A risk assessment method for power information network nodes based on DW-Degree centrality

By adopting the DW-Degree centrality evaluation method in the power information network and combining the attack status data, the problem of insufficient targeted and one-sidedness of the risk assessment method in the existing technology is solved, and higher risk assessment accuracy and classification accuracy are achieved. ensures the security of the power information network.

CN114493246BActive Publication Date: 2024-11-26NORTHEASTERN UNIV CHINA +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210080296.7
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-01-24
Publication Date
2024-11-26
Estimated Expiration
2042-01-24

AI Technical Summary

Technical Problem

The existing risk assessment methods for power information network nodes are insufficiently targeted, one-sided and subjective, and fail to effectively consider the diversity, directionality and heterogeneity of network attacks, resulting in low accuracy of risk assessment results, poor practicality and reference. high.

Method used

The risk assessment method of power information network nodes based on DW-Degree centrality is adopted, and the attack event threat level, type, stage and result are comprehensively considered. The node's importance is measured through DW-Degree centrality, and the attack status data is used to calculate the node's importance. Initial risk coefficient, forming a more accurate risk assessment.

Benefits of technology

The accuracy and practicality of node risk assessment have been improved, and the classification accuracy rate has reached 95.652%, which can more effectively identify high-risk nodes and guide the security measures of the power information network.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114493246B_ABST
    Figure CN114493246B_ABST
Patent Text Reader

Abstract

A method for risk assessment of power information network nodes based on DW‑Degree centrality, comprising the following steps: selecting key factors that affect the security of power information network nodes to characterize the attack status of nodes; forming attack status data based on the power information network attack log; generating risky training data based on the victim IP; calculating the initial risk coefficient of the node through the node risk assessment method; calculating the threat coefficient between nodes based on the attack status data, abstracting the power information network into a directed weighted network; calculating the importance of nodes in the directed weighted network; fusing the initial risk coefficient and importance index of the node to calculate the final risk coefficient of the node. The power information network node risk assessment method proposed in the present invention has a higher accuracy rate and a higher degree of fit with the power information network. The present invention comprehensively considers multiple factors, calculates the initial risk coefficient of the node, and makes the node risk assessment method more compatible with the power information network.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the technical field of power grid intelligent control and relates to a power information network node risk assessment method based on DW-Degree centrality. Background Art

[0002] The power information network is a communication network dedicated to the power system. It carries all services for power production and management, and has an important impact on the safe and stable operation of the power grid. With the advancement of "Made in China 2025", the power industry has begun to upgrade its industry to informatization, digitization, and intelligence. At the same time, as a large number of power equipment are inevitably connected to the Internet, the power information network is also facing increasingly severe network attacks. Node failures may cause power service interruptions or increased delays. Timely and accurate identification of high-risk network nodes in the network is crucial for network security protection. The existing network node risk assessment method is mainly based on the multiplication method, that is, the product of the probability of risk events and the degree of impact of risk events on the network. The multiplication method mainly involves three elements: assets, threats, and vulnerabilities, and can quantitatively calculate the risk value of conventional network equipment.

[0003] The paper "Fan B, Zheng C, Tang L. Risk Assessment of Power Communication Network Based on Node Importance[C]. 2019IEEE 3rd Advanced Information Management, Communicates, Electronic and Automation Control Conference (IMCEC). 2019: 818-821." proposes a risk assessment algorithm for power communication networks based on node importance. This paper abstracts the power communication network as a complex network without direction and weight. First, from the perspective of network topology, the node degree and clustering coefficient are used to quantify the impact of node failure on the network structure; then, the concept of service importance is introduced, and the importance of various services is calculated through the characteristic index evaluation method. The importance of the service carried by the node is used as the impact on the power service transmission process; finally, the information fusion operator is used to fuse the impact of node failure on the network structure and power service, and the risk assessment of node failure on the power communication network is completed.

[0004] The paper "Li J, Liu YQ, Niu Y, et al. Network Node Resource Risk AssessmentBased on Bayesian Belief Network Model[C]. 2018IEEE International Conferenceon Progress in Informatics and Computing(PIC). 2018:404-408." selects memory usage, network traffic and CPU utilization as network node resource indicators, and designs a network node resource risk assessment method based on Bayesian network to quantify the single indicator and overall risk of network node resource risk. First, a Bayesian belief network model graph is constructed for different node resources to calculate the prior probability. Then, the joint probability is calculated based on the given data and the correlation between them. Then, the total probability of network node resources is solved according to the total probability law. Finally, the probability value of risk of different network node resources is calculated based on the total probability.

