the embodiment.

[0053] Figure 3 is Figure 1 is a flowchart of step 103 in the embodiment.

[0054] Figure 4 isFigure 3 The flow chart in step 302.

[0055] Figure 5 Figure 1 The flow chart in step 104.

[0056] Figure 6 Figure 1 Another flow chart in step 104.

[0057] Figure 7 Figure 6 The flow chart in step 601.

[0058] Figure 8 Figure 1 The flow chart in step 105.

[0059] Figure 9 The flow chart of a method for evaluating the security of an infrastructure system according to another embodiment of the present application.

[0060] Figure 10 The structural diagram of an apparatus for evaluating the security of an infrastructure system according to another embodiment of the present application.

[0061] Figure 11 The hardware structural diagram of an electronic device according to another embodiment of the present application. DETAILED DESCRIPTION

[0062] In order to make the objects, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and should not be used to limit the present application.

[0063] It should be noted that although the functional modules are divided in the apparatus schematic diagram and the logical order is shown in the flow chart, in some cases, the steps shown or described can be executed in a manner different from the module division in the apparatus or the order in the flow chart.

[0064] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used in the description herein is for describing the embodiments of the present application only and is not intended to limit the present application.

[0065] ​​​​In the current era of globalization and informatization, the city's key information infrastructure (such as power, transportation, communication, finance, etc.) has formed a highly interconnected and interdependent complex network system. These infrastructures are the "nerve center" of the stable operation of social economy, and the failure of any one link can trigger a cascading failure through close physical connections or logical dependencies, thereby posing a serious threat to the regional network information security of the network system. Therefore, it is important to conduct a forward-looking and systematic cascading security risk assessment of the key information infrastructure.

[0066] However, the existing key information infrastructure security assessment method is usually based on static topology and focuses on identifying vulnerable components in the topology, thereby conducting security assessment on each vulnerable component to obtain the security assessment result of the key information infrastructure. Since the actual infrastructure network is complex, the static topology is difficult to dynamically simulate and quantify the propagation process of failures in complex networks, and the security assessment for a single vulnerable component lacks a perspective from the overall structural resilience of the system, resulting in a low accuracy of the security assessment result obtained by the existing infrastructure security assessment method.

[0067] In order to improve the accuracy of security assessment of infrastructure systems, the embodiments of the present application construct a vulnerability curve for each infrastructure node by using an infrastructure undirected graph, and simulate the propagation of failures based on this, which can dynamically quantify the real failure probability of nodes in complex networks, overcoming the defect that traditional static topology analysis cannot accurately reflect the cascading failure process. Then, by introducing node division constraints, the identified vulnerable nodes are forcibly dispersed into different subsystems in the clustering division stage, achieving "pre-isolation" and structural optimization of system risk to block potential high-risk propagation paths, significantly enhancing the structural resilience of the entire system. Then, by first assessing each divided subsystem after optimization locally, and then combining global dependencies for comprehensive assessment, a complete closed loop from dynamic simulation to structural optimization and then hierarchical assessment is formed, thereby greatly improving the accuracy and forward-looking of the cascading security assessment result of the key information infrastructure.

[0068] The infrastructure system security assessment method, device, electronic equipment and storage medium provided by the embodiments of the present application will be further described below. Referring to Figure 1 , an optional flowchart of the infrastructure system security assessment method provided by the embodiments of the present application, Figure 1 The method can include but is not limited to steps 101-105. It can be understood that the embodiments of the present application are not limited to Figure 1The sequence of steps 101 to 105 is not specifically limited, and the sequence of steps can be adjusted, or some steps can be reduced or added, according to actual needs. The infrastructure system security assessment method provided in the embodiments of the present application can be applied to any server or intelligent processing terminal, etc.

[0069] Step 101: Obtain an infrastructure undirected graph.

[0070] Step 101 will be described in detail below.

[0071] In some embodiments, an infrastructure undirected graph G(V, E) is first obtained, which is generated by the logical connection relationship between a plurality of infrastructures in the infrastructure system. Specifically, a complex infrastructure system in the real world is abstracted into a mathematical model that can be analyzed by calculation. As a functional unit of the infrastructure, it is a node V in graph theory, which can represent a power station, a communication base station, a transportation hub or a financial data center, etc. The logical connection relationship between the nodes V is abstracted as the edge E of the graph, which represents the physical coupling (such as cable connection) or functional dependence (such as the data output of a system is the premise of the operation of another system) between the nodes. By constructing the infrastructure undirected graph, a structured data foundation is laid for subsequent vulnerability analysis and fault propagation simulation.

[0072] Step 102: Generate a vulnerability curve corresponding to each node in the infrastructure undirected graph.

[0073] Step 102 will be described in detail below.

[0074] In some embodiments, after obtaining the infrastructure undirected graph G(V, E), a vulnerability curve corresponding to each node in the infrastructure undirected graph G(V, E) is then generated to quantify the inherent anti-interference ability of each independent node. The vulnerability curve is a function or data model that describes the probability of failure of each node when subjected to different degrees of external pressure (for example, the number or proportion of neighboring nodes connected to it that fail). The generation of the curve takes into account the properties of the node itself (such as the degree of device aging, security protection level) and its connection characteristics in the network, thereby transforming the stability of the node from a simple "normal / failure" binary state into a probability model that changes dynamically with the external environment, providing a core basis for subsequent dynamic simulation.

[0075] How to generate the vulnerability curve of each node will be described further below.

[0076] Referring to Figure 2 Generating a vulnerability curve corresponding to each node in the infrastructure undirected graph includes the following steps 201 to 203.

[0077] Step 201: Determine the failure influence factors of each node according to the node attribute information and node connection relationship of each node.

[0078] Step 202: Obtain the failure stress parameter of the node based on the failure influence factors.

[0079] Step 203: Obtain the vulnerability curve of the node based on the failure stress parameter and the vulnerability function.

[0080] The steps 201 to 203 are described in detail below.

[0081] In some embodiments, when generating the vulnerability curve of each node, first, the failure influence factors of each node are determined according to the node attribute information and node connection relationship of each node. This process builds the basis of the personalized vulnerability model, aiming to identify all internal and external factors that can cause the failure of a node in the infrastructure. The node attribute information is the static characteristics of the node, such as the type of the node (whether it is a power plant or a substation), the ratio of its current design load to actual operating load, and whether it has a redundant backup (i.e., redundancy). The node connection relationship is the dynamic interaction between nodes in the network topology, such as the failure state of its directly connected neighbor nodes, or more specifically, the proportion of failed nodes in the neighbor nodes. By comprehensively collecting and analyzing these failure influence factors, comprehensive and multi-dimensional data input can be provided for subsequent accurate quantification of the pressure-bearing capacity of the node.

[0082] Among them, the failure influence factor is a collection of key indicators for comprehensively quantifying the failure risk faced by a node in the infrastructure. It mainly covers two dimensions: one is the internal attributes of the node itself, such as the type of the node, the current load level, and whether it is equipped with a backup system redundancy, which together determine its basic pressure-bearing capacity; the second is the dynamic pressure from the external network, which is mainly reflected in the failure state of its directly connected neighbor nodes, or more specifically, the proportion of failed nodes in the neighbor nodes, which reflects the direct impact of cascading failures on it.

[0083] Then, based on the failure influence factors, the failure stress parameter of the node is obtained. At this time, the multiple and heterogeneous failure influence factors identified are integrated into a single and quantifiable indicator, i.e., the failure stress parameter S i , which represents the sum or weighted sum of all negative effects acting on the current node. For example, the degree of load overload of the node, the failure state of the key neighbor nodes, and other factors can be integrated through a pre-defined calculation formula (such as accumulation or weighted sum), thereby obtaining a numerical value. This numerical failure stress parameter S iThe risk level currently borne by the node can be intuitively reflected, and is a core independent variable for subsequent construction of a vulnerability curve, so that risks from different sources can be measured on a unified scale.

[0084] Then, based on the failure pressure parameter S i and a preset vulnerability function f i , a vulnerability curve of the node is obtained P i= f i (S i ), the vulnerability curve P i= f i (S i ) is used to indicate the failure probability of the node under different failure pressures. It can be understood that the vulnerability function defines how the failure probability of the node changes with the increase of the failure pressure parameter, and the function can be selected from a classic nonlinear function, for example, a Sigmoid function that can smoothly transition from 0 to 1, or an exponential function that grows rapidly. In a more complex scenario, it can even be fitted by expert experience, a large amount of historical failure data or special simulation data to ensure that it can most truly reflect the failure law of a specific type of node. The obtained failure pressure parameter S i is taken as input, and substituted into the vulnerability function f i , and the output is the corresponding failure probability P i . By calculating the failure probability under a series of different failure pressures, a complete vulnerability curve P i= f i (S i ) is finally drawn, which intuitively shows the pressure-bearing capacity and failure law of the node, and provides a basic unit model for dynamic safety assessment of the entire system.

