Artificial intelligence-based password application risk intelligent assessment method and system

Abstract

The application discloses an intelligent risk assessment method and system for password application based on artificial intelligence, relates to the technical field of password application security assessment, and comprises the following steps: password application risk assessment data is divided in multiple dimensions, data is preprocessed, and four-level labels are constructed; a two-dimensional scoring method is used to screen key nodes, extract associated edges and concatenate paths, and a knowledge graph is constructed; a differentiated collection frequency is set according to the importance level of the key nodes, abnormality is identified and early warning is performed through a normal operation threshold interval; a node importance quantitative index and an initial risk value calculation index are set, a dynamic risk value is calculated in combination with an associated edge conduction strength and a path conduction attenuation factor; four risk levels are divided, a mapping rule table is constructed, an assessment report containing a risk level, a calculation process and a conduction path is generated, and intelligent and accurate assessment and dynamic control of password application risk are realized.

Description

Technical Field

[0001] This invention relates to the field of cryptographic application security assessment technology, specifically to an artificial intelligence-based intelligent assessment method and system for cryptographic application risks. Background Technology

[0002] With the deepening of digital transformation, cryptographic applications have widely penetrated key areas such as finance, government affairs, and healthcare. Their secure and stable operation directly impacts business continuity and data security, making cryptographic application risk assessment a core element in ensuring information security. However, existing risk assessment methods have significant limitations: data segmentation dimensions are too simplistic, resulting in insufficient precision in handling outliers, missing values, and duplicate data; intelligent labeling methods are lacking; key node selection lacks scientific quantitative standards, relying on subjective judgment and insufficient correlation analysis; risk monitoring uses a uniform collection frequency, making it difficult to balance monitoring efficiency and key coverage, and anomaly warnings lack specificity; risk value calculations are mostly static assessments, failing to consider the risk transmission effect between nodes, leading to a disconnect between assessment results and actual risk situations; risk level classification is vague, and assessment reports lack dynamic adjustment basis and transmission path analysis, making it difficult to support precise control decisions. These problems result in insufficient intelligence, accuracy, and dynamism in cryptographic application risk assessment, failing to meet the risk prevention and control needs of complex business scenarios. Summary of the Invention

[0003] The purpose of this invention is to provide an intelligent assessment method and system for cryptographic application risks based on artificial intelligence, so as to solve the problems raised in the prior art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: an intelligent assessment method for cryptographic application risks based on artificial intelligence, the method comprising the following steps: Step S1: Divide the cryptographic application risk assessment into multiple dimensions, collect multi-dimensional data, and process the multi-dimensional data; Step S1-1: Divide the cryptographic application risk assessment data into multiple dimensions according to the actual business scenario; Step S1-2: Perform data processing on the segmented multi-dimensional data, specifically: for outliers, use the 3σ principle to remove outliers; for missing values, use KNN interpolation to fill in missing values; and use hash deduplication to remove duplicate data. Steps S1-3: Automatically label the data using the rule-based BERT semantic matching model to construct a four-level label system: "dimension-entity category-attribute-value".

[0005] This will enable standardized processing and systematic labeling of cryptographic application risk assessment data, improve data quality and classification accuracy, and provide reliable data support for subsequent key node screening and risk assessment.

[0006] Step S2: Select key nodes for data of different dimensions, construct related edges through key nodes, and construct a knowledge graph based on the related edges; Step S2-1: Define key nodes using the two-dimensional scoring method of "risk impact degree × business dependence degree". Based on the actual business scenario, set the risk impact degree sub-indicators as the scope of failure impact, the severity of risk consequences, and compliance risks. Set the business dependence degree sub-indicators as the core business dependence intensity, node call frequency, and availability of alternative solutions. Step S2-2: Set up a 1-5 point scoring mechanism. Professionals score the risk impact sub-indicator and business dependence sub-indicator according to the actual business scenario, assign weights to the risk impact sub-indicator and business dependence sub-indicator, collect historical data from various dimensions, calculate the two-dimensional scores of the historical data from various dimensions, sort the two-dimensional scores of the historical data from various dimensions in ascending order, construct a two-dimensional score set, and select 75% of the scores in the two-dimensional score set as the key node threshold. Step S2-3: Identify multi-dimensional data with a two-dimensional score greater than the key node threshold as key nodes, extract all key nodes, classify key nodes based on the dimension, entity category, and business scenario to which the key nodes belong, and construct a set of key nodes. Step S2-4: Extract adjacent key node pairs. According to the business process sequence, traverse all sub-steps and extract adjacent key nodes of the same category to form related edges. Step S2-5: Using the "key node-related edge-attribute" triplet architecture, connect all related edges globally according to business process logic to form a local path network. Then, integrate all local path networks to build a global path set and form a knowledge graph.