[0005] The paper "Fan B, Zheng C, Tang L. Risk Assessment of Power Communication Network Based on Node Importance [C]. 2019 IEEE 3rd Advanced Information Management, Communicates, Electronic and Automation Control Conference (IMCEC). 2019: 818-821." evaluates the node risk in the power communication network from the perspective of network structure and power service. Among them, the network structure mainly considers the degree and clustering coefficient of the node. However, through experiments, it is found that there is no correlation between the neighbors of all nodes in the power information network, that is, the clustering coefficient is 0, so it is not suitable to use the clustering coefficient to measure its importance. The author abstracts the power communication network as an undirected and unweighted network, which is inconsistent with the directed and heterogeneous characteristics of power information network attacks. That is, node A can attack node B, which does not mean that node B can attack node A, and the impact of attacks between nodes is not exactly the same.

[0006] Important nodes in the network are the key to maintaining the normal operation of the entire network. When the node risk values ​​are the same, important equipment needs to be protected first. The paper "Li J, Liu YQ, Niu Y, et al. Network Node Resource Risk Assessment Based on Bayesian Belief Network Model [C]. 2018 IEEE International Conference on Progress in Informatics and Computing (PIC). 2018: 404-408" ignores the impact of node importance on node risk value, which is still lacking in pertinence for power information networks and makes it difficult to accurately assess the risk value of nodes in power information networks.

[0007] In summary, the existing power information network node risk assessment methods have problems such as lack of pertinence, one-sidedness, subjectivity, and failure to consider the diversity, directionality, and heterogeneity of network attacks, resulting in low accuracy of node risk assessment results, and low practicality and reference value.

[0008] Important nodes in the network are the key to maintaining the normal operation of the entire network. When the node risk values ​​are the same, important equipment needs to be protected first. Summary of the invention

[0009] The present invention proposes a method for risk assessment of power information network nodes that is highly compatible with the power information network and has high accuracy. The method comprehensively considers multiple factors such as the threat level of attack events in the power information network, the type of attack events, the attack stage, and the attack results, and calculates the initial risk coefficient of the node, so that the node risk assessment method is more compatible with the power information network. The present application proposes DW-Degree centrality to measure the importance of nodes. DW-Degree centrality not only considers the weight of the edge, the number of adjacent nodes, and the direction of the edge, but also considers the heterogeneity of the direction of the edge, which can more accurately calculate the importance of nodes in a directed weighted network with directional heterogeneity.

[0010] The case analysis results based on the power information network attack data show that compared with the traditional multiplication method, spectral clustering method and NRAMAS, the node risk classification accuracy of the proposed NRAMDD method reached 95.652%, indicating that the NRAMDD method has achieved an accurate assessment of the node risk of the power information network, thereby guiding the power information network security personnel to quickly and accurately deploy countermeasures. Compared with the traditional multiplication method accuracy of 44.1546% and the spectral clustering method accuracy of 56.8721%, NRAMAS has a higher classification accuracy of 91.0145%, while NRAMDD has the highest classification accuracy of 95.652%. The distribution of risk coefficients and importance obtained by visualizing NRAMDD is shown in the figure below. Figure 2 As shown in the figure, it can be seen that the risk factor and importance of node I5485 are very high, and it is the first node that needs to be processed urgently. Focusing on importance and taking into account the risk factor, the processing order of the nodes is shown in the direction of the broken line arrow in the figure, from top to bottom and from right to left.

[0011] The technical solutions adopted by the present invention are as follows:

[0012] A method for risk assessment of power information network nodes based on DW-Degree centrality includes the following steps:

[0013] Step 1: Select the key factors that affect the security of power information network nodes and characterize the attack status of the nodes.

[0014] Step 2: Based on the power information network attack log attackData, form attack status data.

[0015] Step 3: Extract victim IP queue IPs based on attack status data and generate training data with risk level labels.

[0016] Step 4: Initialize the impact factor set: A, B, C, D.

