An artificial intelligence-based password application scheme verification and evaluation method

Abstract

The application discloses a password application scheme verification and evaluation method based on artificial intelligence, which comprises the following steps: obtaining an original feature data set, pre-processing the original feature data set to obtain a feature data set, weighting each feature data based on a gradient boosting method, and optimizing a decision tree structure; dynamically adjusting the weight of the feature data and the split node of each decision tree based on real-time environmental changes; optimizing the decision tree parameters, split nodes and leaf node settings through iterative training; evaluating the security of the password scheme in the actual environment according to the feature data of the password scheme, and outputting a security evaluation score; generating a security evaluation report and security improvement suggestions according to the security evaluation score; increasing the key length and optimizing the encryption protocol configuration, and further optimizing the password scheme according to the new security improvement suggestions. The application combines the improved XGBoost algorithm and dynamic feature weighting to provide accurate and real-time password scheme security evaluation and optimization.

Description

Technical Field

[0001] This invention relates to the field of cryptography, and in particular to a method for verifying and evaluating cryptographic application schemes based on artificial intelligence. Background Technology

[0002] With the rapid development of information technology, especially its widespread application in fields such as network communication, financial services, and the Internet of Things, data security and privacy protection have become increasingly important. Cryptography, as one of the core technologies of information security, plays a crucial role in protecting the confidentiality, integrity, and reliability of data. To ensure the security of cryptographic systems in various application scenarios, the verification and evaluation of cryptographic schemes have become key steps in ensuring their effectiveness. However, existing methods for verifying and evaluating cryptographic schemes often have certain limitations.

[0003] Traditional methods for verifying and evaluating cryptographic applications primarily rely on manual analysis or automated tools based on pre-defined rules. These methods typically perform static evaluations by analyzing the inherent characteristics of the cryptographic algorithm itself (such as encryption strength and key management). While this approach can provide a basic security assessment in some cases, its lack of adaptability to dynamic environments and real-time threats often makes it difficult to comprehensively evaluate the performance of cryptographic schemes in practical applications. Especially when facing complex attack scenarios or constantly changing network environments, traditional methods often fail to promptly reflect potential vulnerabilities and risks in cryptographic schemes, leading to security risks in practical applications.

[0004] Furthermore, existing evaluation methods have a significant drawback: they typically ignore changes in environmental factors during the evaluation process. Variations in network traffic, different hardware resource configurations, and fluctuations in system load directly affect the actual performance of cryptographic schemes, but traditional evaluation methods do not fully consider these dynamic factors. To address these issues, machine learning-based cryptographic scheme evaluation methods are gaining increasing attention. This is achieved by introducing machine learning techniques, such as the monotonic XGBoost algorithm model. However, despite the excellent performance of the XGBoost algorithm in many applications, existing technologies still lack specific optimizations for cryptographic schemes, making it unable to fully meet the needs of dynamic, real-time, and comprehensive security evaluation in practical applications.

[0005] Therefore, how to provide a method for verifying and evaluating cryptographic application schemes based on artificial intelligence is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0006] One objective of this invention is to propose an artificial intelligence-based method for verifying and evaluating cryptographic application schemes. This invention combines an improved XGBoost algorithm and dynamic feature weighting technology, providing accurate security assessments of cryptographic application schemes by adjusting feature weights and decision tree structures in real time. This method can automatically identify the features most relevant to the current environment based on changes in the actual environment, optimize encryption schemes, and improve the adaptability of cryptographic systems to new types of attacks. Simultaneously, through multi-dimensional security assessment and continuous optimization, it ensures the long-term security of cryptographic schemes in different application scenarios, providing comprehensive security guarantees for cryptographic systems.