[0007] By using scientific and quantitative standards to select key nodes and building a knowledge graph based on business logic, the relationships and paths between nodes are clearly presented, laying a structured foundation for risk transmission analysis.

[0008] Step S3: Monitor the status of key nodes in real time and identify risks based on the status of key nodes; Step S3-1: Integrate the two-dimensional scores of all key nodes to construct a set of two-dimensional scores for key nodes. Sort the two-dimensional score set data in ascending order, and then use the trisection method to divide the importance level of key nodes into low, medium and high levels. Step S3-2: Set different collection frequencies t1, t2, and t3 for different importance levels of key nodes. t1 is a low collection frequency, corresponding to the collection frequency of low-level key nodes; t2 is a medium collection frequency, corresponding to the collection frequency of medium-level key nodes; and t3 is a high collection frequency, corresponding to the collection frequency of high-level key nodes. Collect the running status data of key nodes through API interface and log collection tools. Step S3-3: Collect historical data on the normal operation status of key nodes, timestamp the data on the normal operation status of key nodes, integrate the normal operation status data with the same timestamp, calculate the mean and standard deviation, set the mean as the threshold z for the normal operation status data, set the standard deviation as the fluctuation range b, and set the threshold interval for the normal operation status data as [zb, z+b]. Step S3-4: Determine whether the critical node is abnormal based on the normal operating status data threshold range. When the critical node's operating status data is not within the normal operating status data threshold range, it is determined that the critical node is abnormal. Step S3-5: When a critical node is determined to be abnormal, issue an alert to the maintenance personnel regarding the abnormal critical node, traverse the adjacent critical nodes of the abnormal critical node, generate a list of critical nodes that are likely to be at risk, and send the list of critical nodes that are likely to be at risk and the abnormal critical nodes to the professionals to remind them to repair them.

[0009] By implementing differentiated monitoring based on node importance, abnormal nodes can be accurately identified and potential risks can be warned, enabling early detection and alerts of risks and reducing the probability of risk spread.

[0010] Step S4: Quantify the importance of key nodes and calculate the initial risk value; Step S4-1: Set the quantitative indicators for the importance of key nodes. The quantitative indicators for the importance of key nodes include the two-dimensional score of the node, degree centrality, and betweenness centrality. The degree centrality is the number of associated edges. The specific calculation method is as follows: ; ; In the formula, Let v be the original centrality of the key node. Let v be the mode of the associated edges of the key node v in the undirected graph. Let v be the normalization centrality of the key node v, and n be the total number of key nodes in the knowledge graph. The betweenness centrality is the path transit frequency, and the specific calculation method is as follows: ; ; In the formula, For the target key node The original betweenness centrality, where μ and v are any two distinct key nodes in the knowledge graph. For the target critical node whose betweenness centrality is to be calculated, This represents the total number of shortest paths between key node pairs (μ, v). For node pairs (μ, v) passing through the target critical node The number of shortest paths, As a key node Standardized betweenness centrality; The weights of each quantitative indicator are assigned by the analytic hierarchy process, the corresponding indicator data of each key node are collected, and the weighted summation formula is used to calculate the comprehensive score of node importance. Step S4-2: Set the initial risk value calculation indicators, including the comprehensive score of node importance and the inherent risk sub-indicators. A 1-5 point scoring mechanism is adopted. Professionals score the inherent risk sub-indicators according to the actual business scenario. The comprehensive score of node importance and the inherent risk score are weighted and summed to obtain the initial risk value of each node. Step S4-3: Set the knowledge graph path transmission parameters, including the transmission strength of associated edges and the path transmission attenuation factor. Traverse all transmission paths of abnormal nodes in the knowledge graph. Based on the initial risk value, transmission strength of associated edges, and transmission attenuation factor of each node on the path, calculate the risk transmission adjustment amount for each key node. Specifically: Step S4-3-1: Traverse all reachable paths of the abnormal node in the knowledge graph. Each path starts with the abnormal node and ends with the target node. Record the node sequence and corresponding associated edges on the path. Define the length of the single-player path as the number of associated edges in the path. Step S4-3-2: The source of the associated edge propagation strength is the associated edges between adjacent nodes in the path, based on the business dependency score set in step S2-1, with a value range of [x1, x2]. The source of the path propagation attenuation factor is linear attenuation with increasing path length, and the attenuation formula is... When the calculation result When ≤0.1, take the uniform value. =0.1; Step S4-3-3: Calculate the conduction volume of a single path, specifically as follows: ; In the formula, Let $\frac{k}{k}$ be the single-path risk transmission amount obtained by the target critical node through the $k$-th path. This represents the initial risk value for the abnormal node. For path conduction attenuation factor, This represents the product of the propagation strength of the edges connecting all adjacent nodes on the k-th path; Step S4-3-4: Calculate the risk transmission adjustment amount for key nodes, specifically as follows; ; In the formula, This represents the total risk transmission adjustment obtained by the target critical node from the abnormal critical node. This is a summation operator, indicating that the results from the first path to the second path are accumulated, where k is the path index and m is the total number of reachable paths from the target critical node to the abnormal critical node. The single-path risk transmission volume obtained by the target critical node through the k-th path; The node dynamic risk value is obtained by adding the initial risk value to the risk transmission adjustment.

