Open source component security vulnerability processing method and system

Abstract

The invention relates to the technical field of computers, in particular to an open source component security vulnerability processing method and system.The open source component information and vulnerability information are collected from a project code warehouse and a vulnerability disclosure platform and subjected to standardization processing, multi-dimensional atlas generation and cross correlation analysis, and a dependency graph is generated and dynamically optimized; a distributed Bayesian reasoning and co-evolution model is used for carrying out multi-level data fusion analysis, finally, future vulnerability outbreak is predicted through a space-time regression model and a cross-dimension joint prediction model, and a staged preventive repair strategy is formulated and executed. According to the method, the accuracy and timeliness of vulnerability processing are remarkably improved, and the security risk is effectively reduced.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a method and system for processing security vulnerabilities of open source components. Background Art

[0002] Open source components play a key role in modern software development. Their openness and diversity enable developers to quickly build and expand applications. However, with the widespread use of open source components, the risk of their security vulnerabilities has also increased. The existing open source component security vulnerability processing method (Chinese invention patent, publication number: CN116305167A, name: Open source component security vulnerability processing method and device) mainly relies on periodic vulnerability scanning and manual confirmation. There are multiple problems with this approach: on the one hand, vulnerability scanning tools can only find known vulnerabilities and cannot deal with emerging potential threats; on the other hand, the manual confirmation and processing process is time-consuming and error-prone, especially when faced with complex dependencies and multi-level components. It is difficult to accurately judge and prioritize the repair of the most critical vulnerabilities. These deficiencies make the existing technology inefficient in dealing with open source component security vulnerabilities and unable to fully protect the security of software systems. Summary of the invention

[0003] In view of the many problems existing in the above-mentioned prior art, the present invention provides a method and system for processing security vulnerabilities of open source components. The present invention realizes efficient processing of security vulnerabilities of open source components through technical means such as multi-dimensional graph generation and cross-correlation analysis, adaptive evolutionary algorithm, distributed Bayesian reasoning and co-evolution model. Through these technical means, dependency graphs can be dynamically generated and optimized, and real-time vulnerability prediction and repair strategy formulation can be carried out on this basis, which greatly improves the accuracy and efficiency of vulnerability processing.

[0004] A method for handling open source component security vulnerabilities includes the following steps:

[0005] By collecting open source component information and vulnerability information from project code repositories and vulnerability disclosure platforms, the open source component information and vulnerability information are standardized and preliminary correlation data is generated;

[0006] Through multi-dimensional graph generation and cross-correlation analysis, multi-dimensional dependency graphs are generated based on preliminary correlation data, and the dependency graphs are dynamically evolved and optimized through adaptive evolutionary algorithms to generate verification dependency graphs. The verification dependency graphs are updated in real time through a cross-graph feedback mechanism.

[0007] Based on the verification dependency graph, distributed Bayesian reasoning and co-evolution are performed. Intelligent agents reason and evolve data at different levels, and combine multi-level data fusion algorithms to deeply fuse data of different dimensions. Through the cross-layer multi-agent collaborative game model, strategies are optimized and shared to generate fusion analysis data and game feedback data.

[0008] Based on the fusion analysis data, the time point and spatial location of future vulnerability outbreaks are predicted through the spatiotemporal regression model and the cross-dimensional joint prediction model, and cross-dimensional spatiotemporal prediction data is generated. A preventive repair strategy with multiple repair stages is formulated, and repair operations are performed in stages. Repair feedback data is generated through real-time monitoring;

[0009] Based on cross-dimensional spatiotemporal prediction data and repair feedback data, global optimization decisions are made, and system optimization and adaptive learning are performed through feedback mechanisms to ultimately generate evolutionary learning model data.

[0010] Preferably, the multi-dimensional graph generation and cross-correlation analysis step includes: generating a static dependency graph, a dynamic dependency graph, a similarity graph and a vulnerability propagation graph, and performing cross-correlation analysis on the static dependency graph, the dynamic dependency graph, the similarity graph and the vulnerability propagation graph by using a multi-layer network model to generate multi-dimensional cross-correlation data, wherein the multi-layer network model determines the mutual influence between the graphs by calculating the relationship weights between the nodes in each graph, and uniformly represents the dependency relationships of different dimensions, and the relationship weights W ij The calculation formula is:

[0011]

[0012] Among them, W ij represents the relationship weight between node i and node j; C ij represents the interaction frequency between node i and node j; T i Represent the total active time of node i and node j respectively.

[0013] Preferably, the adaptive evolutionary algorithm generates an evolutionary dependency graph by dynamically adjusting the relationship between node weights and edges in the multidimensional cross-correlation data. The evolutionary dependency graph is updated in real time through a cross-graph feedback mechanism. The cross-graph feedback mechanism is based on the dynamic interaction information between multi-dimensional graphs, and finally generates a verification dependency graph. The update formula of the adaptive evolutionary algorithm is:

[0014] E(t+1)=α×E(t)+β×ΔE

[0015] Where E(t) represents the node weight at time t; ΔE represents the weight increment at time t; α and β are the evolution coefficients respectively.

[0016] Preferably, the distributed Bayesian reasoning and co-evolution step includes: independently performing data reasoning at different levels through multiple intelligent agents, each intelligent agent performs Bayesian reasoning based on the data at its level, and after the reasoning is completed, the reasoning results are shared through the co-evolution model to generate co-evolution reasoning data, and the co-evolution reasoning data is used to update the reasoning models at different levels in real time. The calculation formula of the Bayesian reasoning is:

[0017]

[0018] Among them, P(H|D) represents the posterior probability of hypothesis H being true under data D; P(D|H) represents the likelihood of observing data D when hypothesis H is true; P(H) represents the prior probability of hypothesis H; and P(D) represents the marginal probability of data D.

[0019] Preferably, a multi-level data fusion algorithm is used to deeply fuse the co-evolutionary reasoning data with the data of different dimensions in the verification dependency graph to generate fusion analysis data, and the multi-level data fusion algorithm processes data features of different dimensions in a hierarchical manner.

[0020] Preferably, the cross-layer multi-agent collaborative game model optimizes strategy sharing between intelligent agents through game analysis. The game analysis generates game feedback data based on the strategy execution results of different intelligent agents at each level. The game feedback data is used to further adjust and optimize the fusion analysis data.

[0021] Preferably, a time series analysis of historical system operation data is performed through a space-time regression model to generate space-time regression prediction data, which predicts the time point and spatial location of future vulnerability outbreaks based on the time and space characteristics of the historical system operation data.

[0022] Preferably, the cross-dimensional joint prediction model generates cross-dimensional spatiotemporal prediction data by jointly analyzing the spatiotemporal regression prediction data and the game feedback data. The cross-dimensional joint prediction model generates cross-dimensional spatiotemporal prediction results by combining multi-dimensional data of time, space, dependency and strategy feedback, and formulates a preventive repair strategy including multiple repair stages based on the cross-dimensional spatiotemporal prediction data.

[0023] Preferably, the cross-dimensional spatiotemporal prediction data and the repair feedback data are trained and optimized by an adaptive learning algorithm to generate evolutionary learning model data, and the adaptive learning algorithm continuously adjusts the decision model parameters based on historical decision data and real-time feedback data.