[0007] A method for verifying and evaluating cryptographic application schemes based on artificial intelligence according to an embodiment of the present invention includes the following steps: After obtaining the original feature dataset of the cryptographic application scheme, preprocessing is performed to obtain the feature dataset; The feature dataset is input into the improved XGBoost algorithm model, each feature data is weighted based on the gradient boosting method, and the decision tree structure of the improved XGBoost algorithm model is optimized through each round of training in order to accurately evaluate the security of the cryptographic scheme. Based on real-time environmental changes, the weights of feature data are dynamically adjusted to automatically identify the features most relevant to the current application environment. The splitting nodes of each decision tree are adjusted according to the weights of these feature data to improve the accuracy of security assessment. By iteratively training and optimizing the decision tree parameters, split nodes, and leaf node settings in the XGBoost algorithm model, the model's adaptability to different security threats is improved, and its performance in cryptographic scheme security evaluation is enhanced. The improved XGBoost algorithm model after training is applied to real-time security assessment. Based on the feature data of the cryptographic scheme, the security of the cryptographic scheme in the real environment is evaluated, and the optimized security assessment score is output. A security assessment report is generated based on the optimized security assessment score. The security assessment report details the performance of the cryptographic scheme under various security dimensions, including security risks, vulnerability analysis and optimization suggestions, and generates specific security improvement suggestions. Based on the security improvement recommendations, the relevant security features in the cryptographic scheme are automatically adjusted, including increasing the key length and optimizing the encryption protocol configuration, and the cryptographic scheme is further optimized based on the new security improvement recommendations.

[0008] Optionally, the process of obtaining the original feature dataset of the cryptographic application scheme and then preprocessing it to obtain the feature dataset includes the following steps: The original feature dataset is obtained from the cryptographic application scheme. The original feature dataset includes encryption algorithm type, key length, network traffic, device hardware resource usage, system logs, user behavior data, and operating environment characteristics. The original feature dataset is cleaned to remove missing values, outliers, and duplicate data, ensuring the integrity and accuracy of the dataset. The cleaned original feature dataset is normalized to make the range of each feature data uniform and eliminate the impact of differences in the scale of different features on model training. Missing values ​​are imputed in the normalized original feature dataset using the mean imputation method to ensure the continuity and integrity of the dataset. By applying data standardization techniques, each original feature data after missing values ​​are filled is adjusted to the standard normal distribution range to eliminate bias in the data and ensure the stability of the model during training. Based on correlation analysis, the original feature data most relevant to the security assessment is selected, and redundant or irrelevant original feature data is removed. The original feature data is formatted and transformed into a format suitable for inputting the improved XGBoost algorithm model, forming the final feature dataset for cryptographic applications.

[0009] Optionally, the improved XGBoost algorithm model specifically includes: The decision tree construction module builds decision trees in each round of training and, based on the gradient boosting method, iteratively optimizes the structure of each decision tree by calculating the gradient of the loss function to gradually reduce the prediction error of the improved XGBoost algorithm model. This includes selecting the best split point and adjusting the depth of the decision tree to ensure that each tree can effectively reduce the prediction error and optimize the model's performance. The dynamic feature selection module calculates the information gain of each feature based on real-time environmental changes, selects the feature most relevant to the current environment, and automatically adjusts the feature weights to ensure that the decision tree can better adapt to environmental changes. Through dynamic weighting and feature selection, the model can more accurately reflect the impact of the environment on the data, thereby optimizing the security assessment. The regularization module introduces L1 and L2 regularization methods to control the complexity of the decision tree, limit the excessive growth of the tree depth, avoid overfitting, and optimize the generalization ability of the model. The regularization module works with other modules to adjust the complexity of the improved XGBoost algorithm model in each training process, so that the decision tree can maintain good adaptability to new data while training efficiently. The adaptive learning rate module dynamically adjusts the learning rate based on the performance of the improved XGBoost algorithm model in the current training, ensuring that the learning rate in each training round matches the model's performance. A higher learning rate accelerates the convergence of the improved XGBoost algorithm model, while a lower learning rate enhances the stability of the improved XGBoost algorithm model. This module ensures the accuracy and stability of model training through dynamic adjustment of the learning rate. The multi-objective optimization module is responsible for handling multi-dimensional evaluation objectives. When optimizing the parameters of the decision tree, it takes into account multiple objectives, including prediction accuracy, tree depth, feature selection quality, and regularization. During the optimization process, it minimizes the loss function and achieves a balance among multiple optimization objectives, ensuring that the model can reach the optimal state under multiple objectives and that the model can maintain good adaptability in complex environments. The output layer module combines the optimization results of all modules to output the final security assessment score and generates a detailed security assessment report based on the security assessment score. The security assessment report provides the assessment results of each security dimension and specific security improvement suggestions. By integrating the optimization outputs of each module, the output layer module ensures that the security assessment results of the cryptographic scheme in practical applications can be fully and accurately reflected.