[0017] Step 5: Calculate the initial risk coefficient of the node through the Node Risk Assessment Methodbased on Attack Status (NRAMAS). The algorithm idea is:

[0018] ① Calculate the number of attack state types N:

[0019] N=sizeof(A)+sizeof(B)+sizeof(C)+sizeof(D);

[0020] ② Initialize the attack state matrix S: In the formula, s l ∈S, l=1,2,...,N, k1=1,2,3,4,5, k2=1,2,3, k3=1,2,3,4, k4=1,2;

[0021] ③Construct the network attack risk matrix G: g l =l / size(S), where g l ∈G, l=1,2,...,N, size(G)=size(S);

[0022] ④If i=IPs.begin(), otherwise go to ⑩;

[0023] ⑤ Calculate the attack state probability matrix P of node i based on the training data i :

[0024] ⑥ Calculate the risk value F(i) of node i:

[0025] ⑦If F(i) is within the risk coefficient range corresponding to its risk level label, then i.classLabel = 1, otherwise i.classLabel = 0;

[0026] ⑧Storage f i , f i (i,F(i),i.riskLevel,i.classLevel)∈F;

[0027] ⑨Delete the first element of IPs and go to ④;

[0028] ⑩Output F.

[0029] Step 6: Calculate the threat coefficient between nodes based on the attack status data. The specific steps are as follows:

[0030] ① Calculate the number of attack state types N: N = sizeof(A) + sizeof(B) + sizeof(C) + sizeof(D);

[0031] ② Initialize the attack state matrix S: In the formula, s l ∈S, l=1,2,...,N, k1=1,2,3,4,5, k2=1,2,3, k3=1,2,3,4, k4=1,2;

[0032] ③ Based on attackData statistics, the set of edges in the power information network: e ij ={i,j},e ij∈E, when the attack result of node i pointing to node j is one of attempt, success, or failure, e ij exist;

[0033] ④Build the network attack risk matrix G: g l =l / size(S), where g l ∈G, l=1,2,...,N, size(G)=size(S);

[0034] ⑤If eij=E.begin(), otherwise go to ⑩;

[0035] ⑥ Calculate the attack state probability matrix Pij of edge eij based on attackData: In the formula, Indicates the number of attacks when node i attacks node j with the attack state sl, attackNum ij represents the total number of times node i attacks node j, represents the probability that the attack state of node i is sl when attacking node j, size(P ij ) = size(G);

[0036] ⑦ Calculate the threat coefficient Tij corresponding to edge eij:

[0037] ⑧T.push_back(i,j,Tij);

[0038] ⑨Delete the first element of E and go to ⑤;

[0039] ⑩Output the threat coefficient vector T between nodes.

[0040] Step 7: Calculate the importance of the node through DW-Degree centrality. The specific steps are as follows:

[0041] ① Count the number of adjacent nodes of node i, m i The calculation formula is: i represents the focal node, j represents all other nodes, N is the total number of nodes, X is the adjacency matrix, x ij is an element in the adjacency matrix X. If node i is connected to node j, then x ij =1, otherwise x ij =0;

[0042] ② Calculate the sum of the weights of the edges connected to node i, n i The calculation formula is: Where W is the weight matrix, w ijis an element in the weight matrix W. If node i is connected to node j, then w ij >0, otherwise w ij =0;

[0043] ③Calculate the node strength H out (i) and penetration strength H in (i), the calculation formula is: i is the focal node, λ is a positive adjustment parameter, and λ can be set according to research and data. When the sum of the weights of adjacent edges is the same: when 0<λ<1, the node with more adjacent nodes has a greater degree centrality, and when λ>1, the node with fewer adjacent nodes has a greater degree centrality;

[0044] ④ Calculate the importance of each node i in the power information network. The calculation formula is: In the formula, e refers to a natural constant, whose value is approximately 2.718281828459045, μ and ν are balance factors, μ+ν=1, when the out-degree and in-degree importance of the node are the same, μ=ν=0.5; when the out-degree importance of the node is greater than the in-degree importance, μ>0.5, ν<0.5; when the out-degree importance of the node is less than the in-degree importance, μ<0.5, ν>0.5.