[0024] A system for implementing the open source component security vulnerability processing method, comprising:

[0025] The information collection and preprocessing module is used to collect open source component information and vulnerability information from the project code repository and vulnerability disclosure platform, and to standardize the collected open source component information and vulnerability information to generate preliminary correlation data;

[0026] Multi-dimensional graph generation and analysis module, used to generate multi-dimensional dependency graphs based on preliminary correlation data, generate verification dependency graphs through multi-dimensional graph generation and cross-correlation analysis, and dynamically evolve and optimize the dependency graph through an adaptive evolutionary algorithm. The verification dependency graph is updated in real time through a cross-graph feedback mechanism;

[0027] The collaborative reasoning and data fusion module is used to perform distributed Bayesian reasoning and collaborative evolution based on the verification dependency graph. It uses intelligent agents to perform data reasoning and evolution at different levels, combines multi-level data fusion algorithms to deeply fuse data of different dimensions, generates fusion analysis data and game feedback data, and optimizes and shares strategies through a cross-layer multi-agent collaborative game model.

[0028] The spatiotemporal prediction and repair strategy module is used to predict the time point and spatial location of future vulnerability outbreaks based on fusion analysis data through spatiotemporal regression models and cross-dimensional joint prediction models, generate cross-dimensional spatiotemporal prediction data, formulate preventive repair strategies including multiple repair stages, perform staged repair operations, and generate repair feedback data through real-time monitoring;

[0029] The global optimization and learning module is used to make global optimization decisions based on cross-dimensional spatiotemporal prediction data and repair feedback data, and to perform system optimization and adaptive learning through feedback mechanisms, ultimately generating evolutionary learning model data.

[0030] Compared with the prior art, the advantages and beneficial effects of the present invention are:

[0031] The present invention realizes comprehensive analysis and dynamic optimization of multi-dimensional dependencies of open source components through multi-dimensional graph generation and cross-correlation analysis technology, solving the problem of low efficiency of existing technologies in processing complex dependencies.

[0032] The present invention uses an adaptive evolutionary algorithm and a cross-graph feedback mechanism to achieve real-time updating and optimization of dependency graphs, significantly improving the accuracy and timeliness of vulnerability repairs.

[0033] The application of the distributed Bayesian reasoning and co-evolution method of the present invention ensures the high efficiency of data reasoning and fusion analysis at different levels, further reduces the need for manual intervention, and improves the security and reliability of the overall system. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1It is a schematic diagram of the process of the present invention;

[0035] Figure 2 It is a schematic diagram of multi-dimensional spectrum generation and cross-correlation analysis in the present invention;

[0036] Figure 3 A schematic diagram of the adaptive evolutionary algorithm and real-time feedback mechanism in the present invention;

[0037] Figure 4 It is a structural block diagram of the system of the present invention. DETAILED DESCRIPTION

[0038] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the present disclosure. In the following detailed description, for ease of explanation, many specific details are set forth to provide a comprehensive understanding of the embodiments of the present disclosure. However, it is obvious that one or more embodiments may also be implemented without these specific details. In addition, in the following description, descriptions of known structures and technologies are omitted to avoid unnecessary confusion of the concepts of the present disclosure.

[0039] The terms used herein are only for describing specific embodiments and are not intended to limit the present disclosure. The terms "comprise", "include", etc. used herein indicate the existence of the features, steps, operations and / or components, but do not exclude the existence or addition of one or more other features, steps, operations or components.

[0040] All terms (including technical and scientific terms) used herein have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification, and should not be interpreted in an idealized or overly rigid manner.

[0041] like Figure 1 As shown, a method for handling open source component security vulnerabilities includes the following steps:

[0042] By collecting open source component information and vulnerability information from project code repositories and vulnerability disclosure platforms, the open source component information and vulnerability information are standardized and preliminary correlation data is generated;

[0043] Open source component information includes detailed information about all open source software components used in the project, such as component name, version number, dependencies, etc. This information is usually stored in the project's code repository, such as GitHub, GitLab and other platforms. By extracting this information, you can accurately grasp the dependencies between the various components in the project, which is an important basis for subsequent analysis.

[0044] Vulnerability information mainly comes from vulnerability disclosure platforms, such as the CVE (Common Vulnerabilities and Exposures) database and the NVD (National Vulnerability Database). Vulnerability information includes known security vulnerabilities of each component, including vulnerability number, vulnerability description, impact scope, hazard level, etc. By collecting this information, direct data support can be provided for subsequent security risk assessments.

[0045] In the actual implementation process, data is collected in batches from these platforms and warehouses through automated scripts or API interfaces. These data may be highly heterogeneous in their original form, such as differences in data format, naming conventions, information structure, etc. In order to solve these problems, standardization is applied to the data processing process. The core principle of standardization is to convert data from different sources and formats into a unified standard format for subsequent analysis and processing. This usually includes steps such as field mapping, format conversion, redundant information cleaning, and data filling. For example, the vulnerability numbers in different platforms are unified into the CVE format, or the version number formats of different components are unified into a standard format for subsequent comparison and association.

[0046] Through the above standardization process, preliminary correlation data is generated. The generation of preliminary correlation data not only includes the basic matching of open source components and vulnerability information, but also establishes preliminary correlations between components and their dependencies, and components and known vulnerabilities on this basis. These correlations provide basic data for subsequent multi-dimensional graph generation and cross-correlation analysis. For example, if a component A depends on component B, and component B is known to have a high-risk vulnerability C, the preliminary correlation data will identify that component A is potentially affected by the vulnerability.

[0047] Standardization and the generation of preliminary correlation data significantly improve the accuracy and efficiency of subsequent analysis steps. Data in a unified format makes the construction of dependency graphs more intuitive and accurate, avoiding misjudgments or omissions caused by inconsistent data formats. At the same time, preliminary correlation data provides more accurate input for the inference model, reduces the propagation of errors, and enhances the reliability of the overall analysis.

[0048] Through multi-dimensional graph generation and cross-correlation analysis, multi-dimensional dependency graphs are generated based on preliminary correlation data, and the dependency graphs are dynamically evolved and optimized through adaptive evolutionary algorithms to generate verification dependency graphs. The verification dependency graphs are updated in real time through a cross-graph feedback mechanism.

[0049] Based on the preliminary association data, the system generates dependency graphs in multiple dimensions. These dependency graphs include but are not limited to static dependency graphs, dynamic dependency graphs, similarity graphs, and vulnerability propagation graphs. The static dependency graph shows the fixed dependencies of open source components in the project, such as one library depends on a certain version of another library. The dynamic dependency graph reflects the dynamic loading relationship of components at runtime, capturing dependencies that are only exposed at runtime. Similarity graphs are used to indicate the degree of similarity between components, such as the relationship between components of different versions or components with similar functions. The vulnerability propagation graph shows the path by which known vulnerabilities may propagate between components through dependencies.

[0050] After generating these graphs, the system integrates these multi-dimensional graphs through cross-correlation analysis to obtain a more comprehensive view of component dependencies. For example, the combination of static dependency graphs and dynamic dependency graphs can reveal runtime dependencies that are not captured in static dependencies. The combination of vulnerability propagation graphs helps identify the paths along which vulnerabilities may propagate through complex dependency chains. This cross-correlation analysis integrates multi-dimensional information by calculating the weights and edge relationships between nodes in each graph, providing a richer context for further security analysis.

[0051] The adaptive evolutionary algorithm is applied to the dynamic evolution and optimization of the dependency graph. This algorithm dynamically adjusts the nodes and edges in the dependency graph so that the graph can adapt to new data or dependencies. For example, when a new vulnerability is discovered and added to the system, the dependency graph will automatically adjust the node weights and edge relationships to reflect the impact of this new information. The adaptive evolutionary algorithm uses real-time data to continuously optimize the dependency graph to ensure the accuracy and timeliness of the dependency graph.

[0052] After the evolved dependency graph is generated, the system will update it in real time through the cross-graph feedback mechanism. The cross-graph feedback mechanism monitors the changes between different dependency graphs and dynamically adjusts the node and edge relationships in the dependency graph, so that the final generated verification dependency graph always reflects the latest status of the system. For example, if the dependency of a component changes or a new component is introduced, the verification dependency graph will be automatically updated according to these changes to ensure that it always accurately reflects the current system's dependencies and potential vulnerability propagation paths.