[0010] Optionally, the decision tree structure of the XGBoost algorithm model, which weights each feature data based on the gradient boosting method and optimizes it through each round of training, specifically includes: The feature data is weighted by gradient boosting, and the contribution of each feature to the prediction error of the improved XGBoost algorithm model is calculated. The contribution is used to adjust the weight of the feature data, thereby increasing the influence of important features on the model and suppressing the interference of redundant features. The decision tree is trained using weighted feature data. The decision tree is optimized for each round based on the residuals. The prediction error of the improved XGBoost algorithm model in the previous round is gradually corrected. Through iterative training, the prediction accuracy of the improved XGBoost algorithm model is continuously improved. In each round of training, the gradient of the loss function is calculated, the splitting parameters of each node in the decision tree are updated, and a greedy algorithm is used to find the best splitting point to ensure that each tree can effectively divide the data and reduce the prediction error of the model. For each new decision tree, it is trained based on the residuals of the current model and then weighted and combined with the decision tree from the previous round to optimize the performance of the ensemble model and continuously improve the model’s performance in cryptographic scheme security evaluation. In multiple iterations, the contribution of the decision tree in each round is controlled by adjusting the learning rate, ensuring that the model can converge quickly during training while preventing overfitting and improving the model's generalization ability.

[0011] Optionally, the step of dynamically adjusting the weights of feature data based on real-time environmental changes, automatically identifying the features most relevant to the current application environment, and adjusting the splitting nodes of each decision tree according to the weights of these feature data specifically includes: The system acquires characteristic data of the current application environment through real-time data monitoring, including network traffic, hardware resource usage, system load and user behavior data, as well as security characteristic data related to cryptographic application schemes. The real-time acquired feature data is analyzed to calculate the relevance and importance of each feature in the current environment, and the relevance analysis method is used to automatically identify the feature most relevant to the current application environment. The feature data is dynamically weighted, and the weight of each feature is calculated. The weight is adjusted according to the degree of influence of the feature on the security assessment in the current environment, so as to ensure that the model pays attention to key features in different environments. The weighted feature data is input into the improved XGBoost algorithm model. In each training round, the splitting nodes of the decision tree are optimized based on the real-time adjusted feature weights, and features that can maximize information gain or reduce error are automatically selected for data partitioning. Optionally, the settings for decision tree parameters, split nodes, and leaf nodes in the XGBoost algorithm model optimized through iterative training specifically include: An initial decision tree structure is established by initializing the improved XGBoost algorithm model, where each decision tree is trained using the gradient boosting method. In each round of iterative training, based on the residual of the current model, each decision tree is optimized, the gradient of each feature is calculated, and the best split point of the split node in each tree is determined to reduce the prediction error of the model. Optimize the leaf node settings of each decision tree, and adjust the weight of the leaf node based on the prediction output of each leaf node, so that the output of the final leaf node can match the target value of the training data as accurately as possible, thereby improving the overall prediction performance of the model. During each training round, the depth and number of splits of the decision tree are dynamically adjusted by calculating the gradient information of the loss function to avoid overfitting or underfitting and ensure that the model has strong generalization ability on different datasets. In iterative training, regularization techniques are introduced to constrain each decision tree, control the complexity of the tree, prevent over-growing trees from overfitting the training data, and improve the stability of the model. Setting the learning rate controls the contribution of the new decision tree to the improved XGBoost algorithm model in each training round, avoiding the excessive influence of a single tree on the model. By adjusting the balance between the learning rate and the number of training rounds, the overall training process is optimized to ensure that the model converges at an appropriate speed. Based on the optimization results of each round of training, the decision tree structure in the improved XGBoost algorithm model is adjusted, including the depth of the decision tree, the setting of split nodes, leaf nodes, and training parameters, to gradually improve the model's performance in cryptographic scheme security evaluation, and finally obtain the optimized decision tree set.