[0045] Step 8: Fusion node's initial risk coefficient and importance index to calculate the node's final risk coefficient. The formula is: Where F(i) is the risk coefficient of node i obtained based on NRAMAS, is the DW-Degree centrality of node i, α and β are adjustment factors, and the α and β with the highest classification accuracy can be obtained through the node risk assessment method based on DW-Degree - NRAMDD.

[0046] Methods - The specific steps of NRAMDD are as follows:

[0047] ① Initialize adjustment factors α, β, max(α), max(β), delta, and maximum accuracy max(AC)

[0048] ② Obtain F, f through step 5 i (i,F(i),i.riskLevel,i.classLevel)∈F;

[0049] ③ Calculate the importance of each node (DW-Degree centrality) based on the constructed directed weighted network;

[0050] ④Match node importance and generate data RI, r i(i,F(i),i.riskLevel,i.importance,i.label)∈RI;

[0051] ⑤If α≤max(α), set β=0.1, otherwise go to ⑨;

[0052] ⑥If β≥max(β), α=α+delta, go to ⑤; otherwise, update the risk factor F(i) of each node in RI: If F(i) is within the risk coefficient range corresponding to its risk level label, i.label = 1;

[0053] ⑦Calculate the accuracy AC: In the formula, |X| represents the number of samples in the dataset X, |C i | represents the number of samples of the i-th level data that are correctly classified. If AC>max(AC), max(AC)=AC, α best =α,β best =β;

[0054] ⑧β=β+delta, transfer to ⑥;

[0055] ⑨Update RI:

[0056] ⑩ Output α best , β best , max(AC).

[0057] The beneficial effects of the present invention are as follows:

[0058] (1) Taking into account multiple factors such as the threat level, type, stage and result of attack events in the power information network, the calculated initial risk coefficient of the node is more comprehensive and objective, and has a higher degree of fit with the power information network;

[0059] (2) In view of the fact that traditional power information network abstraction methods do not consider the diversity, directionality and heterogeneity of network attacks, the power information network is abstracted into a directed weighted network to ensure the accuracy of node importance calculation in the power information network;

[0060] (3) The importance of nodes is measured using DW-Degree centrality, which not only considers the weight of the edge, the number of adjacent nodes, and the direction of the edge, but also the heterogeneity of the direction of the edge, thus improving the measurement accuracy of node importance in directed weighted networks with directional heterogeneity.

[0061] (4) Based on the attack status-based node risk assessment method NRAMAS, the improved DW-Degree centrality is combined to quantify the important coefficient of the node, and finally a more accurate node risk coefficient is obtained. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 This is a process flow chart of the risk assessment method for power information network nodes.

[0063] Figure 2 This is the relationship diagram between node importance and risk factor. DETAILED DESCRIPTION

[0064] The specific implementation of the present invention is further described in detail below in conjunction with the drawings and examples.

[0065] This embodiment describes in detail the algorithm flow of a power information network node risk assessment method based on DW-Degree centrality in the implementation and application of the present invention, and is applied to the problem of identifying high-risk network equipment in the field of network security.

[0066] The present invention uses the power information network attack data of a certain power company, which contains 85,207 alarm data. Each alarm record contains 9 features: attack time, attack IP, victim IP, attack event, attack result, threat level, attack stage, number of attacks and attack source. The detailed information is shown in Table 2.

[0067] Table 2 is a table of characteristics of power information network attack data.

[0068]

[0069] Accurately identifying high-risk equipment in the power information network can help network security personnel quickly take effective targeted measures to ensure the safe and stable operation of the power information network. A power information network node risk assessment method based on DW-Degree centrality, the algorithm steps are as follows:

[0070] Step 1: Select four factors, namely, attack event threat level, attack result, attack stage and attack source, to characterize the attack status of the node.

[0071] Step 2: Generate attack status data based on power information network attack data.

[0072] Step 3: Extract victim IP queue IPs based on attack status data, generate 1035 training data with risk level labels, each training data sample contains multiple alarm data. Among them, there are 222 low-risk data (21.45%), 430 medium-risk data (41.55%), 345 high-risk data (33.33%), and 38 severe risk data (3.67%). The node risk is divided into 4 levels: 0-0.25 is low risk; 0.25-0.5 is medium risk; 0.5-0.75 is high risk; 0.75-1 is severe risk.