[0085] Through implementation of the above steps 201 to 203, internal and external factors affecting the stability of the node are comprehensively identified, then quantified by the failure pressure parameter, and then an accurate mapping relationship between pressure and failure probability is established by means of a flexible and configurable vulnerability function, so that the traditional assessment of the fuzzy qualitative description of the node state is converted into an accurate, dynamic and calculable probability model, greatly improving the authenticity and accuracy of subsequent fault propagation simulation, so that the entire safety assessment system can be established on the basis of a solid and reliable micro model, thereby significantly improving the credibility and practical value of the final evaluation result.

[0086] Step 103: obtaining the simulated failure probability of each node based on the vulnerability curve of each node.

[0087] The following describes step 103 in detail.

[0088] In some embodiments, after obtaining the vulnerability curve of each node P i= f i (S i ) can be further based on the vulnerability curves of all nodes, and the cascading propagation process of failures in the entire network is derived through simulation, so as to obtain a node failure risk closer to the actual situation than static analysis, thereby obtaining the simulated failure probability of each node P i .

[0089] Specifically, the simulation process can use methods such as Monte Carlo simulation or graph simulation method, set one or more initial failure nodes, and use the vulnerability curve generated in the last step as the propagation rule to perform hundreds of failure scenario derivations. In each simulation, the failure of a node will exert pressure on its neighbor nodes, and the neighbor nodes will decide whether to also fail according to their own vulnerability curves. By counting the frequency of each node eventually failing in all simulation scenarios, a comprehensive simulated failure probability can be obtained, which not only reflects the vulnerability of the node itself, but more importantly, includes the cascading risk it bears in the entire system network.

[0090] The following will further specifically describe how to use Monte Carlo simulation to obtain the simulated failure probability of each node.

[0091] Referring to Figure 3 , obtaining the simulated failure probability of each node based on the vulnerability curve of each node includes the following steps 301 to step 302.

[0092] Step 301: determining an initial failure node from the plurality of nodes, and performing multiple failure propagation simulations on the plurality of nodes in the infrastructure system based on the vulnerability curve of each node, the initial failure node, and a plurality of failure propagation scenarios to obtain a simulated failure probability of each node under each failure propagation scenario.

[0093] Step 302: In each fault propagation simulation, set the initial fault node to the failure state, and generate an initial propagation queue based on the initial fault node, for each fault node in the initial propagation queue, determine the simulation failure probability of the neighbor node of the fault node, and update the initial propagation queue based on the simulation failure probability of the neighbor node, and determine the simulation failure probability of the neighbor node of the fault node based on the updated initial propagation queue until the initial propagation queue is unchanged.

[0094] Steps 301 to 302 are described in detail below.

[0095] In some embodiments, when Monte Carlo technology is used for fault scenario simulation, the cascading failure risk that the infrastructure network may face is systematically explored through large-scale simulation. First, one or a set of initial fault nodes F0 needs to be set, which can be determined based on historical data, intelligence information or assumed extreme attacks, as the starting point of fault propagation. Then, multiple fault propagation simulations are performed under the framework of multiple fault propagation scenarios, that is, the simulation process is repeated hundreds or thousands of times to cover enough random possibilities to ensure the statistical significance of the results. In each simulation, the propagation rule of the fault completely depends on the vulnerability curve of each node that has been determined P i= f i (S i ). Through this process, the frequency of failure of each node under various possible evolution paths can be statistically obtained, and a comprehensive "simulation failure probability" reflecting the dynamic risk of each node in the entire network can be obtained.

[0096] Specifically, in each fault propagation simulation (simulation times t = 1, 2, …, T), first set all nodes to the "normal state", then set the initial fault node F0 to the "failure state", and generate an initial propagation queue Q based on the initial fault node; Next, for each fault node in the initial propagation queue, traverse all neighbor nodes v of the fault node to determine the simulation failure probability of the neighbor node v of the fault node P v , and update the initial propagation queue Q based on the simulation failure probability of the neighbor node P v , and determine the simulation failure probability of the neighbor node v of the fault node based on the updated initial propagation queue Q P v , until the initial propagation queue is unchanged.

[0097] This process describes the specific execution logic of single fault propagation simulation, which is an iterative algorithm process. When the simulation starts, the selected "initial fault node F0" is marked as "failure" and put into a data structure named "initial propagation queue Q", which is used to store all known fault nodes whose influence has not yet been fully propagated. Then, the algorithm enters a loop: take a fault node from the queue, examine all its neighbor nodes v that have not yet failed. For each neighbor node v, calculate its failure pressure S v based on the failure conditions around it (including the node that has just failed), and determine its simulated failure probability based on its own vulnerability curve P v and make a random decision. If the neighbor node v fails as a result of this decision, it is added to the propagation queue Q. This process continuously updates the initial propagation queue and continues until there are no new nodes in the queue that can be affected by the propagation, i.e. until the initial propagation queue is unchanged, indicating that the propagation process of the cascade failure in this simulation has stabilized. The update of the initial propagation queue can be described as follows.

[0098] Referring to Figure 4 , determine the simulated failure probability of the neighbor nodes of the fault node, and update the initial propagation queue based on the simulated failure probability of the neighbor nodes, including steps 401 to 404.

[0099] Step 401: When the neighbor node of the fault node is in a normal state, obtain the failure influence value of the neighbor node, and determine the failure pressure of the neighbor node based on the failure influence value.

[0100] Step 402: Based on the failure pressure of the neighbor node and the vulnerability curve, obtain the failure probability of the neighbor node, and based on the Bernoulli trial and the failure probability, determine the update state of the neighbor node.

[0101] Step 403: When the update state of the neighbor node is a failure state, add the neighbor node to the initial propagation queue.

[0102] Step 404: When the neighbor node of the fault node is in a failure state, add the neighbor node to the initial propagation queue.

[0103] The steps 401 to 404 are described in detail as follows.

[0104] In some embodiments, the system examines each faulty node i in the initial propagation queue one by one, and for each faulty node i, the system examines all of its neighbor nodes v in the “normal state”. When a neighbor node v of the faulty node i is in the normal state, the system determines the failure impact value corresponding to the failure impact factor of the current neighbor node v, and substitutes the failure impact value into the vulnerability function to determine the failure pressure S of the neighbor node v .

[0105] The failure pressure S obtained is then inputted into the vulnerability curve of the neighbor node itself v , so as to query or calculate the specific failure probability of the neighbor node under the current pressure P v= f v (S v ). P v However, having the failure probability P v does not mean that the node will necessarily fail. Therefore, the present embodiment introduces a Bernoulli trial, which is a single random event with only two outcomes: “success” or “failure”. In this scenario, the system performs a “Bernoulli trial” with the “failure probability” as the success probability, which can be understood as a weighted electronic “coin toss”. The result of the trial will directly “determine the updated state of the neighbor node”, i.e., decide whether it will change to the “failed state” or remain in the “normal state” in the current simulation round.

[0106] Then, when the updated state of the neighbor node v is the failed state, the neighbor node v is added to the initial propagation queue Q. This step is a key operation to realize the spread of the cascading effect in the network. If the result of the “Bernoulli trial” causes the “updated state” of the neighbor node c to be determined and updated to the “failed state”, this originally healthy node now becomes a new source of failure. In order to enable this newly generated failure to continue to have an impact on its own neighbor nodes, the system will “add the neighbor node to the initial propagation queue”. Through this operation, the node will be treated as a “faulty node” in the subsequent simulation loop, thereby advancing the chain reaction of failure one step further.

[0107] Further, when the neighbor node v of the faulty node is originally in the failed state, the neighbor node is added to the initial propagation queue. This step is mainly used as a logical supplement and completeness check to handle the situation where a "neighbor node" of a "faulty node" is found to be in a "failed state" when checking the neighbors of a "faulty node". This situation can occur because the neighbor node has been affected by the failure of other paths in the previous simulation cycle. In order to ensure the integrity of the entire failure propagation process and not to miss any potential propagation source, this step provides that as long as a node is confirmed to be in a "failed state", regardless of how it failed, it should be "added to the initial propagation queue" to ensure that its impact on other neighbors that have not failed can be fully calculated and propagated.