[0011] By quantifying the importance of nodes from multiple dimensions and calculating dynamic risk values ​​in conjunction with risk transmission effects, we can overcome the limitations of static assessment and make the risk quantification results more in line with the actual risk situation.

[0012] Step S5: Dynamically adjust the risk value based on the knowledge graph path, map the risk value to the risk level, and construct a risk assessment report.

[0013] Step S5-1: Divide the risk level into four levels: low risk, medium risk, high risk, and extremely high risk. Collect historical risk handling data and, in conjunction with industry risk management standards, set the dynamic risk value range corresponding to each risk level and construct a "dynamic risk value - risk level" mapping rule table. Step S5-2: Substitute the dynamic risk value of each node into the mapping rule table, match the corresponding risk level, generate a risk level list for each node, integrate the risk level list, the dynamic risk value calculation process and the transmission path analysis results to form a risk assessment report, and send it to professionals.

[0014] Establish clear risk level mapping rules and generate a complete report that includes risk level, calculation process and transmission path, providing professionals with accurate and actionable risk management basis.

[0015] The system includes a data tagging module, a node graph module, a monitoring and early warning module, a risk quantification module, and a level reporting module; The data tagging module is used for multi-dimensional data processing and four-level tag construction; The node graph module is used for key node screening and knowledge graph construction based on associated edges; The monitoring and early warning module is used for monitoring the status of key nodes and providing early warning of anomalies. The risk quantification module is used for quantifying the importance of key nodes and calculating dynamic risk values; The rating report module is used to dynamically map risk values ​​to ratings and generate assessment reports.

[0016] The data tagging module includes a data processing unit and a tag building unit; The data processing unit is used to divide data according to business scenarios and process abnormal, missing, and duplicate data. The label building unit is used to automatically label four levels of tags using rules and the BERT model; The node graph module includes a node filtering unit and a graph construction unit; The node filtering unit is used to define and filter key nodes using a two-dimensional scoring method. The graph construction unit is used to extract related edges, connect paths, and integrate them to form a knowledge graph.

[0017] The monitoring and early warning module includes a status acquisition unit and an anomaly early warning unit; The status acquisition unit is used to classify the importance level of nodes and collect data at the corresponding frequency; The anomaly warning unit is used to set a normal threshold to determine anomalies and push warning information; The risk quantification module includes a quantification calculation unit and a transmission adjustment unit; The quantitative calculation unit is used to set the importance score and initial risk value of the indicator calculation node; The transmission adjustment unit is used to calculate the adjustment amount based on the transmission parameters to obtain the dynamic risk value.

[0018] The rating report module includes a rating mapping unit and a report generation unit; The level mapping unit is used to construct the mapping rules between dynamic risk values ​​and risk levels; The report generation unit is used to match risk level data and integrate it to form an assessment report.

[0019] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention significantly improves the quality of assessment data and the accuracy of classification by standardizing data processing in multiple dimensions and intelligently constructing four-level labels, providing systematic and highly reliable data support for the entire subsequent risk assessment process.

[0020] 2. This invention relies on a differentiated monitoring mechanism to achieve early detection of abnormal risks. It combines node correlation and transmission effect to calculate dynamic risk values, breaking through the limitations of traditional static assessment and improving the pertinence of risk identification and the consistency of quantitative results.