[0012] Optionally, the step of evaluating the security of the cryptographic scheme in a real-world environment based on its feature data and outputting an optimized security assessment score specifically includes: The model evaluates cryptographic schemes based on the weight of each feature and the decision tree structure; Based on the prediction results output by the improved XGBoost algorithm model, and considering the changes in feature data and the current environment, the security of cryptographic schemes in different application environments is evaluated, and a security score is calculated. The security score represents the overall security level of the cryptographic scheme in the current environment. Based on the security score of the cryptographic scheme and related security risk factors, including attack threats and vulnerability severity, a separate evaluation score is generated for each security dimension, including encryption strength, key management and protocol stability. The evaluation scores of each dimension are comprehensively processed to generate the final optimized security evaluation score. The security evaluation score is a weighted average based on the weight of each evaluation dimension and the evaluation results.

[0013] Optionally, the automatic adjustment of relevant security features in the cryptographic scheme based on security improvement recommendations includes increasing the key length and selecting encryption algorithms or optimizing encryption protocol configurations, and further optimizing the cryptographic scheme based on new security improvement recommendations, specifically including: Based on the security improvement suggestions in the security assessment report, the system automatically identifies security weaknesses in the cryptographic scheme and generates relevant optimization strategies based on the characteristic that the security assessment score is lower than a predetermined threshold. Based on the generated optimization strategies, the key length in the cryptographic scheme is automatically increased to enhance the key's resistance to cracking. The increase in key length is automatically selected based on the evaluation results to ensure that the security of the cryptographic scheme meets the latest security standards. The encryption protocol is optimized and configured. If the assessment report shows that the current encryption protocol has vulnerabilities or is not suitable for new attack methods, the encryption protocol settings are automatically optimized to enhance the protocol's resistance to attacks, adopt a stronger authentication mechanism, and increase data integrity protection. By using automated tools to update key management and protocol configuration in cryptographic schemes, the optimized cryptographic schemes can be automatically applied to the cryptographic system, ensuring that new security improvements can be quickly deployed and implemented in real-world environments. After updating the configuration, the security assessment was re-performed using the improved XGBoost algorithm model, and an optimized security score was generated. Based on the new assessment results and security improvement recommendations, we will continue to iterate and optimize the security features of the cryptographic scheme to achieve continuous improvement and long-term optimization, ensuring that the cryptographic scheme is always in the best state of protection against potential future security threats.

[0014] The beneficial effects of this invention are: 1. This invention significantly improves the security assessment accuracy of cryptographic application schemes by introducing an improved XGBoost algorithm combined with the powerful predictive capabilities of gradient boosting tree models. The improved XGBoost algorithm can dynamically weight and iteratively train based on the input feature data. By continuously optimizing the decision tree structure and adjusting the split nodes of each tree, the model can better adapt to different security threats and application environments. This data-driven security assessment method not only reduces the subjective bias caused by traditional manual analysis but also greatly improves the objectivity and credibility of the assessment results.

[0015] 2. This invention, through dynamic adjustment of feature data and real-time environment sensitivity optimization, can respond in real-time to changes in network, hardware, and system environments. Traditional methods often cannot effectively handle complex dynamic changes, while the evaluation method of this invention ensures that the evaluation process always reflects the security in practical applications by adjusting feature weights in real time and identifying the most relevant features in the current environment. This dynamic adaptability enables cryptographic schemes to automatically optimize when faced with new network attacks or changes in hardware resources, thereby improving the security protection capabilities of cryptographic systems.