[0053] Through multi-dimensional graph generation and cross-correlation analysis, the system can obtain a comprehensive, multi-level component dependency view, making it possible to comprehensively and accurately identify potential security risks in open source components. The adaptive evolutionary algorithm and cross-graph feedback mechanism ensure the dynamic update and continuous optimization of these analysis results, which can reflect changes in the system in real time and provide a highly flexible and accurate security situation assessment method.

[0054] like Figure 2As shown, preferably, the multi-dimensional graph generation and cross-correlation analysis steps include: generating a static dependency graph, a dynamic dependency graph, a similarity graph and a vulnerability propagation graph, and performing cross-correlation analysis on the static dependency graph, the dynamic dependency graph, the similarity graph and the vulnerability propagation graph by using a multi-layer network model to generate multi-dimensional cross-correlation data, wherein the multi-layer network model determines the mutual influence between the graphs by calculating the relationship weights between the nodes in each graph, and uniformly represents the dependencies of different dimensions, and the relationship weights W ij The calculation formula is:

[0055]

[0056] Among them, W ij represents the relationship weight between node i and node j; C ij represents the interaction frequency between node i and node j; T i Represents the total active time of node i and node j respectively. Through this formula, the system can quantify the strength of association between different components. The relationship weight between component pairs with high interaction frequency is greater, which means that the dependency between them is closer and the possibility of vulnerability propagation is higher.

[0057] The system generates static dependency graphs, dynamic dependency graphs, similarity graphs, and vulnerability propagation graphs. These graphs represent relationships between components in different dimensions. The static dependency graph shows the fixed dependencies between components at the code level, such as a component requiring a specific version of another component at compile time. The dynamic dependency graph captures the dependencies that components may load at runtime, which are usually difficult to find in static code analysis. Similarity graphs are used to represent the relationships between components with similar functions or versions, which is very important for identifying components that may share vulnerabilities. The vulnerability propagation graph shows the paths that vulnerabilities may spread through dependencies, helping to identify possible risk diffusion points in the system.

[0058] After generating these graphs, the system uses a multi-layer network model to perform cross-correlation analysis on these graphs. The principle of the multi-layer network model is that it superimposes multiple graphs of different dimensions and reveals more complex relationships by analyzing the performance of nodes (i.e. components) in each graph. For example, in a static dependency graph and a dynamic dependency graph, the same node may exhibit different dependency characteristics. The multi-layer network model can combine the information of these two dimensions to obtain a more accurate dependency graph.

[0059] Multidimensional cross-correlation data is generated by comprehensively calculating and analyzing these weights. Multidimensional cross-correlation data uniformly represents dependencies in different dimensions and provides a comprehensive dependency view. This unified representation allows the system to understand the relationship between components at a higher level of abstraction, thereby more accurately evaluating the security status of the system.

[0060] In the actual implementation process, the system generates a graph containing multi-dimensional dependencies based on multi-dimensional cross-correlation data. These graphs not only show the dependencies of the current components, but also reveal the strength of the association between different components through quantified relationship weights. For example, if the interaction frequency between component A and component B is high and their active time is long, then the relationship weight between them will be high, indicating that they are highly dependent on each other in actual operation. This strong dependency also means that if there is a vulnerability in component B, component A may be affected.

[0061] Through the generation of this multi-dimensional graph and cross-correlation analysis, the system can understand the relationship between components in a more comprehensive and accurate way. Multi-dimensional cross-correlation data provides a solid foundation for subsequent security analysis, allowing the system to more accurately predict and prevent potential security risks. Especially in a complex open source component ecosystem, this multi-level correlation analysis can effectively identify risks that may be missed in traditional single-dimensional analysis, improving the overall security of the system.

[0062] like Figure 3 As shown, preferably, the adaptive evolutionary algorithm generates an evolutionary dependency graph by dynamically adjusting the relationship between the node weights and edges in the multidimensional cross-correlation data. The evolutionary dependency graph is updated in real time through a cross-graph feedback mechanism. The cross-graph feedback mechanism is based on the dynamic interaction information between multi-dimensional graphs, and finally generates a verification dependency graph. The update formula of the adaptive evolutionary algorithm is:

[0063] E(t+1)=α×E(t)+β×ΔE

[0064] Among them, E(t) represents the node weight at time t, reflecting the importance or influence of the component in the dependency graph at a certain moment; ΔE represents the weight increment at time t, which is an adjustment calculated based on the latest data and is used to reflect new security information or dependency changes; α and β are evolution coefficients, which determine the proportion of the influence of the current weight and new data on the weight update. α is usually used to maintain the weight of historical data, and β is used to reflect the importance of new information.

[0065] The core of the adaptive evolutionary algorithm lies in its ability to adjust dynamically. This algorithm continuously optimizes the node weights and edge relationships in multi-dimensional cross-correlated data, allowing the system to adapt to new data or dependencies. For example, when a new security vulnerability is disclosed and affects a component, the system can immediately adjust the relationship weights between the component and other components based on the new data, thereby updating the dependency graph to make it more consistent with the current actual situation.

[0066] Through the above formula, the system can dynamically adjust the weight of the node at each time point to generate an evolutionary dependency graph, which not only reflects the current dependency relationship, but also can predict future dependency changes to a certain extent. For example, if a component is frequently updated or affected by multiple vulnerabilities over a period of time, the node weight of the component in the evolutionary dependency graph will gradually increase, prompting the system and administrators to pay attention to the fact that this component may become a potential source of risk.

[0067] Next, the cross-graph feedback mechanism further enhances the accuracy and timeliness of the evolutionary dependency graph. This mechanism adjusts the node weights and edge relationships in the evolutionary dependency graph in real time by monitoring the dynamic interaction information between graphs of different dimensions. For example, if a significant change in the dependency relationship of a component is found between the static dependency graph and the dynamic dependency graph (such as a new dependency or a change in the dependency path), the system will immediately update the evolutionary dependency graph through the cross-graph feedback mechanism to ensure that it is consistent with the actual situation. The final generated verification dependency graph can accurately reflect the dependency structure of the current system and provide a reliable basis for further security analysis and vulnerability remediation strategies.

[0068] The combination of adaptive evolutionary algorithms and cross-graph feedback mechanisms enables the system to maintain a high degree of sensitivity and adaptability in the face of a constantly changing security environment. This dynamic adjustment process can not only reflect changes in dependencies between components in real time, but also identify potential security risks in advance, thus providing strong support for the system's global security posture.

[0069] For example, if a critical open source component is found to have a serious security vulnerability, the system quickly adjusts the weight of the component through an adaptive evolutionary algorithm and updates all dependencies related to it through a cross-graph feedback mechanism. The evolutionary dependency graph can therefore immediately reflect which other components may be affected and help administrators quickly take remediation measures to prevent the vulnerability from spreading.

[0070] Based on the verification dependency graph, distributed Bayesian reasoning and co-evolution are performed. Intelligent agents reason and evolve data at different levels, and combine multi-level data fusion algorithms to deeply fuse data of different dimensions. Through the cross-layer multi-agent collaborative game model, strategies are optimized and shared to generate fusion analysis data and game feedback data.