[0021] 3. This invention clearly presents the risk situation through scientific risk level mapping and an assessment report that includes transmission paths and calculation processes, providing accurate and practical decision-making basis for risk management in cryptographic applications and helping to achieve efficient prevention and control. Attached Figure Description

[0022] Figure 1 This is a flowchart illustrating the intelligent risk assessment method for cryptographic applications based on artificial intelligence, as described in this invention. Figure 2 This is a schematic diagram of the structure of the AI-based intelligent risk assessment system for cryptographic applications according to the present invention. Detailed Implementation

[0023] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] Example 1: As Figures 1-2 As shown, this invention provides a technical solution: an intelligent assessment method for cryptographic application risks based on artificial intelligence. The method includes the following steps: Step S1: Divide the cryptographic application risk assessment into multiple dimensions, collect multi-dimensional data, and process the multi-dimensional data; Step S1-1: Divide the cryptographic application risk assessment data into multiple dimensions according to the actual business scenario; Step S1-2: Perform data processing on the segmented multi-dimensional data, specifically: for outliers, use the 3σ principle to remove outliers; for missing values, use KNN interpolation to fill in missing values; and use hash deduplication to remove duplicate data. Steps S1-3: Automatically label the data using the rule-based BERT semantic matching model to construct a four-level label system: "dimension-entity category-attribute-value".

[0025] This will enable standardized processing and systematic labeling of cryptographic application risk assessment data, improve data quality and classification accuracy, and provide reliable data support for subsequent key node screening and risk assessment.

[0026] Step S2: Select key nodes for data of different dimensions, construct related edges through key nodes, and construct a knowledge graph based on the related edges; Step S2-1: Define key nodes using the two-dimensional scoring method of "risk impact degree × business dependence degree". Based on the actual business scenario, set the risk impact degree sub-indicators as the scope of failure impact, the severity of risk consequences, and compliance risks. Set the business dependence degree sub-indicators as the core business dependence intensity, node call frequency, and availability of alternative solutions. Step S2-2: Set up a 1-5 point scoring mechanism. Professionals score the risk impact sub-indicator and business dependence sub-indicator according to the actual business scenario, assign weights to the risk impact sub-indicator and business dependence sub-indicator, collect historical data from various dimensions, calculate the two-dimensional scores of the historical data from various dimensions, sort the two-dimensional scores of the historical data from various dimensions in ascending order, construct a two-dimensional score set, and select 75% of the scores in the two-dimensional score set as the key node threshold. Step S2-3: Identify multi-dimensional data with a two-dimensional score greater than the key node threshold as key nodes, extract all key nodes, classify key nodes based on the dimension, entity category, and business scenario to which the key nodes belong, and construct a set of key nodes. Step S2-4: Extract adjacent key node pairs. According to the business process sequence, traverse all sub-steps and extract adjacent key nodes of the same category to form related edges. Step S2-5: Using the "key node-related edge-attribute" triplet architecture, connect all related edges globally according to business process logic to form a local path network. Then, integrate all local path networks to build a global path set and form a knowledge graph.

[0027] By using scientific and quantitative standards to select key nodes and building a knowledge graph based on business logic, the relationships and paths between nodes are clearly presented, laying a structured foundation for risk transmission analysis.

[0028] Step S3: Monitor the status of key nodes in real time and identify risks based on the status of key nodes; Step S3-1: Integrate the two-dimensional scores of all key nodes to construct a set of two-dimensional scores for key nodes. Sort the two-dimensional score set data in ascending order, and then use the trisection method to divide the importance level of key nodes into low, medium and high levels. Step S3-2: Set different collection frequencies t1, t2, and t3 for different importance levels of key nodes. t1 is a low collection frequency, corresponding to the collection frequency of low-level key nodes; t2 is a medium collection frequency, corresponding to the collection frequency of medium-level key nodes; and t3 is a high collection frequency, corresponding to the collection frequency of high-level key nodes. Collect the running status data of key nodes through API interface and log collection tools. Step S3-3: Collect historical data on the normal operation status of key nodes, timestamp the data on the normal operation status of key nodes, integrate the normal operation status data with the same timestamp, calculate the mean and standard deviation, set the mean as the threshold z for the normal operation status data, set the standard deviation as the fluctuation range b, and set the threshold interval for the normal operation status data as [zb, z+b]. Step S3-4: Determine whether the critical node is abnormal based on the normal operating status data threshold range. When the critical node's operating status data is not within the normal operating status data threshold range, it is determined that the critical node is abnormal. Step S3-5: When a critical node is determined to be abnormal, issue an alert to the maintenance personnel regarding the abnormal critical node, traverse the adjacent critical nodes of the abnormal critical node, generate a list of critical nodes that are likely to be at risk, and send the list of critical nodes that are likely to be at risk and the abnormal critical nodes to the professionals to remind them to repair them.