[0016] 3. This invention incorporates a multi-objective optimization module during model training, enabling evaluation to go beyond a single dimension of cryptographic security (such as encryption strength) and comprehensively consider multiple security dimensions, such as key management, protocol stability, network traffic, and system load. This multi-dimensional evaluation approach provides a more comprehensive security assessment for cryptographic applications, effectively identifying potential security vulnerabilities and offering targeted improvement suggestions. Through this method, this invention can provide more accurate and comprehensive security assessment reports for cryptographic systems in actual deployment, helping developers to promptly identify and resolve potential security vulnerabilities.

[0017] 4. This invention not only provides accurate security assessments in the initial evaluation stage, but also enables continuous optimization of the cryptographic scheme in practical applications through a continuous iterative optimization process. As network environments and attack patterns constantly change, the security requirements of cryptographic systems also evolve. Through feedback from each security assessment, the system can automatically adjust key features of the cryptographic scheme (such as increasing key length, optimizing encryption protocol configuration, etc.) based on new security improvement suggestions, ensuring that the cryptographic scheme always remains in an optimal security state. Attached Figure Description

[0018] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is an overall flowchart of a method for verifying and evaluating cryptographic application schemes based on artificial intelligence, as proposed in this invention. Figure 2 This is a flowchart illustrating the decision tree optimization and training process for a method for verifying and evaluating cryptographic application schemes based on artificial intelligence, as proposed in this invention. Figure 3 This is a schematic diagram of the improved XGBoost algorithm structure for a cryptographic application scheme verification and evaluation method based on artificial intelligence proposed in this invention. Detailed Implementation

[0019] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0020] refer to Figure 1-3 A method for verifying and evaluating cryptographic application schemes based on artificial intelligence includes the following steps: After obtaining the original feature dataset of the cryptographic application scheme, preprocessing is performed to obtain the feature dataset; The feature dataset is input into the improved XGBoost algorithm model, each feature data is weighted based on the gradient boosting method, and the decision tree structure of the improved XGBoost algorithm model is optimized through each round of training in order to accurately evaluate the security of the cryptographic scheme. Based on real-time environmental changes, the weights of feature data are dynamically adjusted to automatically identify the features most relevant to the current application environment. The splitting nodes of each decision tree are adjusted according to the weights of these feature data to improve the accuracy of security assessment. By iteratively training and optimizing the decision tree parameters, split nodes, and leaf node settings in the XGBoost algorithm model, the model's adaptability to different security threats is improved, and its performance in cryptographic scheme security evaluation is enhanced. The improved XGBoost algorithm model after training is applied to real-time security assessment. Based on the feature data of the cryptographic scheme, the security of the cryptographic scheme in the real environment is evaluated, and the optimized security assessment score is output. A security assessment report is generated based on the optimized security assessment score. The security assessment report details the performance of the cryptographic scheme under various security dimensions, including security risks, vulnerability analysis and optimization suggestions, and generates specific security improvement suggestions. Based on the security improvement recommendations, the relevant security features in the cryptographic scheme are automatically adjusted, including increasing the key length and optimizing the encryption protocol configuration, and the cryptographic scheme is further optimized based on the new security improvement recommendations.

[0021] This invention utilizes an improved XGBoost algorithm for dynamic and security evaluation of cryptographic application schemes. Employing a gradient-boosting-based decision tree optimization method, it automatically performs real-time security analysis of cryptographic schemes and adjusts feature data weights according to different application scenarios, resulting in a more accurate evaluation process. Continuous optimization of the decision tree structure reduces prediction errors and improves the accuracy and adaptability of cryptographic system security evaluation. Furthermore, through iterative training and optimization of the XGBoost algorithm model, this invention ensures that the cryptographic scheme reflects the security of the current application environment in real time, avoiding the problem of traditional evaluation methods failing to adapt to attack patterns and environmental changes in real time.