[0073] Step 4: Initialize the impact factor set: A[4] = {low risk, medium risk, high risk, critical}, B[3] = {attempt, successful attack, compromise}, C[5] = {reconnaissance, purpose execution, intrusion, lateral penetration, command and control}, D[2] = {power information network, Internet}.

[0074] Step 5: Calculate the initial risk factor of the node through NRAMAS, including:

[0075] ① Calculate the number of attack state types N = 120 through N = sizeof(A) + sizeof(B) + sizeof(C) + sizeof(D);

[0076] ② Initialize the attack state matrix S: In the formula, s l ∈S, l=1,2,...,N, k1=1,2,3,4,5,k2=1,2,3,k3=1,2,3,4,k4=1,2

[0077] ③Construct the network attack risk matrix G: g l =l / size(S), where g l ∈G, l=1,2,...,N;

[0078] ④If i=IPs.begin(), otherwise go to ⑩;

[0079] ⑤ Calculate the attack state probability matrix P of node i based on the training data i :

[0080] ⑥ Calculate the risk value F(i) of node i:

[0081] ⑦If F(i) is within the risk coefficient range corresponding to its risk level label, then i.classLabel = 1, otherwise i.classLabel = 0;

[0082] ⑧Storage f i , f i (i,F(i),i.riskLevel,i.classLevel)∈F;

[0083] ⑨Delete the first element of IPs and go to ④;

[0084] ⑩Output F.

[0085] The results of part F obtained in step five are shown in Table 4. Among the 1035 training data, the risk assessment accurately identified 942 items, with an accuracy rate of 91.0145%.

[0086] Table 4 shows some results of the initial risk index of nodes calculated by NRAMAS

[0087]

[0088] Step 6: Calculate the threat coefficient between nodes based on the attack status data, including:

[0089] ① Calculate the number of attack state types N = 120 through N = sizeof(A) + sizeof(B) + sizeof(C) + sizeof(D);

[0090] ② Initialize the attack state matrix S: In the formula, s l ∈S, l=1,2,...,N, k1=1,2,3,4,5, k2=1,2,3, k3=1,2,3,4, k4=1,2;

[0091] ③ Based on attackData statistics, the set of edges in the power information network: e ij ={i,j},e ij ∈E, when the attack result of node i pointing to node j is one of attempt, success, or failure, e ij exist;

[0092] ④Construct the network attack risk matrix G: g l =l / size(S), where g l ∈G, l=1,2,...,N;

[0093] ⑤If e ij =E.begin(), otherwise go to ⑩;

[0094] ⑥ Calculate edge e based on attackData ij The attack state probability matrix P ij : In the formula, Indicates that the attack state of node i when attacking node j is s l The number of attacks, attackNum ij represents the total number of times node i attacks node j, Indicates that the attack state of node i when attacking node j is s l The probability of

[0095] ⑦Calculate edge e ij The corresponding threat coefficient T ij :

[0096] ⑧T.push_back(i,j,T ij );

[0097] ⑨Delete the first element of E and go to ⑤;

[0098] ⑩Output the threat coefficient vector T between nodes.

[0099] The results of the threat coefficient vector T between nodes obtained in step 6 are shown in Table 5.

[0100] Attack node i Victim node j <![CDATA[Threat coefficient T ij > I16166 I5950 0.825 E3062 I6235 0.825 I12683 I5386 0.8125 I4474 I5484 0.80893 I16166 I5484 0.805472 E7025 I5386 0.803623 E7011 I5386 0.803472 E7112 I5386 0.802976 I2683 I5386 0.802083 I12788 I5386 0.802083 I12774 I5386 0.800926 I12782 I5386 0.800926 I5877 I3984 0.791833 I5933 I10743 0.791667 I5947 I10745 0.791667

[0101] Step 7: Calculate the importance of each node i in the power information network through DW-Degree centrality, including:

[0102] ① Count the number of adjacent nodes of node i, m i The calculation formula is: i represents the focal node, j represents all other nodes, N is the total number of nodes, X is the adjacency matrix, x ij is an element in the adjacency matrix X. If node i is connected

[0103] To node j, then x ij =1, otherwise x ij =0;

[0104] ② Calculate the sum of the weights of the edges connected to node i, n i The calculation formula is: Where W is the weight matrix, w ij is an element in the weight matrix W. If node i is connected to node j, then w ij > 0, otherwise w ij =0;

[0105] ③Calculate the output strength H of the node out (i) and penetration strength H in (i), the calculation formula is: λ = 1.2;

[0106] ④ Calculate the importance of each node i in the power information network. The calculation formula is: In the formula, e refers to a natural constant, whose value is approximately 2.718281828459045, μ=0.4, and v=1-μ=0.6.