[0108] Through the close cooperation of the above steps 401 to 404, starting from calculating the stress on the healthy node, the uncertainty of node failure is truly simulated through the probabilistic judgment of combining the vulnerability curve and Bernoulli trial, and the iterative diffusion of cascading failure in the network is realized through the way of dynamically updating the propagation queue, so that the macro failure simulation process is divided into a series of rigorous and repeatable micro operations, ensuring that each simulation not only follows physical and logical laws, but also contains randomness in the real world, so that the final simulation failure probability can highly realistically reflect the dynamic risk of the system, providing a solid algorithmic guarantee for the accuracy of the evaluation.

[0109] In one example, the failure propagation scenario is generated by a Monte Carlo method or a graph simulation method, and the number of simulations is not less than a preset threshold T. For each simulation t = 1, 2,..., T, the specific process includes the following steps.

[0110] Step 1: initialize the state, set all node states to "normal", mark the initial failure node F0 as "failed", and initialize the propagation queue Q.

[0111] Step 2: propagation loop, for each node in the current propagation queue, traverse its neighbor nodes v, if the neighbor node has not failed, calculate its current failure stress S v , then substitute it into the vulnerability function f to get the failure probability, perform a Bernoulli trial according to the failure probability, if v is new, add it to the propagation queue Q.

[0112] Step 3: record the propagation result, record the set of all nodes that failed in this round Ft t , for all nodes, if in the set Ft, its failure count +1.

[0113] Step 4: repeat T rounds of simulation, count the failure frequency P i of all nodes iThe number of failures in all scenarios / T.

[0114] Through the implementation of steps 301 to 302, all potential failure propagation paths in the network are systematically explored by setting the initial failure and performing large-scale repeated simulation, the vulnerability curve previously constructed is used as the dynamic propagation rule, and the cascading reaction of failure from one node to another is accurately simulated by iteratively updating the propagation queue, so that the security assessment is no longer limited to static topology analysis, but can dynamically and quantitatively reveal the hidden cascading risk in the network, accurately identify the key nodes that are not only vulnerable themselves but also susceptible to the influence of neighbors and fail, and the final simulation failure probability of each node provides a highly reliable and insightful data basis for subsequent risk isolation and systematic security assessment.

[0115] Step 104: Obtain node partition constraints of the infrastructure system, generate a similarity matrix based on the simulation failure probability of each node, and perform cluster partitioning of the infrastructure system based on the similarity matrix and the node partition constraints to obtain a plurality of partitioned subsystems.

[0116] Step 104 is described in detail below.

[0117] In some embodiments, after obtaining the simulation failure probability of each node, the node partition constraints of the infrastructure system are then obtained, a similarity matrix is generated based on the simulation failure probability of each node, and the infrastructure system is cluster partitioned based on the similarity matrix and the node partition constraints to obtain a plurality of partitioned subsystems. This process is a key link to achieve active risk isolation and system structure optimization. First, according to the simulation failure probability obtained in the previous step, the similarity between nodes is calculated, and nodes with similar failure probabilities are considered to have similar risk characteristics, thereby constructing a similarity matrix W. At the same time, the core "node partition constraint" is introduced, which is a set of pre-set rules that clearly specify that "at most one vulnerable node is included in each partitioned subsystem" (i.e., the vulnerable node must be divided into different areas to achieve isolation), and "there is at least one partitioned subsystem that includes multiple nodes with the same function" (i.e., functionally cooperative units should be divided in the same area to ensure integrity). Finally, a constraint spectral clustering algorithm that can integrate data and rules is used to partition the entire infrastructure undirected graph based on the similarity matrix E and the node partition constraints, and finally form a plurality of independent "partitioned subsystems".

[0118] The node partition constraint includes a same partition constraint (Must-Link constraint) and a fragile independent partition constraint (Cannot-Link constraint). The Must-Link constraint is used to specify that the functional units corresponding to a certain node must be partitioned into the same area, and the Cannot-Link constraint is used to specify that a key fragile node cannot be partitioned into the same area. The fragile node can be a node whose simulated failure probability is higher than a certain failure threshold after the simulated failure probability of each node is obtained above, or a certain node that is prone to failure.

[0119] How to generate the similarity matrix will be further described below.

[0120] Referring to Figure 5 , the similarity matrix is generated based on the simulated failure probability of each node, including the following steps 501 to 503.

[0121] Step 501: Based on the square of the difference between the simulated failure probabilities of each two nodes, divided by the square of the control parameter, the node probability difference is obtained.

[0122] Step 502: Based on the negative value of the node probability difference, exponential processing is performed to obtain the similarity between each two nodes.

[0123] Step 503: Based on all similarities, the similarity matrix is obtained.

[0124] The steps 501 to 503 will be described in detail below.

[0125] In some embodiments, when generating the similarity matrix W, first, based on the square of the difference between the simulated failure probabilities of each two nodes (such as node i and node j), divided by the square of the control parameter a, the node probability difference (P i -P j ) 2 / 2a 2 This step is the starting calculation link for quantifying the risk similarity between any two infrastructure nodes. The system will traverse all node pairs in the network. For each pair of nodes, the "failure probability P" obtained in the previous stage through failure propagation simulation will be extracted. Then, the difference between the two "failure probability" values is calculated, and the square of the difference is processed. The advantage of this is that the result is positive regardless of whether the difference is positive or negative, only reflecting the size of the gap. Subsequently, the square difference is "divided by the square of the control parameter", which is a key technical feature. It plays a role in mathematics to adjust the scale, which is used to control the sensitivity of the failure probability difference to the final similarity. Through this series of operations, a standardized intermediate value, the "node probability difference", is finally obtained, which directly reflects the difference between the two nodes in the dynamic cascade risk.

[0126] where the control parameter a can be set as all P i The standard deviation, the more similar the nodes are, the higher the weight is.

[0127] Then, based on the negative value of the node probability difference, an exponential processing is performed to obtain the similarity between each two nodes. This step aims to convert the difference measure obtained in the previous step into an intuitive similarity measure, i.e., the similarity w ij = exp(-(P i -P j ) 2 / 2a 2 ). First, the "negative value" of the "node probability difference" is taken, and the purpose of this step is to obtain an intuitive result after subsequent exponential processing: the smaller the difference, the higher the similarity. Subsequently, the negative value is "exponentially processed", which usually refers to the exponential operation with the natural constant e, i.e., the Gaussian kernel function (RBF) is applied. This similarity value is between 0 and 1: when the failure probabilities of two nodes are exactly the same, the node probability difference is 0, and the similarity is e to the power of 0, i.e., 1 (completely similar); when the difference in failure probability is huge, the node probability difference is large, and its negative value is a large negative number, and the similarity tends to 0 (completely dissimilar).

[0128] After the similarity w ij is calculated between all possible pairs of nodes in the network, the system organizes these values into a two-dimensional square matrix, i.e., a similarity matrix W. The rows and columns of the matrix represent all nodes in the infrastructure system, and the element value in the ith row and jth column of the matrix is equal to the "similarity" calculated between node i and node j. This "similarity matrix" is a symmetric matrix (because the similarity between node i and j is equal to the similarity between node j and i), which completely and structurally stores the mutual relationships between all nodes in the network based on dynamic cascading risk, providing a core data input for subsequent clustering algorithm.

[0129] Through the rigorous calculation of steps 501 to 503 above, by applying the Gaussian kernel function, the difference in failure probability of the nodes is skillfully mapped to a standardized similarity space, so that nodes with similar risk characteristics obtain high similarity, and vice versa, thus providing a quantitative and intuitive basis for the subsequent clustering algorithm, enabling the algorithm to clearly determine which nodes exhibit similar behavior patterns in the face of cascading failures, and further providing a key data foundation for realizing risk-based intelligent partitioning and actively isolating high-risk node groups, and is the core bridge connecting dynamic risk assessment and active structure optimization.

[0130] How to perform clustering partition on the infrastructure system based on the similarity matrix and the node partition constraints will be further described below.

[0131] With reference to Figure 6 , the infrastructure system is partitioned based on the similarity matrix and the node partition constraints to obtain a plurality of partition subsystems, including the following steps 601 to 605.

[0132] Step 601: constraint updating the similarity matrix based on the node partition constraints to obtain a constraint similarity matrix.

[0133] The step 601 will be described in detail below.

[0134] In some embodiments, when the infrastructure system is partitioned, the similarity matrix W is first constraint updated based on the node partition constraints (including the same partition constraint and the fragile independent partition constraint) to obtain a constraint similarity matrix. This process is the core of achieving active risk isolation and functional aggregation, which forcibly injects the preset expert knowledge or operation rules into the data-driven clustering process. In this process, it is defined based on the pre-defined node partition constraints which nodes must be partitioned in the same area (Must-Link) and which key fragile nodes must not be partitioned in the same area (Cannot-Link). Then, the system will modify the similarity matrix W using the node partition constraints to obtain the updated constraint similarity matrix, which is described as follows.