[0071] The verification dependency graph is the result of early dynamic optimization through multi-dimensional graph generation, cross-correlation analysis, and adaptive evolutionary algorithms. It provides a comprehensive view of dependencies and vulnerability propagation paths that are updated in real time. Based on this dependency graph, the system begins to perform distributed Bayesian reasoning. Bayesian reasoning is a probabilistic reasoning method that calculates the possibility of a hypothesis based on existing evidence. In the present invention, distributed Bayesian reasoning is performed independently by multiple intelligent agents at different system levels. The intelligent agents reason about the security risks of specific components or dependencies in a distributed manner. For example, if a component is found to have a potential vulnerability, the intelligent agent will calculate the probability of the vulnerability affecting other parts of the system based on its position in the verification dependency graph and its relationship with other components.

[0072] After each intelligent agent completes its own Bayesian reasoning, the system will use the co-evolution mechanism to integrate and optimize the reasoning results of each level. Co-evolution means that while intelligent agents independently evolve reasoning models at different levels, they achieve overall evolution and optimization through information sharing and feedback mechanisms. In this process, intelligent agents not only make adjustments based on their own reasoning results, but also optimize their own models by sharing information with other agents. For example, if an agent at a certain level finds that a component is in a critical position in multiple dependency paths, other agents will take this information into account and adjust their reasoning strategies accordingly.

[0073] The system uses a multi-level data fusion algorithm to deeply fuse data of different dimensions. These data include the relationship weights in the verification dependency graph, the probability data generated by distributed Bayesian reasoning, and the adjustment information shared by each agent in the co-evolution process. The core of the multi-level data fusion algorithm is to integrate data from different levels to generate a unified analysis view. This fusion can not only improve the comprehensiveness and accuracy of data analysis, but also reduce the deviation caused by single-dimensional data. For example, by combining the probability of vulnerability propagation with the frequency of component use and the strength of dependency relationships, the system can more accurately identify the security risks that need the most attention.

[0074] The system optimizes and shares the generated strategies through a cross-layer multi-agent collaborative game model. The game model is a mathematical tool for decision optimization, where multiple intelligent agents simulate different strategy combinations to find the optimal security protection strategy for the entire system. Cross-layer games mean that these strategies are not only optimized within a single layer, but also consider the interactive effects across layers. For example, a strategy is effective at the application layer, but if its dependencies on the underlying libraries are not taken into account, then the strategy may not be optimal overall. Through the cross-layer game model, intelligent agents can collaborate with each other to jointly develop the best strategy for the system as a whole.

[0075] The system generates fusion analysis data and game feedback data. Fusion analysis data provides a comprehensive assessment of the system security status after multi-level data fusion, while game feedback data contains the optimal strategy combination between different levels. These data not only provide a basis for real-time security monitoring and decision-making, but also provide direct guidance for the formulation of subsequent security protection measures and vulnerability repair strategies.

[0076] This process significantly enhances the system's ability to respond to complex and dynamic security threats. Through distributed Bayesian reasoning and co-evolution, the system can simultaneously analyze and predict potential risks at multiple levels and dimensions, thereby more accurately identifying the propagation path and impact range of security vulnerabilities. Multi-level data fusion ensures the comprehensiveness and reliability of the analysis results, while the cross-layer multi-agent collaborative game model provides the system with the optimal security protection strategy, enabling the system to maintain a high degree of security and resilience in a complex open source component ecosystem.

[0077] Preferably, the distributed Bayesian reasoning and co-evolution step includes: independently performing data reasoning at different levels through multiple intelligent agents, each intelligent agent performs Bayesian reasoning based on the data at its level, and after the reasoning is completed, the reasoning results are shared through the co-evolution model to generate co-evolution reasoning data, and the co-evolution reasoning data is used to update the reasoning models at different levels in real time. The calculation formula of the Bayesian reasoning is:

[0078]

[0079] Among them, P(H|D) represents the posterior probability of hypothesis H being true under data D; P(D|H) represents the likelihood of observing data D when hypothesis H is true; P(H) represents the prior probability of hypothesis H; and P(D) represents the marginal probability of data D.

[0080] The system uses multiple intelligent agents to independently perform data reasoning at different levels. Each intelligent agent is responsible for the security assessment of components or dependencies at a specific level. The intelligent agent uses the Bayesian reasoning method to deduce the probability that a component has a security risk based on known data (such as historical vulnerability data of the component, dependencies, usage frequency, etc.). The core of Bayesian reasoning is to calculate the posterior probability that hypothesis H is true under new data D by updating the prior probability (that is, the initial hypothesis when there is no new evidence).

[0081] In the present invention, this reasoning process is used to evaluate the security risk of a component (hypothesis H) given its current dependencies and known vulnerability data (data D). For example, if an open source component is recently found to have multiple new high-risk vulnerabilities, Bayesian reasoning will use this new information to adjust the previous assumptions about the security status of the component and derive new posterior probabilities that reflect the latest results of the current risk assessment.

[0082] After the intelligent agent completes the Bayesian reasoning within its level, the system will share and integrate these reasoning results through the co-evolution model. The principle of co-evolution is to let the reasoning models at each level influence each other through information sharing, so as to achieve overall evolution and optimization. For example, if a low-level library component is assessed as high risk, the reasoning models of other upper-level components that depend on the library will automatically adjust to reflect this new security information, thereby improving the risk perception ability of the overall system.

[0083] The generated co-evolutionary reasoning data not only includes the reasoning results of each level, but also integrates the information interaction between different levels to form a more comprehensive risk assessment view. These data are used to update the reasoning models at each level in real time to ensure that the system can quickly adjust as the security environment changes. For example, if the system detects an outbreak of a high-risk vulnerability, the co-evolutionary reasoning data will immediately reflect this situation and quickly propagate to related components and dependencies through the real-time updated reasoning model, ensuring that the system responds to potential risks in the first place.

[0084] The combination of distributed Bayesian reasoning and co-evolutionary models enables the system to perform complex security assessments at multiple levels simultaneously, and these assessments can adapt to emerging security threats in real time. This multi-level, multi-agent reasoning method greatly improves the sensitivity and response speed of the system, and can more accurately predict and respond to complex security risks. In addition, by sharing and integrating information through the co-evolutionary model, the system can achieve cross-level risk perception, so that security assessments are not limited to a single component or dependency, but extend to the entire system level, forming a global security protection network.

[0085] For example, if a low-level component is assessed as high risk by intelligent agent A, and this component is relied upon by multiple upper-level applications, then through co-evolution, the risk assessment models of all upper-level applications involved will be quickly adjusted and the security risks of these applications will be recalculated. This process ensures that the system can maintain a highly consistent risk perception across different levels and take effective protective measures.

[0086] Preferably, a multi-level data fusion algorithm is used to deeply fuse the co-evolutionary reasoning data with the data of different dimensions in the verification dependency graph to generate fusion analysis data, and the multi-level data fusion algorithm processes data features of different dimensions in a hierarchical manner.

[0087] In the early stage, the system generated different levels of reasoning results and dependency views through co-evolutionary reasoning and verification dependency graphs. The co-evolutionary reasoning data contains the security risk assessment results inferred by intelligent agents from different levels, while the verification dependency graph provides a component dependency graph after multi-dimensional graph generation, dynamic optimization and real-time update. Although these data each reflect part of the system status, in order to obtain a global security assessment, these multi-dimensional data must be deeply integrated.

[0088] The multi-level data fusion algorithm is designed to solve this problem. The algorithm processes data features of different dimensions in layers and gradually fuses data from different sources to generate a more complete and accurate security assessment view. Specifically, the data fusion process is usually divided into several levels:

[0089] Primary fusion layer: At this level, the system first performs basic fusion of data from different dimensions. For example, the node weights in the verification dependency graph and the risk probabilities in the co-evolutionary reasoning data will be fused at this level. At this point, data fusion mainly relies on simple weighted average or geometric average methods, with the goal of combining information from different data sources without losing the data characteristics of each source.