[0022] In this embodiment, the process of obtaining the original feature dataset of the cryptographic application scheme and then preprocessing it to obtain the feature dataset includes the following steps: The original feature dataset is obtained from the cryptographic application scheme. The original feature dataset includes encryption algorithm type, key length, network traffic, device hardware resource usage, system logs, user behavior data, and operating environment characteristics. The original feature dataset is cleaned to remove missing values, outliers, and duplicate data, ensuring the integrity and accuracy of the dataset. The cleaned original feature dataset is normalized to make the range of each feature data uniform and eliminate the impact of differences in the scale of different features on model training. Missing values ​​are imputed in the normalized original feature dataset using the mean imputation method to ensure the continuity and integrity of the dataset. By applying data standardization techniques, each original feature data after missing values ​​are filled is adjusted to the standard normal distribution range to eliminate bias in the data and ensure the stability of the model during training. Based on correlation analysis, the original feature data most relevant to the security assessment is selected, and redundant or irrelevant original feature data is removed. The original feature data is formatted and transformed into a format suitable for inputting the improved XGBoost algorithm model, forming the final feature dataset for cryptographic applications.

[0023] This invention's innovative steps in feature data preprocessing combine multiple steps, including data cleaning, standardization, and missing value imputation, ensuring the integrity and accuracy of the input data. This method effectively removes redundant information and outlier data, eliminates dimensional differences between different features, and employs a scientific missing value imputation strategy to guarantee the quality of the training dataset. Through this comprehensive preprocessing approach, this invention provides a high-quality feature dataset, offering an accurate and stable data foundation for subsequent security assessments and significantly improving the reliability of cryptographic scheme security assessment results.

[0024] In this embodiment, the improved XGBoost algorithm model specifically includes: The decision tree construction module builds decision trees in each round of training and, based on the gradient boosting method, iteratively optimizes the structure of each decision tree by calculating the gradient of the loss function to gradually reduce the prediction error of the improved XGBoost algorithm model. This includes selecting the best split point and adjusting the depth of the decision tree to ensure that each tree can effectively reduce the prediction error and optimize the model's performance. The dynamic feature selection module calculates the information gain of each feature based on real-time environmental changes, selects the feature most relevant to the current environment, and automatically adjusts the feature weights to ensure that the decision tree can better adapt to environmental changes. Through dynamic weighting and feature selection, the model can more accurately reflect the impact of the environment on the data, thereby optimizing the security assessment. The regularization module introduces L1 and L2 regularization methods to control the complexity of the decision tree, limit the excessive growth of the tree depth, avoid overfitting, and optimize the generalization ability of the model. The regularization module works with other modules to adjust the complexity of the improved XGBoost algorithm model in each training process, so that the decision tree can maintain good adaptability to new data while training efficiently. The adaptive learning rate module dynamically adjusts the learning rate based on the performance of the improved XGBoost algorithm model in the current training, ensuring that the learning rate in each training round matches the model's performance. A higher learning rate accelerates the convergence of the improved XGBoost algorithm model, while a lower learning rate enhances the stability of the improved XGBoost algorithm model. This module ensures the accuracy and stability of model training through dynamic adjustment of the learning rate. The multi-objective optimization module is responsible for handling multi-dimensional evaluation objectives. When optimizing the parameters of the decision tree, it takes into account multiple objectives, including prediction accuracy, tree depth, feature selection quality, and regularization. During the optimization process, it minimizes the loss function and achieves a balance among multiple optimization objectives, ensuring that the model can reach the optimal state under multiple objectives and that the model can maintain good adaptability in complex environments. The output layer module combines the optimization results of all modules to output the final security assessment score and generates a detailed security assessment report based on the security assessment score. The security assessment report provides the assessment results of each security dimension and specific security improvement suggestions. By integrating the optimization outputs of each module, the output layer module ensures that the security assessment results of the cryptographic scheme in practical applications can be fully and accurately reflected.