[0135] With reference to Figure 7 , the similarity matrix is constraint updated based on the node partition constraints to obtain a constraint similarity matrix, including the following steps 701 to 703.

[0136] Step 701: in the similarity matrix, the first target element indicated in the same partition constraint is updated to a first numerical value.

[0137] Step 702: in the similarity matrix, the second target element indicated in the fragile independent partition constraint is updated to a second numerical value.

[0138] Step 703: based on the updated similarity matrix, a constraint similarity matrix is obtained.

[0139] The steps 701 to 703 will be described in detail below.

[0140] In some embodiments, when modifying the similarity matrix W, first update the first target elements indicated by the same partition constraint to a first numerical value in the similarity matrix. In this process, the same partition constraint (i.e. Must-Link constraint) explicitly specifies which nodes must be partitioned together due to their close functional coupling or strong dependency in business logic. Therefore, for each pair of nodes specified by this constraint, the system locates their corresponding elements, i.e. the first target elements, in the similarity matrix and updates their numerical values to the first numerical value. This first numerical value is usually set to a very large positive number, such as 1 or a number much larger than other similarity values. Through such an operation, no matter what the original risk similarity of the two nodes is, they will show a very strong binding relationship in the updated matrix, thereby guiding the subsequent clustering algorithm to regard them as an indivisible whole.

[0141] Then update the second target elements indicated by the fragile independent partition constraint to a second numerical value in the similarity matrix W. The purpose of this process is to forcibly isolate high-risk nodes to actively block potential cascading failure propagation paths. In this process, the fragile independent partition constraint (i.e. Cannot-Link constraint) explicitly specifies the relationship that high-risk fragile nodes cannot be partitioned into the same subsystem. For each pair of fragile nodes specified by this constraint, the system locates their corresponding elements, i.e. the second target elements, in the similarity matrix W and updates their numerical values to the second numerical value. This second numerical value is usually set to 0 or a very small positive number. Such an update operation is equivalent to artificially cutting off the connection between these high-risk nodes, no matter how high their original risk similarity is, they will show a very strong repulsive relationship after updating, thereby ensuring that the clustering algorithm will divide them into different areas.

[0142] After completing the processing of all "same partition constraints" and "fragile independent partition constraints", i.e. updating the numerical values of all constraint-specified elements in the "similarity matrix" W, the modified matrix that combines data-driven similarity and rule-driven mandatory requirements, i.e. the constrained similarity matrix W', will serve as the direct input for subsequent spectral clustering analysis, carrying all the information to achieve intelligent, target-oriented system partitioning.

[0143] Through the accurate operations of steps 701 to 703, by performing targeted forced updating on the similarity matrix, the method can ensure that units that need to be cooperated in function are tightly bound, while effectively isolating high-risk vulnerable nodes, so that the subsequent clustering division is no longer a pure data mining process, but a "smart design" process with a clear optimization goal, so that the final "constrained similarity matrix" can guide the entire evaluation process to generate a system partition scheme that not only conforms to the internal laws of data, but also meets the needs of the proactive security defense strategy, greatly improving the practicality and forward-looking of security evaluation.

[0144] Step 602: obtaining a graph Laplacian matrix based on the difference between the node degree matrix and the constrained similarity matrix.

[0145] Step 603: obtaining a plurality of eigenvectors by solving the graph Laplacian matrix, and selecting a preset number of eigenvectors as an embedding matrix according to the order of the plurality of eigenvectors, the embedding matrix corresponding to an embedding space.

[0146] Step 604: performing clustering processing in the embedding space by using a clustering algorithm to obtain a clustering identifier of each node.

[0147] Step 605: dividing a plurality of nodes based on the clustering identifier to obtain a plurality of divided subsystems.

[0148] The steps 602 to 605 will be described in detail below.

[0149] In some embodiments, then, based on the difference between the node degree matrix D and the constrained similarity matrix W', a graph Laplacian matrix L = D - W' is obtained. This step is the standard core calculation link of the spectral clustering algorithm, which aims to convert the connection information of the graph into a matrix with excellent mathematical properties for spectral analysis. First, the node degree matrix D needs to be calculated, which is a diagonal matrix, and the ith element on the diagonal is equal to the sum of all elements in the ith row (or the ith column) of the constrained similarity matrix, representing the total connection "weight" or "degree" of node i. Then, subtract the "constrained similarity matrix" from this "node degree matrix", and the "difference" is the "graph Laplacian matrix". This matrix contains all the key information about the structure and connectivity of the graph, and its eigenvalues and eigenvectors can effectively reveal the best way to cut the graph.

[0150] Then, a plurality of eigenvectors are solved from the graph Laplacian matrix L, and according to the ordering of the plurality of eigenvectors, eigenvectors of a preset feature quantity k are selected as an embedding matrix, and the embedding matrix corresponds to an embedding space. This step is an embodiment of the dimension reduction idea of spectral clustering, and the purpose is to map the nodes from the original high-dimensional complex relationship space to a low-dimensional and easy-to-cluster space. By performing eigenvalue decomposition on the graph Laplacian matrix L, all eigenvectors of the graph Laplacian matrix L can be solved. These “eigenvectors” correspond to different vibration modes of the graph, and the eigenvectors corresponding to the smallest non-zero eigenvalues usually contain the optimal partition information of the graph. Therefore, the system will “select eigenvectors of a preset feature quantity” (the quantity is equal to the number k of the finally wanted partition subsystems) “according to the ordering of the plurality of eigenvectors” (usually in the order of the corresponding eigenvalues from small to large), and combine them to form a new matrix, that is, an “embedding matrix”. Each row of the “embedding matrix” represents a node in the original graph, but at this time, the node has been represented as a point in a k-dimensional space, and this new k-dimensional space is called an “embedding space”.

[0151] Then, a clustering algorithm is used to perform clustering processing in the embedding space to obtain a clustering identifier of each node. This step is to complete the final node grouping in the dimension-reduced space. In the “embedding space”, the nodes in the original graph that are structurally close, similar in risk characteristics and meet the partition constraints will be close to each other in geometric position, forming obvious clusters. At this time, a clustering algorithm, such as the classic K-Means algorithm, can be used to perform efficient “clustering processing” on the points in the k-dimensional space. The clustering algorithm will automatically divide the points into k different clusters, and assign a “clustering identifier” to each point (i.e., each node), which (such as 1, 2, …, k) explicitly indicates the cluster to which it belongs.

[0152] Finally, a plurality of nodes are divided based on the clustering identifier to obtain a plurality of partition subsystems. This step is to perform the final logical partitioning of the infrastructure system in the real world according to the calculation result of clustering. The system will traverse all infrastructure nodes and check the “clustering identifier” assigned to them. All nodes with the same “clustering identifier” will be grouped together to form a “partition subsystem”. Since the entire process starts from the “constraint similarity matrix” containing the mandatory constraints, the “partition subsystems” obtained finally naturally meet the dual goals of “effectively isolating key vulnerable nodes” and “completely retaining functional coupling units”, and the intelligent and optimized partitioning of the infrastructure system is completed.

[0153] By implementing the above steps 601 to 605, the complex graph partitioning problem is transformed into a simple clustering problem in a low-dimensional space through spectral analysis, so that not only can nodes with similar risk characteristics be aggregated together like traditional clustering, but more importantly, it can strictly enforce the pre-set partitioning constraints, thereby achieving "active design" of the system rather than passive analysis, so that the final partition subsystem has stronger structural resilience, effectively curbing the potential propagation path of cascading failures, and providing an optimized partitioning scheme with foresight and operability for subsequent safety assessment and actual operation and management.

[0154] Step 105: Perform safety assessment on each partition subsystem to obtain the safety assessment result of the infrastructure system.

[0155] The following describes step 105 in detail.

[0156] In some embodiments, after completing the optimized partitioning of the infrastructure system, in order to obtain a comprehensive and structured final evaluation conclusion, safety assessment of the infrastructure system is performed in two levels. First, for each independent partition subsystem, internal local safety assessment is performed to analyze the internal risks and resilience. Subsequently, from a global perspective, considering the macro dependence relationship between the partition subsystems and the potential cross-region propagation path, the evaluation results of all subsystems are synthesized and weighted to finally form a hierarchical and accurate "safety assessment result" of the entire infrastructure system. This evaluation method not only reflects the safety status of each part, but also reflects the overall safety level of the system after active risk isolation and structural optimization, as described below.

[0157] With reference to Figure 8 Performing safety assessment on each partition subsystem to obtain the safety assessment result of the infrastructure system includes the following steps 801 to 803.