[0029] By implementing differentiated monitoring based on node importance, abnormal nodes can be accurately identified and potential risks can be warned, enabling early detection and alerts of risks and reducing the probability of risk spread.

[0030] Step S4: Quantify the importance of key nodes and calculate the initial risk value; Step S4-1: Set the quantitative indicators for the importance of key nodes. The quantitative indicators for the importance of key nodes include the two-dimensional score of the node, degree centrality, and betweenness centrality. The degree centrality is the number of associated edges. The specific calculation method is as follows: ; ; In the formula, Let v be the original centrality of the key node. Let v be the mode of the associated edges of the key node v in the undirected graph. Let v be the normalization centrality of the key node v, and n be the total number of key nodes in the knowledge graph. The betweenness centrality is the path transit frequency, and the specific calculation method is as follows: ; ; In the formula, For the target key node The original betweenness centrality, where μ and v are any two distinct key nodes in the knowledge graph. For the target critical node whose betweenness centrality is to be calculated, This represents the total number of shortest paths between key node pairs (μ, v). For node pairs (μ, v) passing through the target critical node The number of shortest paths, As a key node Standardized betweenness centrality; The weights of each quantitative indicator are assigned by the analytic hierarchy process, the corresponding indicator data of each key node are collected, and the weighted summation formula is used to calculate the comprehensive score of node importance. Step S4-2: Set the initial risk value calculation indicators, including the comprehensive score of node importance and the inherent risk sub-indicators. A 1-5 point scoring mechanism is adopted. Professionals score the inherent risk sub-indicators according to the actual business scenario. The comprehensive score of node importance and the inherent risk score are weighted and summed to obtain the initial risk value of each node. Step S4-3: Set the knowledge graph path transmission parameters, including the transmission strength of associated edges and the path transmission attenuation factor. Traverse all transmission paths of abnormal nodes in the knowledge graph. Based on the initial risk value, transmission strength of associated edges, and transmission attenuation factor of each node on the path, calculate the risk transmission adjustment amount for each key node. Specifically: Step S4-3-1: Traverse all reachable paths of the abnormal node in the knowledge graph. Each path starts with the abnormal node and ends with the target node. Record the node sequence and corresponding associated edges on the path. Define the length of the single-player path as the number of associated edges in the path. Step S4-3-2: The source of the associated edge propagation strength is the associated edges between adjacent nodes in the path, based on the business dependency score set in step S2-1, with a value range of [x1, x2]. The source of the path propagation attenuation factor is linear attenuation with increasing path length, and the attenuation formula is... When the calculation result When ≤0.1, take the uniform value. =0.1; Step S4-3-3: Calculate the conduction volume of a single path, specifically as follows: ; In the formula, Let $\frac{k}{k}$ be the single-path risk transmission amount obtained by the target critical node through the $k$-th path. This represents the initial risk value for the abnormal node. For path conduction attenuation factor, This represents the product of the propagation strength of the edges connecting all adjacent nodes on the k-th path; Step S4-3-4: Calculate the risk transmission adjustment amount for key nodes, specifically as follows; ; In the formula, This represents the total risk transmission adjustment obtained by the target critical node from the abnormal critical node. This is a summation operator, indicating that the results from the first path to the second path are accumulated, where k is the path index and m is the total number of reachable paths from the target critical node to the abnormal critical node. The single-path risk transmission volume obtained by the target critical node through the k-th path; The node dynamic risk value is obtained by adding the initial risk value to the risk transmission adjustment.

[0031] By quantifying the importance of nodes from multiple dimensions and calculating dynamic risk values ​​in conjunction with risk transmission effects, we can overcome the limitations of static assessment and make the risk quantification results more in line with the actual risk situation.

[0032] Step S5: Dynamically adjust the risk value based on the knowledge graph path, map the risk value to the risk level, and construct a risk assessment report.

[0033] Step S5-1: Divide the risk level into four levels: low risk, medium risk, high risk, and extremely high risk. Collect historical risk handling data and, in conjunction with industry risk management standards, set the dynamic risk value range corresponding to each risk level and construct a "dynamic risk value - risk level" mapping rule table. Step S5-2: Substitute the dynamic risk value of each node into the mapping rule table, match the corresponding risk level, generate a risk level list for each node, integrate the risk level list, the dynamic risk value calculation process and the transmission path analysis results to form a risk assessment report, and send it to professionals.

[0034] Establish clear risk level mapping rules and generate a complete report that includes risk level, calculation process and transmission path, providing professionals with accurate and actionable risk management basis.