[0090] Intermediate fusion layer: Based on the primary fusion, the intermediate fusion layer further considers the interaction and mutual influence between data. For example, in the verification dependency graph, if a component has a high node weight and is also assessed as high risk in the co-evolutionary reasoning data, the intermediate fusion layer will further amplify the importance of the component in the global context. This process can be achieved through more complex fusion functions or machine learning algorithms, so that the fused data can better reflect the real risk situation of the system.

[0091] Advanced fusion layer: At the advanced fusion layer, the system globally optimizes the fused data to generate the final fused analysis data. This level of processing usually takes into account global dependencies and cross-level interactions. For example, if a component that seems to have little risk at a low level, but its dependent components show high risk at a higher level, the advanced fusion layer will adjust this situation to increase the risk assessment value of the component. The algorithms that may be used in the advanced fusion layer include graph neural networks or other deep learning methods to capture deeper correlations and complex patterns in the data.

[0092] Through this multi-level data fusion, the fusion analysis data ultimately generated by the system not only integrates information from multiple dimensions, but also fully considers the complex relationships and mutual influences between data in each dimension, thereby providing the system with a more comprehensive and accurate security assessment view.

[0093] The biggest advantage of using a multi-level data fusion algorithm is that it can improve the overall risk perception of the system. By fusing data from different sources and different levels, the system can better identify potential security vulnerabilities and their possible propagation paths, avoiding the limitations of single-dimensional analysis. For example, if the system only relies on the verification dependency graph, it may underestimate the potential risks in some complex dependency chains; and if it only relies on co-evolutionary reasoning data, it may ignore the actual dependencies between components. Through multi-level data fusion, the system can organically combine these two aspects of information to generate a more complete and reliable security assessment result.

[0094] For example, in a large open source project, a low-level library may not directly expose a vulnerability, but a high-level component it depends on is seriously affected by a vulnerability. It is difficult to accurately assess this risk through simple reasoning or dependency analysis. However, through multi-level data fusion algorithms, such complex dependencies and potential risks can be accurately identified, ensuring that the system does not miss key security threats during the assessment.

[0095] Preferably, the cross-layer multi-agent collaborative game model optimizes strategy sharing between intelligent agents through game analysis. The game analysis generates game feedback data based on the strategy execution results of different intelligent agents at each level. The game feedback data is used to further adjust and optimize the fusion analysis data.

[0096] The core of the cross-layer multi-agent collaborative game model is to use the principles of game theory to optimize and coordinate the strategies between different intelligent agents. Each intelligent agent operates independently at different levels and formulates corresponding security policies based on the components or dependencies it is responsible for. However, in a complex open source software ecosystem, there are often close dependencies between different levels and components, and the strategy of a single intelligent agent may affect the decision-making of other agents. Therefore, the local optimal strategy of a single agent may not necessarily achieve the global optimality of the system. At this time, by introducing game analysis, the system can simulate the interaction between multiple intelligent agents and explore the best combination of their respective strategies.

[0097] Game analysis is based on the policy execution results of different intelligent agents at various levels. Each intelligent agent selects the optimal security policy based on its own reasoning and decision-making model, such as which components to repair first or how to deal with potential vulnerability propagation paths. Game analysis simulates the execution effects of these policies in different scenarios and evaluates the impact of different policy combinations on the overall security of the system through comparative analysis. For example,

[0098] At one level, if an intelligent agent decides to postpone patching a vulnerability in a component, while an agent at another level has already prioritized its dependencies, this combination of strategies may reduce the overall security risk of the system.

[0099] Through this game analysis, the system generates game feedback data. This feedback data contains the effect evaluation of different strategy combinations, helping the system to identify which strategy combinations can achieve the best security effect globally. For example, a game analysis may show that although a strategy performs well at a single level, it may not be ideal when combined with strategies at other levels. Through this feedback data, the system can adjust the intelligent agent strategies at each level in real time, thereby achieving better coordination between different levels.

[0100] The system uses this game feedback data to further adjust and optimize the fusion analysis data. The fusion analysis data has been previously generated through a multi-level data fusion algorithm and includes the overall security posture assessment of the system. However, the introduction of game feedback data allows this data to be further optimized at the policy level. For example, based on the feedback from game analysis, the system may readjust the security risk weights of certain components or adjust the priority processing order of certain dependencies, thereby more accurately reflecting the optimized security policy combination at the data level.

[0101] The cross-layer multi-agent collaborative game model enables the system to achieve overall security optimization in a complex multi-level dependency environment through strategy optimization. This process ensures that intelligent agents at different levels can effectively collaborate rather than just operate independently, thereby improving the security of the system globally. For example, if a single intelligent agent does not take into account the dependencies of other levels when formulating a strategy, it may lead to inefficient execution of security measures or even negative effects. Through game analysis, the system can identify and avoid these potential problems, ensuring that the strategies of each level are consistent with the global optimal strategy.

[0102] For example, suppose in an open source project, the intelligent agent at the application layer decides to postpone the vulnerability fix for a component, but the agent at the base library layer has discovered that the component is a critical dependency and plans to prioritize its vulnerability. Without game analysis, the decisions of these two levels may conflict and even lead to an increase in overall security risks. However, through cross-layer game analysis, the system is able to identify this potential inconsistency and adjust the application layer agent's strategy to coordinate with the strategy of the base library layer, thereby achieving more efficient vulnerability repair and risk control on a global scale.

[0103] Based on the fusion analysis data, the time point and spatial location of future vulnerability outbreaks are predicted through the spatiotemporal regression model and the cross-dimensional joint prediction model, and cross-dimensional spatiotemporal prediction data is generated. A preventive repair strategy with multiple repair stages is formulated, and repair operations are performed in stages. Repair feedback data is generated through real-time monitoring;

[0104] The system uses fusion analysis data, which has integrated information from different levels and dimensions through a multi-level data fusion algorithm, providing a comprehensive security assessment view. This data not only includes component dependencies and vulnerability propagation paths, but also integrates security policy information that has been optimized through game analysis. Therefore, fusion analysis data provides a reliable foundation for the prediction model, making subsequent predictions more accurate and targeted.

[0105] The system introduces a spatiotemporal regression model to predict the time and spatial location of future vulnerability outbreaks. The principle of the spatiotemporal regression model is that it can combine information in the two dimensions of time and space, and predict when and where a vulnerability may be triggered or exploited in the future by analyzing historical data and the current system status. Specifically, the spatiotemporal regression model calculates the possible outbreak point of the vulnerability based on the time series features (such as vulnerability discovery time, repair time, component update frequency, etc.) and spatial features (such as the distribution of components in different systems, the geographical distribution of dependencies, etc.) in the fusion analysis data. For example, if a component is widely used in systems in multiple geographical locations, and its vulnerability repair time is long, then the spatiotemporal regression model may predict that the component is vulnerable to attack at a specific time and location in the future.

[0106] On the basis of the spatiotemporal regression model, the system also uses a cross-dimensional joint prediction model to further improve the accuracy and comprehensiveness of the prediction. The cross-dimensional joint prediction model can combine data from multiple dimensions such as time, space, dependencies, and risk levels for joint analysis. For example, the model may simultaneously consider the historical outbreak records of vulnerabilities, the frequency of component use, the complexity of dependency chains, and the behavior patterns of potential attackers. Through this cross-dimensional comprehensive analysis, the system can more accurately predict the specific manifestations and scope of impact of vulnerabilities in the future. The generated cross-dimensional spatiotemporal prediction data provides a direct basis for formulating preventive repair strategies.