[0158] Step 801: For each partition subsystem, obtain node information of all nodes in the partition subsystem and node association information between nodes, and determine a sub-safety assessment result of each partition subsystem based on all node information and node association information.

[0159] Step 802: Obtain system dependence relationship between the partition subsystems, and determine propagation influence relationship between the partition subsystems based on the system dependence relationship.

[0160] Step 803: Perform cascading safety assessment on all partition subsystems based on the propagation influence relationship and all sub-safety assessment results to obtain the safety assessment result of the infrastructure system.

[0161] The following describes steps 801 to 803 in detail.

[0162] In some embodiments, first, for each partitioned subsystem, node information of all nodes in the partitioned subsystem and node association information between the nodes are obtained, and a sub-security evaluation result of each partitioned subsystem is determined based on all the node information and the node association information. This step marks the focus of the evaluation work from global partition to local fine-grained analysis. For each "partitioned subsystem" formed in the previous step, the system takes it as an independent analysis unit. Within this unit, the system will comprehensively "obtain" the "node information of all nodes" (such as the vulnerability, load, etc. of the nodes) and the "node association information between the nodes" (i.e. the local topology within the subsystem) contained therein. Based on these detailed local data, a corresponding risk evaluation model or algorithm is used to calculate and "determine" a quantitative "sub-security evaluation result", which represents the robustness or security level of this particular partitioned subsystem itself.

[0163] Then, system dependency relationships between the partitioned subsystems are obtained, and a propagation influence relationship between the partitioned subsystems is determined based on the system dependency relationships. This step aims to elevate the analysis perspective from the subsystem to the macro level between the subsystems, and build a dependency model of "system of systems". The system will identify and "obtain" the "system dependency relationships" between different "partitioned subsystems", for example, the output of the power subsystem is the premise for the operation of the communication subsystem. Such dependency relationships are higher-level logical associations beyond individual node connections. Subsequently, based on these dependency relationships, the system will further quantify and model to "determine" the "propagation influence relationship between the partitioned subsystems", which can be a weighted directed graph or an influence matrix, which accurately describes how much and in what way a failure of one subsystem will affect other subsystems with which it has dependency relationships.

[0164] Finally, a cascading security evaluation is performed on all the partitioned subsystems based on the propagation influence relationship and all the sub-security evaluation results, to obtain a security evaluation result of the infrastructure system. In this process, the system integrates the results of the previous two steps: on the one hand, the health status of each subsystem itself, i.e. "all sub-security evaluation results"; on the other hand, the failure propagation paths between the subsystems, i.e. "propagation influence relationship". Through a "cascading security evaluation" algorithm, the system simulates how the failure of one or more subsystems with a lower security level spreads to other subsystems through the propagation influence relationship, and comprehensively considers all possible cascading paths and consequences. Through this global dynamic deduction, a "security evaluation result of the infrastructure system" is finally "obtained", which is no longer a simple superposition of individual risks, but a deep insight into the overall resilience of the entire system in the face of cascading failures.

[0165] Through the implementation of the above steps 801 to 803, the refined internal evaluation of each subsystem optimized by division is performed, and then on this basis, a macro inter-system dependence and influence model is constructed, and finally a global evaluation conclusion is obtained through simulation of the cascading process, so that it can be clearly revealed that the safety condition of the system depends not only on the strength of a single component or subsystem, but more profoundly on the coupling mode between them, so that the final safety evaluation result has an unprecedented depth and breadth, which can not only accurately locate the local risk source, but also can prospectively predict the cascading collapse risk at the system level, providing a comprehensive, three-dimensional and highly instructive risk view for decision makers.

[0166] Referring to Figure 9 is a flowchart of a method for evaluating the safety of an infrastructure system provided by an embodiment of the present application. As shown in Figure 9 , a complete technical flow of a constraint spectral clustering-based basic cascade safety evaluation method proposed by the present application is shown. First, the top of the flowchart shows the modeling process of a complex critical infrastructure system, that is, from a complex network composed of multiple industries (such as industries 1, 2, and 3) connected to each other, an infrastructure undirected graph required for processing by the scheme provided by the present application is abstracted. Then, for each node in the graph, the vulnerability curve of the node is determined by analyzing its internal attributes and external pressure, which quantifies the change law of the node failure probability with the increase of the pressure. On this basis, by simulating the propagation of one or more initial failure nodes (such as the red nodes shown in the figure) in the network, a failure scenario is generated, and the dynamic simulation failure probability of each node under the cascading effect is obtained. Further, the constraint partitioning region is divided: the risk similarity between nodes is calculated based on the failure probability obtained in the foregoing, and the preset node partitioning constraint (such as the constraint that the two red vulnerable nodes in the graph are forced to be divided into different regions) is combined, and the constraint spectral clustering algorithm is used to divide the entire system into multiple more resilient partition subsystems. Finally, the internal evaluation of each partition subsystem is performed, and the dependence relationship between the subsystems is comprehensively considered, and the final safety evaluation is performed, and an evaluation report with comprehensive, accurate and forward-looking risk insight is obtained.

[0167] The infrastructure system safety evaluation method, device, electronic equipment and storage medium provided by the embodiments of the present application comprise the following steps: first, an infrastructure undirected graph is obtained, which is generated by the logical connection relationship between a plurality of infrastructures and the infrastructures in an infrastructure system; then, the failure influence factor of each node is determined according to the node attribute information and the node connection relationship of each node, the failure pressure parameter of the node is obtained based on the failure influence factor, the vulnerability curve of the node is obtained based on the failure pressure parameter and the vulnerability function, and the vulnerability curve is used to indicate the failure probability of the node under different failure pressures; next, an initial fault node is determined from the plurality of nodes, and the plurality of nodes in the infrastructure system are simulated for a plurality of times of fault propagation based on the vulnerability curve of each node, the initial fault node and a plurality of fault propagation scenarios, to obtain the simulated failure probability of each node under each fault propagation scenario. In each fault propagation simulation, the initial fault node is set to a failure state, and an initial propagation queue is generated based on the initial fault node. For each fault node in the initial propagation queue, when the neighbor node of the fault node is in a normal state, the failure influence value of the neighbor node is obtained, and the failure pressure of the neighbor node is determined based on the failure influence value. Based on the failure pressure of the neighbor node and the vulnerability curve, the failure probability of the neighbor node is obtained. Based on the Bernoulli test and the failure probability, the update state of the neighbor node is determined. When the update state of the neighbor node is a failure state, the neighbor node is added to the initial propagation queue. When the neighbor node of the fault node is in a failure state, the neighbor node is added to the initial propagation queue. The simulated failure probability of the neighbor node of the fault node is determined based on the updated initial propagation queue, until the initial propagation queue is unchanged. Then, the node division constraint of the infrastructure system is obtained. Based on the difference between the simulated failure probabilities of each two nodes, divided by the square of the control parameter, the node probability difference value is obtained. Based on the negative value of the node probability difference value, the similarity between each two nodes is obtained by performing exponential processing. Based on all the similarities, a similarity matrix is obtained. In the similarity matrix, the first target element indicated in the same division constraint is updated to a first value, and the second target element indicated in the vulnerability independent division constraint is updated to a second value. Based on the updated similarity matrix, a constraint similarity matrix is obtained. Based on the difference between the node degree matrix and the constraint similarity matrix, a graph Laplacian matrix is obtained. A plurality of eigenvectors are obtained by solving the graph Laplacian matrix. According to the sorting of the plurality of eigenvectors, eigenvectors with a preset number of characteristics are selected as an embedding matrix. The embedding matrix corresponds to an embedding space. Clustering processing is performed in the embedding space by using a clustering algorithm to obtain a clustering identifier of each node. Based on the clustering identifier, the plurality of nodes are divided to obtain a plurality of division subsystems. Each division subsystem includes at most one vulnerable node, and at least one division subsystem includes a plurality of nodes with the same function. The vulnerable node is at least one of the plurality of nodes.Finally, for each divisional subsystem, node information of all nodes in the divisional subsystem and node association information between the nodes are obtained, and a sub-security evaluation result of each divisional subsystem is determined based on all the node information and the node association information, system dependency relationships between the divisional subsystems are obtained, and a propagation influence relationship between the divisional subsystems is determined based on the system dependency relationships, and a cascading security evaluation is performed on all the divisional subsystems based on the propagation influence relationship and all the sub-security evaluation results, to obtain a security evaluation result of the infrastructure system.