[0107] The results of calculating the important index of each node i in the power information network obtained in step 7 are shown in Table 6.

[0108] Table 6 shows some results of node importance calculated based on DW-Degree.

[0109] Device node i code The importance index of node i I14070 0.998433 I4690 0.991437 I4715 0.998363 I5303 0.991316 I5379 0.999982 I5386 1 I5394 1 I5484 0.999999 I5536 0.9856 I5767 0.981034 I5791 0.970768 I5802 0.951061 I5950 0.954564 I5955 0.976754 I6005 0.99714 I6158 0.999229 I6166 0.995954 I6217 0.985788 I6246 0.993173 I66 0.999639

[0110] Step 8: Fusion node's initial risk factor and importance index to calculate the node's final risk factor, including:

[0111] ① Initialize adjustment factors α=0.1, β=0.1, max(α), max(β), delta=0.05, and maximum accuracy max(AC)=0.0. Since 0≤F(i)≤1, In order to make 0≤F'(i)≤1, max(α)=1.0, max(β)=1.0;

[0112] ② Obtain F, f through step 5 i (i,F(i),i.riskLevel,i.classLevel)∈F;

[0113] ③ Calculate the importance of each node based on the constructed directed weighted network;

[0114] ④Match node importance and generate data RI, r i (i,F(i),i.riskLevel,i.importance,i.label)∈RI;

[0115] ⑤If α≤max(α), set β=0.1, otherwise go to ⑨;

[0116] ⑥If β≥max(β), α=α+delta, go to ⑤; otherwise, update the risk factor F(i) of each node in RI: F(i) is the initial risk factor of node i obtained in step 5, is the reuse index of node i obtained in step 7. If F(i) is within the risk coefficient range corresponding to its risk level label, i.label = 1;

[0117] ⑦ Calculate the accuracy AC. If AC>max(AC), max(AC)=AC, α best =α,β best =β;

[0118] ⑧β=β+delta, transfer to ⑥;

[0119] ⑨Update RI:

[0120] ⑩ Output α best , β best , max(AC).

[0121] Through step 8, we can see that when α=0.9, β=0.1, the accuracy of risk assessment is the highest, which is 95.652%. best =0.9,β best =0.1.