[0035] The system includes a data tagging module, a node graph module, a monitoring and early warning module, a risk quantification module, and a level reporting module; The data tagging module is used for multi-dimensional data processing and four-level tag construction; The node graph module is used for key node screening and knowledge graph construction based on associated edges; The monitoring and early warning module is used for monitoring the status of key nodes and providing early warning of anomalies. The risk quantification module is used for quantifying the importance of key nodes and calculating dynamic risk values; The rating report module is used to dynamically map risk values ​​to ratings and generate assessment reports.

[0036] The data tagging module includes a data processing unit and a tag building unit; The data processing unit is used to divide data according to business scenarios and process abnormal, missing, and duplicate data. The label building unit is used to automatically label four levels of tags using rules and the BERT model; The node graph module includes a node filtering unit and a graph construction unit; The node filtering unit is used to define and filter key nodes using a two-dimensional scoring method. The graph construction unit is used to extract related edges, connect paths, and integrate them to form a knowledge graph.

[0037] The monitoring and early warning module includes a status acquisition unit and an anomaly early warning unit; The status acquisition unit is used to classify the importance level of nodes and collect data at the corresponding frequency; The anomaly warning unit is used to set a normal threshold to determine anomalies and push warning information; The risk quantification module includes a quantification calculation unit and a transmission adjustment unit; The quantitative calculation unit is used to set the importance score and initial risk value of the indicator calculation node; The transmission adjustment unit is used to calculate the adjustment amount based on the transmission parameters to obtain the dynamic risk value.

[0038] The rating report module includes a rating mapping unit and a report generation unit; The level mapping unit is used to construct the mapping rules between dynamic risk values ​​and risk levels; The report generation unit is used to match risk level data and integrate it to form an assessment report.

[0039] Example 2: The analytic hierarchy process (AHP) was used to determine the two-dimensional scoring weight of nodes as 0.4, the degree centrality weight as 0.3, and the betweenness centrality weight as 0.3. The two-dimensional scores for key nodes are A=4 points, B=3.5 points, C=3.8 points, and D=3 points; If A is associated with B, C, and D, then d(v) = 3; if B is associated with A and C, then d(v) = 2; if C is associated with A, B, and D, then d(v) = 3; if D is associated with A and C, then d(v) = 2. Standardization centrality: A=1, B≈0.67, C=1.0; Betweenness centrality: When calculating node A, all node pairs (μ,v) are (B,C), (B,D), and (C,D). The total number of shortest paths is 1 and all of them pass through A. Therefore, σ(μ,v|A)=0.5. Similarly, we can get B=0, C=0.5, and D=0. The overall score for the importance of the compute nodes is 1.401; Inherent risk sub-indicator scores: A (password storage) = 4 points, B (password transmission) = 3 points, C (password verification) = 3.5 points, D (password reset) = 2.5 points; Calculate the initial risk value: A = 2.05 × 0.5 + 4 × 0.5 = 1.025 + 2 = 3.025; B=1.601×0.5+3×0.5=0.8005+1.5=2.3005; C=1.97×0.5+3.5×0.5=0.985+1.75=2.735; D=1.401×0.5+2.5×0.5=0.7005+1.25=1.9505; The range of the associated edge conduction strength is [0.7, 1.0]. Based on the business dependency rating settings: AB=0.9, AC=0.95, AD=0.8, BC=0.85; The path conduction attenuation factor formula α(L) = 0.8^(L-1) ≤ 0.1 is taken as 0.1; Traverse the reachable paths of the abnormal node A: Path 1: A→B (L=1), single-path conduction ΔRnk=2.7225; Path 2: A→C (L=1), single-path conduction ΔRnk=2.87375; Path 3: A→D (L=1), single-path conduction ΔRnk=2.42; Path 4: A→B→C (L=2), single-path conduction ΔRnk≈1.842; Calculate the risk transmission adjustment for each node: Node B: Reachable only by path 1, ΔRn=2.7225; Node C: Path 2 + Path 4, ΔRn≈4.71575; Node D: Reachable only by path 3, ΔRn=2.42; Calculate the dynamic risk value: B = 2.3005 + 2.7225 = 5.023; C = 2.735 + 4.71575 ≈ 7.45075; D = 1.9505 + 2.42 = 4.3705.