[0168] The embodiment of the application can dynamically quantify the real failure probability of a node in a complex network by constructing a vulnerability curve for each infrastructure node using an infrastructure undirected graph and performing failure propagation simulation based on the vulnerability curve, overcomes the defect that traditional static topology analysis cannot accurately reflect the cascading failure process, and further introduces node division constraints to forcibly disperse the identified vulnerable nodes to different subsystems in the clustering division stage, thereby achieving "pre-isolation" and structural optimization of system risks to block potential high-risk propagation paths, significantly enhancing the structural resilience of the entire system. Then, the embodiment of the application forms a complete closed loop from dynamic simulation to structural optimization and then hierarchical evaluation, thereby greatly improving the accuracy and forward-looking nature of the cascading security evaluation results of the key information infrastructure. In addition, the embodiment of the application comprehensively identifies internal and external factors affecting the stability of a node, quantifies the factors through a failure pressure parameter, and then establishes an accurate mapping relationship between the pressure and the failure probability by means of a flexible and configurable vulnerability function, so that the traditional evaluation of the fuzzy qualitative description of the node state is converted into an accurate, dynamic and calculable probability model, greatly improving the authenticity and accuracy of subsequent failure propagation simulation, so that the entire security evaluation system can be established on the basis of a solid and reliable micro model, thereby significantly improving the credibility and practical value of the final evaluation results. Furthermore, the embodiment of the application starts from calculating the pressure on a healthy node, combines the vulnerability curve and the probability judgment of Bernoulli trials to realistically simulate the uncertainty of node failure, and realizes the iterative diffusion of cascading failures in the network by dynamically updating the propagation queue, so that the macro failure simulation process is decomposed into a series of rigorous and repeatable micro operations, ensuring that each simulation not only follows physical and logical laws, but also contains randomness in the real world, so that the final simulation failure probability can highly realistically reflect the dynamic risks of the system, thereby providing a solid algorithm guarantee for the accuracy of the evaluation. Moreover, by setting an initial failure and performing large-scale repeated simulation, the embodiment of the application systematically explores all potential failure propagation paths in the network, uses the previously constructed vulnerability curve as a dynamic propagation rule, and accurately simulates the chain reaction of failure from one node to another node by iteratively updating the propagation queue, so that the security evaluation is no longer limited to static topology analysis, but can dynamically and quantitatively reveal the hidden cascading risks in the network, accurately identify key nodes that are not only vulnerable themselves but also easily affected by neighbors to fail, so that the final simulation failure probability of each node provides a highly reliable and insightful data basis for subsequent risk isolation and systematic security evaluation.In addition, by applying the Gaussian kernel function, the failure probability difference of the nodes is skillfully mapped to a standardized similarity space, so that the nodes with similar risk characteristics obtain high similarity, and vice versa, thereby providing a quantitative and intuitive basis for the subsequent clustering algorithm, so that the algorithm can clearly determine which nodes exhibit similar behavior patterns in the face of cascading failures, thereby providing a key data basis for realizing risk-based intelligent partitioning and actively isolating high-risk node groups, and is the core bridge connecting dynamic risk assessment and active structure optimization. In addition, by performing targeted forced updating on the similarity matrix, this method can ensure that units that need to cooperate in function are tightly bound, while effectively isolating high-risk vulnerable nodes, so that the subsequent clustering division is no longer a purely data mining process, but an "intelligent design" process with a clear optimization goal, so that the final "constrained similarity matrix" can guide the entire evaluation process to generate a system partitioning scheme that not only conforms to the internal laws of the data, but also meets the strategic needs of active security defense, greatly improving the practicality and forward-looking nature of security evaluation. In addition, by performing spectral analysis, the complex graph division problem is converted into a simple clustering problem in a low-dimensional space, so that not only can nodes with similar risk characteristics be aggregated together like traditional clustering, but more importantly, it can strictly implement the preset division constraints, thereby achieving "active design" of the system rather than passive analysis, so that the final partitioning subsystem already has stronger resilience in structure, effectively containing the potential propagation path of cascading failures, and providing an optimized partitioning scheme with forward-looking and operability for subsequent security evaluation and actual operation and management. Finally, the internal evaluation of each subsystem after optimized partitioning is performed in detail, and then a macro inter-system dependency and influence model is constructed, and finally the global evaluation conclusion is obtained through simulation of the cascading process, so that the safety condition of the system depends not only on the strength of individual components or subsystems, but also on the coupling mode between them, thereby making the final security evaluation result have unprecedented depth and breadth, which can not only accurately locate the local risk source, but also can forwardly predict the cascading collapse risk of the system level, thereby providing a comprehensive, three-dimensional, and highly instructive risk view for decision makers.

[0169] The embodiment of the present application also provides an infrastructure system security evaluation device, which can implement the infrastructure system security evaluation method. Figure 10 The device 1000 comprises:

[0170] The undirected graph acquisition module 1010 is configured to acquire an infrastructure undirected graph, wherein the infrastructure undirected graph is generated by the logical connection relationship between a plurality of infrastructures in the infrastructure system and the infrastructures.

[0171] The curve generation module 1020 is configured to generate a vulnerability curve corresponding to each node in the infrastructure undirected graph.

[0172] The failure probability simulation module 1030 is configured to obtain a simulated failure probability of each node based on the vulnerability curve of the node.

[0173] The subsystem division module 1040 is configured to obtain node division constraints of the infrastructure system, generate a similarity matrix based on the simulated failure probability of each node, and perform cluster division on the infrastructure system based on the similarity matrix and the node division constraints to obtain a plurality of division subsystems, each of which includes at most one vulnerable node, and at least one of the division subsystems includes a plurality of nodes with the same function, and the vulnerable node is at least one of the plurality of nodes.

[0174] The security evaluation module 1050 is configured to perform security evaluation on each division subsystem to obtain a security evaluation result of the infrastructure system.

[0175] In some embodiments, the curve generation module 1020 is further configured to:

[0176] determine a failure influence factor of the node according to the node attribute information and the node connection relationship of each node;

[0177] obtain a failure pressure parameter of the node based on the failure influence factor;

[0178] obtain the vulnerability curve of the node based on the failure pressure parameter and the vulnerability function, and the vulnerability curve is used to indicate a failure probability of the node under different failure pressures.

[0179] In some embodiments, the failure probability simulation module 1030 is further configured to:

[0180] determine an initial failure node from the plurality of nodes, and perform a plurality of failure propagation simulations on the plurality of nodes in the infrastructure system based on the vulnerability curve of each node, the initial failure node, and a plurality of failure propagation scenarios to obtain a simulated failure probability of each node under each failure propagation scenario;

[0181] In each failure propagation simulation, the initial failure node is set to a failure state, an initial propagation queue is generated based on the initial failure node, for each failure node in the initial propagation queue, a simulated failure probability of a neighbor node of the failure node is determined, and the initial propagation queue is updated based on the simulated failure probability of the neighbor node, and the simulated failure probability of the neighbor node of the failure node is determined based on the updated initial propagation queue until the initial propagation queue is unchanged.

[0182] In some embodiments, the failure probability simulation module 1030 is further configured to:

[0183] When the neighbor node of the faulty node is in the normal state, an influence value of the neighbor node is obtained, and a failure pressure of the neighbor node is determined based on the influence value;

[0184] Based on the failure pressure of the neighbor node and the fragility curve, a failure probability of the neighbor node is obtained, and an update state of the neighbor node is determined based on the Bernoulli trial and the failure probability;

[0185] When the update state of the neighbor node is the failure state, the neighbor node is added to the initial propagation queue;

[0186] When the neighbor node of the faulty node is in the failure state, the neighbor node is added to the initial propagation queue.

[0187] In some embodiments, the subsystem division module 1040 is further configured to:

[0188] Based on the square of the difference between the simulated failure probabilities of each two nodes, and then divided by the square of the control parameter, a node probability difference value is obtained;

[0189] Based on the negative value of the node probability difference value, and then exponential processing, a similarity between each two nodes is obtained;

[0190] Based on all the similarities, a similarity matrix is obtained.

[0191] In some embodiments, the subsystem division module 1040 is further configured to:

[0192] Based on the node division constraint, the similarity matrix is updated to obtain a constraint similarity matrix;

[0193] Based on the difference between the node degree matrix and the constraint similarity matrix, a graph Laplacian matrix is obtained;

[0194] Solving the graph Laplacian matrix obtains a plurality of feature vectors, and according to the sorting of the plurality of feature vectors, a feature vector with a preset feature quantity is selected as an embedding matrix, and the embedding matrix corresponds to an embedding space;

[0195] Using a clustering algorithm, clustering processing is performed in the embedding space to obtain a clustering identifier of each node;

[0196] Based on the clustering identifier, a plurality of nodes are divided to obtain a plurality of division subsystems.