[0107] Based on these prediction data, the system will develop a preventive repair strategy that includes multiple repair phases. The core of this strategy is that it is not a simple one-time repair of the vulnerability, but a phased repair operation based on the predicted time point and spatial location of the vulnerability outbreak. For example, if the spatiotemporal regression model predicts that a vulnerability will pose a threat to systems in a specific area within a certain period of time within two months, the system may prioritize repairing the relevant systems before that time period, rather than waiting until the vulnerability has been exploited before taking action. Through this phased repair strategy, the system can better allocate resources and avoid wasting too many repair resources during periods of low vulnerability threats.

[0108] At the same time, when the system performs these repair operations, it will generate repair feedback data through real-time monitoring. This feedback data includes information such as the implementation effect of the repair operation, new problems found during the repair process, and the system status after the repair. Real-time monitoring can not only ensure that the repair strategy is executed as planned, but also promptly discover and respond to emergencies during the repair process. For example, if it is found during the repair process that the dependency chain of a component is more complex than expected, or the repair operation has an unexpected impact on other systems, the system can immediately adjust the repair strategy to avoid greater security risks.

[0109] Through the spatiotemporal regression model and the cross-dimensional joint prediction model, the system can predict the potential vulnerability outbreak time and spatial location in advance, so as to formulate a more accurate and effective preventive repair strategy. This process significantly improves the system's active defense capabilities, making vulnerability repair no longer just a passive response, but proactively taking preventive measures before the vulnerability breaks out. The combination of phased repair operations and real-time monitoring ensures the effectiveness and timeliness of repair measures, while minimizing the interference with the normal operation of the system during the repair process.

[0110] For example, suppose a common component in a large open source project is found to have a high-risk vulnerability, and the spatiotemporal regression model predicts that the vulnerability may pose a serious threat to certain key systems in Asia in the next three months. Based on this prediction, the system can prioritize the repair of related systems in Asia during this period, while closely monitoring the status of systems in other regions so that timely measures can be taken when necessary. In this way, the system can not only effectively prevent vulnerability risks, but also ensure that repair operations can play the greatest role when resources are limited.

[0111] Preferably, a time series analysis of historical system operation data is performed through a space-time regression model to generate space-time regression prediction data, which predicts the time point and spatial location of future vulnerability outbreaks based on the time and space characteristics of the historical system operation data.

[0112] The system uses a spatiotemporal regression model to analyze historical system operation data. Historical system operation data includes various behaviors and event records of the system in the past period of time, such as component usage frequency, update logs, vulnerability repair time, geographic location of vulnerability occurrence, etc. These data usually have strong temporal and spatial characteristics. For example, the vulnerability of a component may recur in a specific time period or be more concentrated in certain geographic areas. The task of the spatiotemporal regression model is to combine these temporal and spatial characteristics to identify potential patterns and trends.

[0113] Time series analysis is one of the basic technologies of the spatiotemporal regression model. Time series analysis reveals the periodicity, trend, and burst patterns in the data by studying the state changes of the system at different time points. For example, a component may have a security vulnerability discovered every once in a while. By analyzing this time series data, the spatiotemporal regression model can predict the time point when the next vulnerability may appear. Time series analysis not only focuses on the data changes in a single time dimension, but also takes into account the driving factors behind these changes, such as system updates, changes in the frequency of external attacks, etc.

[0114] In the space-time regression model, in addition to the time dimension, spatial features are also key analysis objects. Spatial features include the geographical location of the vulnerability outbreak, the distribution of affected systems, etc. By analyzing these spatial features, the model can identify the propagation pattern of vulnerabilities in different geographical locations. For example, a vulnerability recurs in systems in certain geographical areas, perhaps because the system configurations in these areas have common vulnerabilities or attackers pay more attention to these areas. By combining time series data with spatial feature data, the space-time regression model can predict the location of future vulnerability outbreaks, not only accurately locating them in time, but also locking in high-risk areas in space.

[0115] The generated spatiotemporal regression prediction data is a direct result of this analysis process. Based on the temporal and spatial characteristics of historical system operation data, these data provide a comprehensive prediction of future vulnerability outbreaks. For example, spatiotemporal regression prediction data may show that in a certain month in the future, certain servers located in Europe are at a higher risk of being attacked by a specific type of vulnerability. This prediction not only helps to prevent potential attacks in advance, but also helps system administrators to reasonably allocate security resources and repair components that may be affected in advance.

[0116] The spatiotemporal regression model greatly improves the system's predictability and proactive defense capabilities. Through in-depth analysis of historical data, the system is able to identify potential security threats in advance and take preventive measures to prevent vulnerabilities from causing significant losses in the future. For example, a component may have a high-risk vulnerability approximately every six months over the past few years, and these vulnerabilities are mainly concentrated on specific servers in Asia. Based on this pattern, the spatiotemporal regression model can predict that the next vulnerability may reappear on these servers in three months, allowing system administrators to take precautions in advance.

[0117] Preferably, the cross-dimensional joint prediction model generates cross-dimensional spatiotemporal prediction data by jointly analyzing the spatiotemporal regression prediction data and the game feedback data. The cross-dimensional joint prediction model generates cross-dimensional spatiotemporal prediction results by combining multi-dimensional data of time, space, dependency and strategy feedback, and formulates a preventive repair strategy including multiple repair stages based on the cross-dimensional spatiotemporal prediction data.

[0118] The system generates spatiotemporal regression prediction data, which is obtained by analyzing the temporal and spatial characteristics of historical system operation data. These data can predict the time and spatial location of future vulnerabilities. However, relying solely on information in the spatiotemporal dimension may not be able to fully capture complex system dependencies and potential security risks. To this end, the system introduces game feedback data, which is derived from a cross-layer multi-agent collaborative game model. By simulating the strategy combination of different intelligent agents, the overall security strategy of the system is optimized.

[0119] The core of the cross-dimensional joint prediction model is to generate a more comprehensive cross-dimensional spatiotemporal prediction result by jointly analyzing the spatiotemporal regression prediction data with the game feedback data, combining multi-dimensional information such as time, space, dependency, and strategy feedback. The principle of this joint analysis is that the system can better understand and predict potential security threats in the system by integrating data from multiple dimensions. For example, spatiotemporal regression prediction data may show that a vulnerability may break out at a certain point in time and location in the future, while game feedback data may reveal that if the system does not fix the vulnerability in time, it may cause the collapse of the entire dependency chain. Through the cross-dimensional joint prediction model, the system can identify these complex associations and provide a more comprehensive security assessment.

[0120] Specifically, the cross-dimensional joint prediction model integrates information from different data dimensions in a variety of ways. For example, it may use a multivariate regression analysis method, taking spatiotemporal prediction data as independent variables and game feedback data as dependent variables, to determine the relationship between data of different dimensions. At the same time, the model also uses machine learning algorithms, such as random forests or deep learning models, to mine complex patterns hidden in multidimensional data. This multi-level analysis ensures that the system can capture potential security risks that may be missed by single-dimensional data analysis.

[0121] Cross-dimensional spatiotemporal prediction data is a direct product of this joint analysis process. These data not only take into account the time and space factors of vulnerability outbreaks, but also integrate the complexity of system dependencies and game feedback data to form a comprehensive security risk forecast. For example, cross-dimensional spatiotemporal prediction data may show that a vulnerability that is about to break out in the future will not only affect certain specific geographical locations, but may also spread rapidly to other areas through dependencies, thereby causing greater security risks.

[0122] Based on these cross-dimensional spatiotemporal prediction data, the system can formulate more detailed and scientific preventive repair strategies. Unlike traditional one-time repair strategies, preventive repair strategies usually include multiple stages, aiming to gradually repair potential vulnerabilities in steps and regions. For example, if the cross-dimensional joint prediction model predicts that a vulnerability will first break out in North America within the next month, the system can prioritize repairing related components in North America, while repairs in other regions can be performed later. This phased repair strategy not only improves the utilization efficiency of repair resources, but also better responds to large-scale security threats and avoids new problems caused by improper operations during the repair process.