Claims

1. A power information network node risk assessment method based on DW-Degree centrality, characterized in that: The following steps are involved: Step 1: Select key factors that affect the security of power information network nodes and characterize the attack status of nodes; Step 2: Based on the attack data of the power information network attack log, attackData is generated; Step 3: Extract victim IP queue IPs based on attack status data and generate training data with risk level labels; Step 4: Initialize the impact factor set: A, B, C, D; Step 5: Calculate the initial risk coefficient of the node through the node risk assessment method NRAMAS based on the attack state; the specific algorithm is: ① Calculate the number of attack state types N: N=sizeof(A)+sizeof(B)+sizeof(C)+sizeof(D); ② Initialize the attack state matrix S: In the formula, s l ∈S, l=1,2,...,N, k1=1,2,3,4,5,k2=1,2,3,k3=1,2,3,4,k4=1,2; ③Construct the network attack risk matrix G: g l =l / size(S), where g l ∈G, l=1,2,...,N, size(G)=size(S); ④If i=IPs.begin(), otherwise go to ⑩; ⑤ Calculate the attack state probability matrix P of node i based on the training data i : ⑥ Calculate the risk value F(i) of node i: ⑦If F(i) is within the risk coefficient range corresponding to its risk level label, then i.classLabel = 1, otherwise i.classLabel = 0; ⑧ Store f i , f i (i, F(i), i.riskLevel, i.classLevel) ∈ F; ⑨Delete the first element of IPs and go to ④; Output F; Step 6: Calculate the threat coefficient between nodes based on the attack status data; Step 7: Calculate the importance of the node through DW-Degree centrality; the steps are as follows: ① Count the number of adjacent nodes of node i, m i The calculation formula is: i represents the focal node, j represents all other nodes, N is the total number of nodes, X is the adjacency matrix, x ij is an element in the adjacency matrix X. If node i is connected to node j, then x ij =1, otherwise x ij =0; ② Calculate the sum of the weights of the edges connected to node i, n i The calculation formula is: Where W is the weight matrix, w ij is an element in the weight matrix W. If node i is connected to node j, then w ij >0, otherwise w ij =0; ③Calculate the node strength H out (i) and penetration strength H in (i), the calculation formula is: i is the focal node, λ is a positive adjustment parameter, and λ can be set according to research and data; when the sum of the weights of adjacent edges is the same: when 0<λ<1, the node with more adjacent nodes has a greater degree centrality, and when λ>1, the node with fewer adjacent nodes has a greater degree centrality; Calculate the importance of each node i in the power information network, the calculation formula is: Where, e refers to a natural constant, whose value is approximately 2.718281828459045, μ and ν are balance factors, μ+ν=1, when the out-degree and in-degree importance of the node are the same, μ=ν=0.5; when the out-degree importance of the node is greater than the in-degree importance, μ>0.5, ν<0.5; when the out-degree importance of the node is less than the in-degree importance, μ<0.5, ν>0.5; Step 8: Fusion node's initial risk coefficient and importance index to calculate the node's final risk coefficient; the formula is: Where F(i) is the risk coefficient of node i obtained based on NRAMAS, is the DW-Degree centrality of node i, α and β are adjustment factors, and the α and β with the highest classification accuracy can be obtained through the node risk assessment method based on DW-Degree - NRAMDD; The specific steps of the node risk assessment method based on DW-Degree - NRAMDD are as follows: ① Initialize adjustment factors α, β, max(α), max(β), delta, and maximum accuracy max(AC) ② Obtain F, f through step 5 i (i,F(i),i.riskLevel,i.classLevel)∈F; ③ Calculate the importance of each node (DW-Degree centrality) based on the constructed directed weighted network; ④Match node importance and generate data RI, r i (i,F(i),i.riskLevel,i.importance,i.label)∈RI; ⑤If α≤max(α), set β=0.1, otherwise go to ⑨; ⑥If β≥max(β), α=α+delta, go to ⑤; Otherwise, update the risk factor F(i) of each node in RI: If F(i) is within the risk coefficient range corresponding to its risk level label, i.label = 1; ⑦Calculate the accuracy AC: In the formula, |X| represents the number of samples in the dataset X, |C i | represents the number of samples of the i-th level data that are correctly classified. If AC>max(AC), max(AC)=AC, α best =α,β best =β; ⑧β=β+delta, transfer to ⑥; ⑨Update RI: outputs best ,b best ,max(AC).

2. According to claim 1, a method for risk assessment of power information network nodes based on DW-Degree centrality is characterized in that: The steps for step six are as follows: ① Calculate the number of attack state types N: N = sizeof(A) + sizeof(B) + sizeof(C) + sizeof(D); ② Initialize the attack state matrix S: In the formula, s l ∈S, l=1,2,...,N, k1=1,2,3,4,5,k2=1,2,3,k3=1,2,3,4,k4=1,2; ③ Based on attackData statistics, the set of edges in the power information network: e ij ={i,j},e ij ∈E, when the attack result of node i pointing to node j is one of attempt, success, or failure, e ij exist; ④Build the network attack risk matrix G: g l =l / size(S), where g l ∈G, l=1,2,...,N, size(G)=size(S); ⑤If eij=E.begin(), otherwise go to ⑩; ⑥ Calculate the attack state probability matrix Pij of edge eij based on attackData: In the formula, Indicates the number of attacks when node i attacks node j with the attack state sl, attackNum ij represents the total number of times node i attacks node j, represents the probability that the attack state of node i is sl when attacking node j, size(P ij )=size(G); ⑦ Calculate the threat coefficient Tij corresponding to edge eij: ⑧T.push_back(i,j,Tij); ⑨Delete the first element of E and go to ⑤; ⑩Output the threat coefficient vector T between nodes.