[0197] In some embodiments, the subsystem division module 1040 is further configured to:

[0198] In the similarity matrix, the first target element indicated in the same division constraint is updated to the first numerical value;

[0199] In the similarity matrix, the second target element indicated in the fragile independent division constraint is updated to the second numerical value;

[0200] Based on the updated similarity matrix, a constraint similarity matrix is obtained.

[0201] In some embodiments, the security assessment module 1050 is further configured to:

[0202] For each divided subsystem, node information of all nodes in the divided subsystem and node association information between the nodes are obtained, and a sub-security assessment result of each divided subsystem is determined based on all the node information and the node association information;

[0203] A system dependency relationship between the divided subsystems is obtained, and a propagation influence relationship between the divided subsystems is determined based on the system dependency relationship;

[0204] Cascade security assessment is performed on all the divided subsystems based on the propagation influence relationship and all the sub-security assessment results, and a security assessment result of the infrastructure system is obtained.

[0205] In the above embodiments, the description of each embodiment has its own focus, and the specific embodiments of the infrastructure system security assessment apparatus are basically the same as the specific embodiments of the above infrastructure system security assessment method, which will not be described here.

[0206] The infrastructure system security evaluation device provided by the embodiment of the present application can construct a vulnerability curve for each infrastructure node by using an infrastructure undirected graph, and can simulate failure propagation based on the same, can dynamically quantify the real failure probability of the node in the complex network, and overcomes the defect that the traditional static topology analysis cannot accurately reflect the cascading failure process, and then by introducing node division constraints, the identified vulnerable nodes are forcibly dispersed into different subsystems in the clustering division stage, the "pre-isolation" and structural optimization of system risk are realized, the potential high-risk propagation path is blocked, the structural resilience of the whole system is significantly enhanced, then the local evaluation of each division subsystem after optimization is combined with global dependence for comprehensive evaluation, forming a complete closed loop from dynamic simulation to structural optimization and hierarchical evaluation, thereby greatly improving the accuracy and forward-looking of the cascading security evaluation result of the key information infrastructure; in addition, the internal and external factors affecting the stability of the node are comprehensively identified, and then the failure pressure parameter is used for quantization, and then the flexible and configurable vulnerability function is used to establish an accurate mapping relationship between the pressure and the failure probability, so that the fuzzy qualitative description of the node state in the traditional evaluation is changed into an accurate, dynamic and calculable probability model, which greatly improves the authenticity and accuracy of the subsequent failure propagation simulation, so that the whole security evaluation system can be established on the basis of a solid and reliable micro model, thereby significantly improving the credibility and practical value of the final evaluation result; and starting from the pressure borne by the healthy node, the uncertainty of node failure is truly simulated by combining the vulnerability curve and the probability judgment of Bernoulli test, and the iterative diffusion of cascading failure in the network is realized by dynamically updating the propagation queue, so that the macro failure simulation process is decomposed into a series of rigorous and repeatable micro operations, ensuring that each simulation not only follows the physical and logical laws, but also contains the randomness in the real world, so that the final simulation failure probability can highly realistically reflect the dynamic risk of the system, providing a solid algorithm guarantee for the accuracy of the evaluation; and by setting an initial failure and performing large-scale repeated simulation, all potential failure propagation paths in the network are systematically explored, the vulnerability curve constructed in advance is used as a dynamic propagation rule, and the iterative updating of the propagation queue is used to accurately simulate the chain reaction of failure from one node to another, so that the security evaluation is no longer limited to static topology analysis, but can dynamically and quantitatively reveal the hidden cascading risk in the network, accurately identify those key nodes which are not only vulnerable themselves, but also easy to fail under the influence of neighbors, so that the final simulation failure probability of each node provides a highly reliable and insightful data basis for subsequent risk isolation and systematic security evaluation.In addition, by applying the Gaussian kernel function, the failure probability difference of the nodes is skillfully mapped to a standardized similarity space, so that nodes with similar risk characteristics obtain high similarity, and vice versa, thus providing a quantitative and intuitive basis for the subsequent clustering algorithm, so that the algorithm can clearly determine which nodes exhibit similar behavior patterns when facing cascading failures, thus providing a key data basis for realizing risk-based intelligent partitioning and actively isolating high-risk node groups, and is the core bridge connecting dynamic risk assessment and active structure optimization. In addition, by performing targeted forced updating on the similarity matrix, this method can ensure that functionally cooperative units are tightly bound, while effectively isolating high-risk vulnerable nodes, so that the subsequent clustering division is no longer a purely data mining process, but an "intelligent design" process with a clear optimization goal, so that the final "constrained similarity matrix" can guide the entire evaluation process to generate a system partitioning scheme that not only conforms to the internal laws of the data, but also meets the strategic needs of active security defense, greatly improving the practicality and forward-looking nature of security evaluation. In addition, by spectral analysis, the complex graph division problem is converted into a simple clustering problem in low-dimensional space, so that not only can nodes with similar risk characteristics be aggregated together like traditional clustering, but more importantly, it can strictly enforce pre-set division constraints, thus achieving "active design" of the system rather than passive analysis, so that the final partitioning subsystem already has stronger resilience in structure, effectively containing the potential propagation path of cascading failures, and providing an optimized partitioning scheme with forward-looking and operational nature for subsequent security evaluation and actual operation and management. Finally, the optimized partitioned subsystems are subjected to fine internal evaluation, and then a macro inter-system dependency and influence model is constructed on this basis, and finally the global evaluation conclusion is obtained through simulation of the cascading process, so that the safety condition of the system depends not only on the strength of individual components or subsystems, but also on the coupling mode between them, so that the final security evaluation result has an unprecedented depth and breadth, which can not only accurately locate the local risk source, but also can forward-lookingly predict the system-level cascading collapse risk, thus providing a comprehensive, three-dimensional, and highly instructive risk view for decision-makers.

[0207] The embodiment of the present application also provides an electronic device, comprising:

[0208] at least one memory;

[0209] at least one processor;

[0210] at least one program;

[0211] The program is stored in the memory, and the processor executes the at least one program to implement the infrastructure system security assessment method described above. The electronic device can be any smart terminal including a mobile phone, a tablet computer, a personal digital assistant (PDA), a vehicle-mounted computer, etc.

[0212] Please refer to Figure 11 , Figure 11 The hardware structure of the electronic device of another embodiment is illustrated, and the electronic device includes:

[0213] The processor 1101 can be implemented in the form of a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, etc., for executing related programs to implement the technical solutions provided by the embodiments of the present application.

[0214] The memory 1102 can be implemented in the form of a ROM (Read Only Memory), a static storage device, a dynamic storage device, or a RAM (Random Access Memory), etc. The memory 1102 can store an operating system and other application programs, and when the technical solutions provided by the embodiments of the present application are implemented by software or firmware, the related program codes are saved in the memory 1102 and called and executed by the processor 1101 to implement the infrastructure system security assessment method of the embodiments of the present application.

[0215] The input / output interface 1103 is used to realize information input and output.

[0216] The communication interface 1104 is used to realize the communication interaction between the device and other devices, which can realize communication through wired means (such as USB, network cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.).

[0217] The bus 1105 transmits information between various components (such as the processor 1101, the memory 1102, the input / output interface 1103, and the communication interface 1104) of the device.

[0218] Among them, the processor 1101, the memory 1102, the input / output interface 1103, and the communication interface 1104 realize the communication connection between each other inside the device through the bus 1105.

[0219] The embodiment of the present application further provides a storage medium, which is a computer readable storage medium, and stores a computer program. The computer program is executed by a processor to implement the infrastructure system security assessment method.

[0220] The memory is a non-transitory computer readable storage medium, and can be used to store a non-transitory software program and a non-transitory computer executable program. In addition, the memory can include a high-speed random access memory, and can further include a non-transitory memory, such as at least one magnetic disk storage device, a flash memory device, or other non-transitory solid-state memory device. In some embodiments, the memory can optionally include a memory remotely arranged relative to the processor, and the remote memory can be connected to the processor through a network. Examples of the network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and a combination thereof.

[0221] The embodiments described in the embodiments of the present application are used to more clearly illustrate the technical solutions of the embodiments of the present application, and do not constitute a limitation on the technical solutions provided by the embodiments of the present application. Those skilled in the art can understand that, with the evolution of technology and the appearance of new application scenarios, the technical solutions provided by the embodiments of the present application are also applicable to similar technical problems.

[0222] Those skilled in the art can understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of the present application, and can include more or fewer steps than those shown in the figures, or combine certain steps, or different steps.

[0223] The device embodiments described above are merely schematic, and the units described as separate components can or can not be physically separated, that is, can be located in one place, or can be distributed on multiple network units. According to actual needs, part or all of the modules can be selected to achieve the purpose of the embodiments of the present application.