[0123] The cross-dimensional joint prediction model greatly improves the system's predictive and defensive capabilities. By jointly analyzing data from multiple dimensions such as time, space, dependencies, and policy feedback, the system can more comprehensively understand and predict potential security threats. This multi-dimensional comprehensive analysis not only improves the accuracy of predictions, but also helps the system develop more flexible and effective preventive repair strategies, so that it can proactively respond to vulnerabilities before they break out.

[0124] For example, suppose a component in an open source project is predicted to be vulnerable to a major attack in the next two months, and the component has dependencies in multiple systems. Through the cross-dimensional joint prediction model, the system can identify which dependencies are the most vulnerable and which systems are most likely to be affected. Based on this information, the system can prioritize repairing the most critical parts and dynamically adjust strategies during the repair process to respond to new security threats and changes.

[0125] Based on cross-dimensional spatiotemporal prediction data and repair feedback data, global optimization decisions are made, and system optimization and adaptive learning are performed through feedback mechanisms to ultimately generate evolutionary learning model data.

[0126] Preferably, the cross-dimensional spatiotemporal prediction data and the repair feedback data are trained and optimized by an adaptive learning algorithm to generate evolutionary learning model data, and the adaptive learning algorithm continuously adjusts the decision model parameters based on historical decision data and real-time feedback data.

[0127] The system makes global optimization decisions through cross-dimensional spatiotemporal prediction data and repair feedback data. The cross-dimensional spatiotemporal prediction data is generated by the previous multi-dimensional joint prediction model, providing predictions of the possible outbreak time points, spatial locations, and diffusion paths of future vulnerabilities. These data provide forward-looking guidance for the system's decision-making. Repair feedback data comes from the actual repair operations performed, including repair effects, unforeseen security issues, and the status of the system after repair. These data can reflect the actual changes in the system after the implementation of security measures, providing an important realistic basis for subsequent decision-making.

[0128] The core of global optimization decision-making is that the system does not simply rely on historical data or static models to make decisions, but adjusts and optimizes its security protection strategy in real time through comprehensive analysis of cross-dimensional spatiotemporal prediction data and repair feedback data. For example, if the system learns from prediction data that a vulnerability may break out in a specific area in the next few weeks, and the repair feedback data shows that the repair strategy in similar situations was not effective before, the system can choose to adjust the strategy, such as changing the priority of the repair, adjusting the order of the repair steps, or allocating more resources to handle these specific areas.

[0129] In order to achieve this dynamic adjustment, the system introduces an adaptive learning algorithm. The principle of the adaptive learning algorithm is that it can automatically adjust and optimize the parameters of the decision model based on the continuously accumulated historical decision data and real-time feedback data. Specifically, the adaptive learning algorithm analyzes past decisions and actual execution results to identify which strategies perform better under specific conditions, so that these strategies are given priority in future decisions. At the same time, the system will quickly capture the latest effects of the repair operation through a real-time feedback mechanism, so that the decision model can adapt to the new security environment in a timely manner.

[0130] A key advantage of adaptive learning algorithms is their ability to adjust dynamically. Traditional decision-making models usually rely on preset parameters and rules, which are difficult to maintain in an optimal state in a complex and rapidly changing environment. Adaptive learning algorithms, on the other hand, can keep models flexible and efficient in the face of new security threats through continuous learning and adjustment. For example, if the spread of a vulnerability exceeds the expectations of the prediction model, the adaptive learning algorithm can quickly adjust the model parameters based on real-time feedback data, allowing the system to respond in a timely manner, thereby avoiding potential security losses.

[0131] The final generated evolutionary learning model data is the core result of this process. The evolutionary learning model data not only contains the parameters and strategies of the current system's optimal decision-making model, but also records the knowledge and experience accumulated during the adaptive learning process. These data provide the system with a dynamic adjustment and continuous optimization capability, enabling the system to always maintain forward-looking and responsive capabilities in a constantly changing security environment.

[0132] Through global optimization decisions and adaptive learning, the system can continue to maintain a high degree of sensitivity and response capabilities to security threats. The introduction of adaptive learning algorithms enables the system to continuously optimize based on historical and real-time data, so that it can always make the best decisions when facing complex security challenges. This not only improves the overall protection effect of the system, but also reduces the waste of resources and errors in repair operations.

[0133] For example, in a complex open source project, the system predicts that a vulnerability in a key component will cause security issues on certain specific servers in the next few days. However, previous repair feedback data shows that simple repair operations cannot completely solve the problem in similar situations. Through adaptive learning, the system can combine these feedbacks and optimize the repair strategy, and may take more phased and regional repair operations to ensure that the vulnerability will not break out at the predicted time and location, thereby protecting the overall security of the system.

[0134] like Figure 4 As shown, a system for implementing the open source component security vulnerability processing method includes:

[0135] The information collection and preprocessing module is used to collect open source component information and vulnerability information from the project code repository and vulnerability disclosure platform, and standardize the collected open source component information and vulnerability information to generate preliminary related data; the system automatically collects open source component information and related vulnerability data from the project code repository and vulnerability disclosure platform. These data are standardized to generate preliminary related data as the basis for subsequent analysis. Standardization ensures that data from different sources can be uniformly formatted, thereby improving data compatibility and processing efficiency.

[0136] The multi-dimensional graph generation and analysis module is used to generate multi-dimensional dependency graphs based on preliminary correlation data, generate verification dependency graphs through multi-dimensional graph generation and cross-correlation analysis, and dynamically evolve and optimize the dependency graph through an adaptive evolutionary algorithm. The verification dependency graph is updated in real time through a cross-graph feedback mechanism. Based on preliminary correlation data, the system generates multi-dimensional dependency graphs. These graphs reflect the complex dependencies between components. Through multi-dimensional graph generation and cross-correlation analysis, the system can generate verification dependency graphs, and dynamically optimize and evolve them through an adaptive evolutionary algorithm. This process is updated in real time through a cross-graph feedback mechanism to ensure that the dependency graph always reflects the current actual situation.

[0137] The collaborative reasoning and data fusion module is used to perform distributed Bayesian reasoning and collaborative evolution based on the verification dependency graph. It uses intelligent agents to perform data reasoning and evolution at different levels, and combines multi-level data fusion algorithms to deeply fuse data of different dimensions, generate fusion analysis data and game feedback data, and optimize and share strategies through cross-layer multi-agent collaborative game models. Based on the verification dependency graph, the system performs distributed Bayesian reasoning and collaborative evolution, and uses intelligent agents to reason and fuse data at different levels. Combined with multi-level data fusion algorithms, the system can extract more global and reliable fusion analysis data from data of different dimensions, and generate game feedback data at the same time. These data provide the basis for the system to optimize and share strategies.

[0138] The spatiotemporal prediction and repair strategy module is used to predict the time point and spatial location of future vulnerability outbreaks based on fusion analysis data through spatiotemporal regression models and cross-dimensional joint prediction models, generate cross-dimensional spatiotemporal prediction data, and formulate preventive repair strategies including multiple repair stages, perform staged repair operations, and generate repair feedback data through real-time monitoring; the system predicts the time and spatial location of future vulnerability outbreaks through spatiotemporal regression models and cross-dimensional joint prediction models. These prediction data help the system formulate preventive repair strategies for multiple stages, perform repair operations in stages, and obtain repair feedback data through real-time monitoring to ensure that the repair process is effective and does not introduce new security issues.