[0040] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. An intelligent risk assessment method for cryptographic applications based on artificial intelligence, characterized by: The method includes the following steps: Step S1: Divide the password application risk assessment data into multiple dimensions according to the actual business scenario, collect multi-dimensional data, and process the multi-dimensional data; The specific steps of step S1 are as follows: Step S1-1: Divide the cryptographic application risk assessment data into multiple dimensions according to the actual business scenario; Step S1-2: Perform data processing on the segmented multi-dimensional data, specifically: remove outliers, use KNN interpolation to fill in missing values, and use hash deduplication to remove duplicate data. Steps S1-3: Automatically label the data using the rule-based BERT semantic matching model, constructing a four-level label system of "dimension-entity category-attribute-value". Step S2: Select key nodes for data of different dimensions, construct related edges through key nodes, and construct a knowledge graph based on the related edges; The specific steps of step S2 are as follows: Step S2-1: Define key nodes using the two-dimensional scoring method of "risk impact degree × business dependence degree". Based on the actual business scenario, set the risk impact degree sub-indicators as the scope of failure impact, the severity of risk consequences, and compliance risks. Set the business dependence degree sub-indicators as the core business dependence intensity, node call frequency, and availability of alternative solutions. Step S2-2: Set up a 1-5 point scoring mechanism. Professionals score the risk impact sub-indicator and business dependence sub-indicator according to the actual business scenario, assign weights to the risk impact sub-indicator and business dependence sub-indicator, collect historical data from various dimensions, calculate the two-dimensional scores of the historical data from various dimensions, sort the two-dimensional scores of the historical data from various dimensions in ascending order, construct a two-dimensional score set, and select the 75th percentile in the two-dimensional score set as the key node threshold. Step S2-3: Determine the nodes corresponding to the multi-dimensional data with two-dimensional scores greater than the key node threshold as key nodes, extract all key nodes, classify the key nodes based on the dimension, entity category, and business scenario to which the key nodes belong, and construct a set of key nodes. Step S2-4: Extract adjacent key node pairs. According to the business process sequence, traverse all sub-steps and extract adjacent key nodes of the same category to form related edges. Step S2-5: Using the "key node-related edge-attribute" triplet architecture, connect all related edges globally according to the business process logic to form a local path network. Then, integrate all local path networks to build a global path set and form a knowledge graph. Step S3: Monitor the status of key nodes in real time and identify risks based on the status of key nodes; The specific steps of step S3 are as follows: Step S3-1: Integrate the two-dimensional scores of all key nodes to construct a set of two-dimensional scores for key nodes. Sort the two-dimensional score set data in ascending order, and then use the trisection method to divide the importance level of key nodes into low, medium and high levels. Step S3-2: Set different collection frequencies t1, t2, and t3 for different importance levels of key nodes. t1 is a low collection frequency, corresponding to the collection frequency of low-level key nodes; t2 is a medium collection frequency, corresponding to the collection frequency of medium-level key nodes; and t3 is a high collection frequency, corresponding to the collection frequency of high-level key nodes. Collect the running status data of key nodes through API interface and log collection tools. Step S3-3: Collect historical data on the normal operation status of key nodes, timestamp the data on the normal operation status of key nodes, integrate the normal operation status data with the same timestamp, calculate the mean and standard deviation, set the mean as the threshold z for the normal operation status data, set the standard deviation as the fluctuation range b, and set the threshold interval for the normal operation status data as [zb, z+b]. Step S3-4: Determine whether the critical node is abnormal based on the normal operating status data threshold range. When the critical node's operating status data is not within the normal operating status data threshold range, it is determined that the critical node is abnormal. Step S3-5: When a critical node is determined to be abnormal, issue an alert to the operation and maintenance personnel regarding the abnormal critical node, traverse the adjacent critical nodes of the abnormal critical node, generate a list of critical nodes that are likely to be at risk, and send the list of critical nodes that are likely to be at risk and the abnormal critical nodes to the professionals to remind them to repair them. Step S4: Quantify the importance of key nodes and calculate the initial risk value; The specific steps of step S4 are as follows: Step S4-1: Set the quantitative indicators for the importance of key nodes. The quantitative indicators for the importance of key nodes include the two-dimensional score of the node, degree centrality, and betweenness centrality. The degree centrality is the number of associated edges. The specific calculation method is as follows: ； ； In the formula, Let v be the original centrality of the key node. Let v be the number of associated edges of the critical node v in the undirected graph. Let v be the normalization centrality of the key node, and n be the total number of key nodes in the knowledge graph. The betweenness centrality is the path transit frequency, and the specific calculation method is as follows: ； ； In the formula, For the target key node The original betweenness centrality, where μ and v are any two distinct key nodes in the knowledge graph. For the target critical node whose betweenness centrality is to be calculated, This represents the total number of shortest paths between key node pairs (μ, v). For node pairs (μ, v) passing through the target critical node The number of shortest paths, As a key node Standardized betweenness centrality; The weights of each quantitative indicator are assigned using the analytic hierarchy process (AHP), corresponding indicator data for each key node are collected, and a weighted summation formula is used to calculate the overall importance score of the node. Step S4-2: Set the initial risk value calculation indicators, including the comprehensive score of node importance and the inherent risk sub-indicators. A 1-5 point scoring mechanism is adopted. Professionals score the inherent risk sub-indicators according to the actual business scenario. The comprehensive score of node importance and the inherent risk score are weighted and summed to obtain the initial risk value of each node. Step S4-3: Set the knowledge graph path transmission parameters, including the transmission strength of associated edges and the path transmission attenuation factor. Traverse all transmission paths of abnormal nodes in the knowledge graph. Based on the initial risk value, transmission strength of associated edges, and transmission attenuation factor of each node on the path, calculate the risk transmission adjustment amount for each key node. Specifically: ； In the formula, This represents the total risk transmission adjustment obtained by the target critical node from the abnormal critical node. This is a summation operator, indicating that the results from the first path to the second path are accumulated, where k is the path index and m is the total number of reachable paths from the target critical node to the abnormal critical node. The single-path risk transmission volume obtained by the target critical node through the k-th path; The node dynamic risk value is obtained by adding the initial risk value to the total risk transmission adjustment. In step S4-3, the source of the associated edge propagation strength is the associated edge between adjacent nodes in the path, based on the business dependency score set in step S2-1, with a set value range of [x1, x2]. The source of the path propagation attenuation factor is linear attenuation with increasing path length, and the attenuation formula is... When the calculation result When ≤0.1, take the uniform value. =0.1; The calculation of single-path risk transmission is as follows: ； In the formula, Let $\frac{k}{k}$ be the single-path risk transmission amount obtained by the target critical node through the $k$-th path. This represents the initial risk value for the abnormal node. For path conduction attenuation factor, This represents the product of the propagation strength of the edges connecting all adjacent nodes on the k-th path; Step S5: Dynamically adjust the risk value based on the knowledge graph path, map the risk value to the risk level, and construct a risk assessment report; The specific steps of step S5 are as follows: Step S5-1: Divide the risk level into four levels: low risk, medium risk, high risk, and extremely high risk. Collect historical risk handling data and, in conjunction with industry risk management standards, set the dynamic risk value range corresponding to each risk level and construct a "dynamic risk value - risk level" mapping rule table. Step S5-2: Substitute the dynamic risk value of each node into the mapping rule table, match the corresponding risk level, generate a risk level list for each node, integrate the risk level list, the dynamic risk value calculation process and the transmission path analysis results to form a risk assessment report, and send it to professionals.