[0224] Those skilled in the art can understand that all or some steps in the method disclosed above, the function modules / units in the system and the device can be implemented as software, firmware, hardware and appropriate combinations thereof.

[0225] The terms "first", "second", "third", "fourth", and the like in the description of this application and in the claims hereof, if any, are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the use of the terms so termed herein is solely for the convenience of the reader and does not limit the scope of the application. It is also to be understood that the description and examples in this application are intended to cover all possible combinations where any of the several elements can represent one or more elements.

[0226] It should be understood that, in this application, "at least one" means one or more, "multiple" means two or more. "And / or" is used to describe the relationship between associated objects, which means that there can be three relationships, for example, "A and / or B" can mean: only A, only B, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally represents an "or" relationship between the associated objects. "At least one of the following" or similar expressions means any combination of these items, including any combination of single or multiple items. For example, at least one of a, b or c, can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0227] In several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only illustrative, for example, the division of the above-mentioned units is only a logical function division, and actual implementation can have another division manner, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. The coupling or direct coupling or communication connection between the displayed or discussed each other can be indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0228] The units described above as separate components can or can not be physically separated, and the components shown as units can or can not be physical units, that is, they can be located in one place, or they can be distributed on a plurality of network units. According to actual needs, part or all of the units can be selected to achieve the purpose of the embodiment of the present application.

[0229] In addition, each of the functional units in the embodiments of the present application can be integrated in one processing unit, or each unit can exist alone physically, or two or more units can be integrated in one unit. The integrated unit can be implemented in the form of hardware, or in the form of a software functional unit.

[0230] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer readable storage medium. Based on such an understanding, the technical solutions of the present application, essentially or partially, or all or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes multiple instructions used to cause a computer device (such as a personal computer, a server, or a network device) to perform all or part of the steps of the methods in the embodiments of the present application. The foregoing storage medium includes: various memories (such as a read-only memory, a random access memory, a flash memory, or the like) and a magnetic disk or an optical disk and the like.

[0231] The preferred embodiments of the embodiments of the present application are described above with reference to the accompanying drawings, and are not intended to limit the scope of the embodiments of the present application. Any modification, equivalent replacement, and improvement made by those skilled in the art without departing from the scope and spirit of the embodiments of the present application shall fall within the scope of the embodiments of the present application.

Claims

1. A method for infrastructure system security assessment, characterized in that: The method comprises: Acquire an infrastructure undirected graph, where the infrastructure undirected graph is generated by logical connection relationships between a plurality of infrastructures and the infrastructure in the infrastructure system; generating a vulnerability curve corresponding to each node in the infrastructure undirected graph; Obtaining a simulated failure probability of each node based on a vulnerability curve of each node; Obtaining node partitioning constraints for the infrastructure system, generating a similarity matrix based on a simulated failure probability of each node, and clustering the infrastructure system based on the similarity matrix and the node partitioning constraints to obtain a plurality of partitioned subsystems, each of which includes at most one vulnerable node, and at least one of which includes multiple nodes with the same function, wherein the vulnerable node is at least one of the multiple nodes; A security assessment is performed on each of the divided subsystems to obtain a security assessment result of the infrastructure system.

2. The infrastructure system security assessment method according to claim 1, characterized in that: Generating a vulnerability curve corresponding to each node in the infrastructure undirected graph includes: Determining failure influencing factors of the node according to the node attribute information and node connection relationship of each node; Based on the failure influencing factors, obtaining a failure pressure parameter of the node; Based on the failure pressure parameter and the fragility function, a fragility curve of the node is obtained, where the fragility curve is used to indicate the failure probability of the node under different failure pressures.

3. The infrastructure system security assessment method according to claim 2, characterized in that: The obtaining of a simulated failure probability of each node based on the vulnerability curve of each node includes: Determining an initial fault node from the plurality of nodes, and performing multiple fault propagation simulations on the plurality of nodes in the infrastructure system based on the vulnerability curve of each node, the initial fault node, and multiple fault propagation scenarios to obtain a simulated failure probability of each node under each fault propagation scenario; In each fault propagation simulation, the initial fault node is set to a failed state, and an initial propagation queue is generated based on the initial fault node. For each fault node in the initial propagation queue, the simulated failure probability of the neighboring nodes of the fault node is determined, and the initial propagation queue is updated based on the simulated failure probability of the neighboring nodes. The simulated failure probability of the neighboring nodes of the fault node is determined based on the updated initial propagation queue until the initial propagation queue remains unchanged.

4. The infrastructure system security assessment method according to claim 3, characterized in that: The determining of the simulated failure probabilities of neighboring nodes of the faulty node and updating the initial propagation queue based on the simulated failure probabilities of the neighboring nodes includes: When the neighboring node of the faulty node is in a normal state, obtaining a failure impact value of the neighboring node, and determining a failure pressure of the neighboring node based on the failure impact value; Obtaining a failure probability of the neighboring node based on the failure pressure of the neighboring node and the fragility curve, and determining an update state of the neighboring node based on a Bernoulli test and the failure probability; When the update status of the neighbor node is in an invalid state, adding the neighbor node to the initial propagation queue; When the neighboring node of the faulty node is in a failed state, the neighboring node is added to the initial propagation queue.

5. The infrastructure system security assessment method according to claim 1, characterized in that: Generating a similarity matrix based on the simulated failure probability of each node includes: A node probability difference is obtained by dividing the square of the difference between the simulated failure probabilities of each two nodes by the square of the control parameter; Based on the negative value of the node probability difference, exponential processing is performed to obtain the similarity between each two nodes; Based on all the similarities, the similarity matrix is ​​obtained.

6. The infrastructure system security assessment method according to claim 1, characterized in that: The infrastructure system is clustered and divided based on the similarity matrix and the node division constraint to obtain a plurality of divided subsystems, including: Performing constraint updating on the similarity matrix based on the node partitioning constraint to obtain a constrained similarity matrix; Obtaining a graph Laplacian matrix based on a difference between the node degree matrix and the constraint similarity matrix; Solving the graph Laplacian matrix to obtain a plurality of eigenvectors, and selecting a preset number of eigenvectors as an embedding matrix according to the order of the plurality of eigenvectors, wherein the embedding matrix corresponds to an embedding space; Using a clustering algorithm, clustering is performed in the embedding space to obtain a cluster identifier for each node; Based on the cluster identification, the plurality of nodes are divided to obtain a plurality of divided subsystems.

7. The infrastructure system security assessment method according to claim 6, characterized in that: The node partitioning constraints include identical partitioning constraints and fragile independent partitioning constraints. The similarity matrix is ​​updated based on the node partitioning constraints to obtain a constrained similarity matrix, including: In the similarity matrix, updating the first target element indicated in the same partition constraint to a first value; In the similarity matrix, updating a second target element indicated in the fragile independent partition constraint to a second value; Based on the updated similarity matrix, the constraint similarity matrix is ​​obtained.

8. The infrastructure system security assessment method according to claim 1, characterized in that: The performing of a security assessment on each of the divided subsystems to obtain a security assessment result of the infrastructure system includes: For each of the divided subsystems, obtaining node information of all nodes in the divided subsystem and node association information between the nodes, and determining a sub-security assessment result of each of the divided subsystems based on all the node information and the node association information; Acquiring system dependencies between the divided subsystems, and determining a propagation influence relationship between the divided subsystems based on the system dependencies; A cascade security assessment is performed on all the divided subsystems based on the propagation impact relationship and all the sub-security assessment results to obtain the security assessment result of the infrastructure system.

9. An infrastructure system security assessment device, characterized in that: The device comprises: An undirected graph acquisition module, configured to acquire an infrastructure undirected graph, wherein the infrastructure undirected graph is generated by logical connection relationships between a plurality of infrastructures and the infrastructure in the infrastructure system; A curve generation module, configured to generate a vulnerability curve corresponding to each node in the infrastructure undirected graph; A failure probability simulation module, configured to obtain a simulated failure probability of each node based on a vulnerability curve of each node; a subsystem partitioning module, configured to obtain node partitioning constraints of the infrastructure system, generate a similarity matrix based on the simulated failure probability of each node, and cluster the infrastructure system based on the similarity matrix and the node partitioning constraints to obtain a plurality of partitioned subsystems, each of which includes at most one vulnerable node, and at least one of which includes multiple nodes with the same function, wherein the vulnerable node is at least one of the multiple nodes; The security assessment module is used to perform a security assessment on each of the divided subsystems to obtain a security assessment result of the infrastructure system.

10. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the infrastructure system security assessment method according to any one of claims 1 to 8 is implemented.

11. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the infrastructure system security assessment method according to any one of claims 1 to 8 is implemented.

Images