[0139] The global optimization and learning module is used to make global optimization decisions based on cross-dimensional spatiotemporal prediction data and repair feedback data, and to perform system optimization and adaptive learning through feedback mechanisms, and finally generate evolutionary learning model data. Based on cross-dimensional spatiotemporal prediction data and repair feedback data, the system makes global optimization decisions. Through the feedback mechanism, the adaptive learning algorithm can continuously optimize the decision model of the system, allowing it to evolve over time and as the environment changes, and finally generate evolutionary learning model data, thereby further improving the overall security and responsiveness of the system.

[0140] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems or computer program products. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware.

[0141] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. A method for handling security vulnerabilities of open source components. It is characterized in that The following steps are involved: By collecting open source component information and vulnerability information from project code repositories and vulnerability disclosure platforms, the open source component information and vulnerability information are standardized and preliminary correlation data is generated; Through multi-dimensional graph generation and cross-correlation analysis, multi-dimensional dependency graphs are generated based on preliminary correlation data, and the dependency graphs are dynamically evolved and optimized through adaptive evolutionary algorithms to generate verification dependency graphs. The verification dependency graphs are updated in real time through a cross-graph feedback mechanism. Based on the verification dependency graph, distributed Bayesian reasoning and co-evolution are performed. Intelligent agents reason and evolve data at different levels, and combine multi-level data fusion algorithms to deeply fuse data of different dimensions. Through the cross-layer multi-agent collaborative game model, strategies are optimized and shared to generate fusion analysis data and game feedback data. Based on the fusion analysis data, the time point and spatial location of future vulnerability outbreaks are predicted through the spatiotemporal regression model and the cross-dimensional joint prediction model, and cross-dimensional spatiotemporal prediction data is generated. A preventive repair strategy with multiple repair stages is formulated, and repair operations are performed in stages. Repair feedback data is generated through real-time monitoring; Based on cross-dimensional spatiotemporal prediction data and repair feedback data, global optimization decisions are made, and system optimization and adaptive learning are performed through feedback mechanisms to ultimately generate evolutionary learning model data.

2. According to the method for handling open source component security vulnerabilities according to claim 1, It is characterized in that The multi-dimensional graph generation and cross-correlation analysis step includes: generating a static dependency graph, a dynamic dependency graph, a similarity graph, and a vulnerability propagation graph, and performing cross-correlation analysis on the static dependency graph, the dynamic dependency graph, the similarity graph, and the vulnerability propagation graph by using a multi-layer network model to generate multi-dimensional cross-correlation data, wherein the multi-layer network model determines the mutual influence between the graphs by calculating the relationship weights between the nodes in each graph, and uniformly represents the dependency relationships of different dimensions, and the relationship weights W ij The calculation formula is: Among them, W ij represents the relationship weight between node i and node j; C ij represents the interaction frequency between node i and node j; T i Represent the total active time of node i and node j respectively.

3. According to claim 2, a method for handling open source component security vulnerabilities, It is characterized in that The adaptive evolutionary algorithm generates an evolutionary dependency graph by dynamically adjusting the relationship between node weights and edges in the multi-dimensional cross-correlation data. The evolutionary dependency graph is updated in real time through a cross-graph feedback mechanism. The cross-graph feedback mechanism is based on the dynamic interaction information between multi-dimensional graphs, and finally generates a verification dependency graph. The update formula of the adaptive evolutionary algorithm is: E(t+1)=α×E(t)+β×ΔE Where E(t) represents the node weight at time t; ΔE represents the weight increment at time t; α and β are the evolution coefficients respectively.

4. According to claim 1, a method for handling open source component security vulnerabilities, It is characterized in that The distributed Bayesian reasoning and co-evolution step includes: performing data reasoning independently at different levels through multiple intelligent agents, each intelligent agent performs Bayesian reasoning based on the data at its level, and after the reasoning is completed, the reasoning results are shared through the co-evolution model to generate co-evolution reasoning data, and the co-evolution reasoning data is used to update the reasoning models at different levels in real time. The calculation formula of the Bayesian reasoning is: Among them, P(H|D) represents the posterior probability of hypothesis H being true under data D; P(D|H) represents the likelihood of observing data D when hypothesis H is true; P(H) represents the prior probability of hypothesis H; and P(D) represents the marginal probability of data D.

5. According to claim 4, a method for handling open source component security vulnerabilities, It is characterized in that A multi-level data fusion algorithm is used to deeply fuse the co-evolutionary reasoning data with the data of different dimensions in the verification dependency graph to generate fusion analysis data. The multi-level data fusion algorithm processes data features of different dimensions in a hierarchical manner.

6. The method for processing open source component security vulnerabilities according to claim 1, It is characterized in that The cross-layer multi-agent collaborative game model optimizes the strategy sharing between intelligent agents through game analysis. The game analysis generates game feedback data based on the strategy execution results of different intelligent agents at each level. The game feedback data is used to further adjust and optimize the fusion analysis data.

7. The method for processing open source component security vulnerabilities according to claim 1, It is characterized in that The time series analysis of historical system operation data is performed through a space-time regression model to generate space-time regression prediction data. The space-time regression prediction data is based on the time and space characteristics of the historical system operation data to predict the time point and spatial location of future vulnerability outbreaks.

8. A method for handling open source component security vulnerabilities according to claim 7, It is characterized in that The cross-dimensional joint prediction model generates cross-dimensional spatiotemporal prediction data by jointly analyzing the spatiotemporal regression prediction data and the game feedback data. The cross-dimensional joint prediction model generates cross-dimensional spatiotemporal prediction results by combining multi-dimensional data of time, space, dependency and strategy feedback, and formulates a preventive repair strategy including multiple repair stages based on the cross-dimensional spatiotemporal prediction data.

9. The method for processing open source component security vulnerabilities according to claim 1, It is characterized in that The cross-dimensional spatiotemporal prediction data and the repair feedback data are trained and optimized through an adaptive learning algorithm to generate evolutionary learning model data. The adaptive learning algorithm continuously adjusts the decision model parameters based on historical decision data and real-time feedback data.

10. A system for implementing the open source component security vulnerability processing method according to any one of claims 1 to 9, It is characterized in that include: The information collection and preprocessing module is used to collect open source component information and vulnerability information from the project code repository and vulnerability disclosure platform, and to standardize the collected open source component information and vulnerability information to generate preliminary correlation data; Multi-dimensional graph generation and analysis module, used to generate multi-dimensional dependency graphs based on preliminary correlation data, generate verification dependency graphs through multi-dimensional graph generation and cross-correlation analysis, and dynamically evolve and optimize the dependency graph through an adaptive evolutionary algorithm. The verification dependency graph is updated in real time through a cross-graph feedback mechanism; The collaborative reasoning and data fusion module is used to perform distributed Bayesian reasoning and collaborative evolution based on the verification dependency graph. It uses intelligent agents to perform data reasoning and evolution at different levels, combines multi-level data fusion algorithms to deeply fuse data of different dimensions, generates fusion analysis data and game feedback data, and optimizes and shares strategies through a cross-layer multi-agent collaborative game model. The spatiotemporal prediction and repair strategy module is used to predict the time point and spatial location of future vulnerability outbreaks based on fusion analysis data through spatiotemporal regression models and cross-dimensional joint prediction models, generate cross-dimensional spatiotemporal prediction data, formulate preventive repair strategies including multiple repair stages, perform staged repair operations, and generate repair feedback data through real-time monitoring; The global optimization and learning module is used to make global optimization decisions based on cross-dimensional spatiotemporal prediction data and repair feedback data, and to perform system optimization and adaptive learning through feedback mechanisms, ultimately generating evolutionary learning model data.

Images