2. An AI-based intelligent risk assessment system for cryptographic applications, used to execute the AI-based intelligent risk assessment method for cryptographic applications as described in claim 1, characterized in that: The system includes a data tagging module, a node graph module, a monitoring and early warning module, a risk quantification module, and a level reporting module; The data tagging module is used for multi-dimensional data processing and four-level tag construction; The data tagging module includes a data processing unit and a tag building unit; The data processing unit is used to divide data according to business scenarios and process abnormal, missing, and duplicate data. The label building unit is used to automatically label four levels of tags using rules and the BERT model; The node graph module includes a node filtering unit and a graph construction unit; The node filtering unit is used to define and filter key nodes using a two-dimensional scoring method. The graph construction unit is used to extract related edges, connect paths, and integrate them to form a knowledge graph; The node graph module is used for key node screening and knowledge graph construction based on associated edges; The monitoring and early warning module is used for monitoring the status of key nodes and providing early warning of anomalies. The monitoring and early warning module includes a status acquisition unit and an anomaly early warning unit; The status acquisition unit is used to classify the importance level of nodes and collect data at the corresponding frequency; The anomaly warning unit is used to set a normal threshold to determine anomalies and push warning information; The risk quantification module includes a quantification calculation unit and a transmission adjustment unit; The quantitative calculation unit is used to set the importance score and initial risk value of the indicator calculation node; The transmission adjustment unit is used to calculate the adjustment amount based on the transmission parameters to obtain a dynamic risk value; The risk quantification module is used for quantifying the importance of key nodes and calculating dynamic risk values; The risk level reporting module is used to dynamically map risk values ​​to risk levels and generate assessment reports. The rating report module includes a rating mapping unit and a report generation unit; The level mapping unit is used to construct the mapping rules between dynamic risk values ​​and risk levels; The report generation unit is used to match risk level data and generate an assessment report.

Images