[0025] This invention improves the XGBoost algorithm model by adding dynamic feature selection, regularization, and multi-objective optimization modules, significantly enhancing the accuracy and stability of cryptographic scheme evaluation. By automatically identifying the features most relevant to the current environment and adjusting weights according to real-time changes, the model can better adapt to complex security environment changes. Simultaneously, the introduction of L1 and L2 regularization techniques avoids overfitting and improves the model's generalization ability when facing unknown data. The multi-objective optimization module ensures security evaluation across multiple dimensions, thus providing a comprehensive security evaluation system for cryptographic schemes.

[0026] In this embodiment, the decision tree structure of the XGBoost algorithm model, which weights each feature data based on the gradient boosting method and optimizes it through each round of training, specifically includes: Using gradient boosting methods: ; in, Let i be the predicted value of the i-th sample in the t-th iteration. Let i be the predicted value of the i-th sample in the (t-1)-th iteration. The learning rate controls the magnitude of model updates in each iteration. The output of the decision tree model in round t is based on the input feature data. The calculated residuals.

[0027] The feature data is weighted, and the contribution of each feature to the prediction error of the improved XGBoost algorithm model is calculated. The contribution is then used to adjust the weights of the feature data. ; in, Let the weight of the i-th feature be the updated weight in the (t+1)-th round of training. Let i be the current weight of the i-th feature in the t-th round of training. This is the regularization coefficient, used to control overfitting and limit the excessive growth of feature weights. Let i be the value of the i-th feature on the j-th sample. Let be the gradient of the i-th feature, representing the contribution of that feature to the model error. ,in, The loss function measures the true value. and predicted value To address the differences between these features, this invention uses the mean squared error loss function. This enhances the influence of important features on the model and suppresses the interference of redundant features. The decision tree is trained using weighted feature data. The decision tree is optimized for each round based on the residuals. The prediction error of the improved XGBoost algorithm model in the previous round is gradually corrected. Through iterative training, the prediction accuracy of the improved XGBoost algorithm model is continuously improved. In each round of training, the splitting parameters of each node in the decision tree are updated by calculating the gradient of the loss function, and a greedy algorithm is used to find the optimal splitting point: ; in, The optimal split point is defined by argmin(), where argmin() is the independent variable that minimizes the function. Let be the predicted value of the i-th sample at the split point t. The total number of samples represents the number of samples in the training dataset. The regularization coefficient controls the complexity of the model; a larger one... The value limits the complexity of the model and prevents the decision tree from growing excessively. The parameters of the decision tree in the k-th round are... The total number of decision trees. This is a regularization term used to control model complexity and prevent overfitting. It ensures that each tree can effectively partition the data, reducing the model's prediction error. For each new decision tree, it is trained based on the residuals of the current model and then weighted and combined with the decision tree from the previous round to optimize the performance of the ensemble model and continuously improve the model’s performance in cryptographic scheme security evaluation. In multiple iterations, the contribution of the decision tree in each round is controlled by adjusting the learning rate, ensuring that the model can converge quickly during training while preventing overfitting and improving the model's generalization ability.

[0028] This invention employs a gradient boosting method to weight each feature data point and optimizes the decision tree structure in each training round, ensuring the evaluation model accurately captures key features. During each iteration, the model dynamically adjusts feature weights by calculating the contribution of each feature to the prediction error, thereby enabling effective security evaluation of cryptographic schemes in various application environments. By optimizing the splitting nodes and leaf nodes of the decision tree, this invention further improves the model's prediction accuracy while reducing errors and maintaining computational efficiency, significantly enhancing the evaluation quality of cryptographic schemes.

[0029] In this embodiment, the step of dynamically adjusting the weights of feature data based on real-time environmental changes, automatically identifying the features most relevant to the current application environment, and adjusting the split nodes of each decision tree according to the weights of these feature data specifically includes: The system acquires characteristic data of the current application environment through real-time data monitoring, including network traffic, hardware resource usage, system load and user behavior data, as well as security characteristic data related to cryptographic application schemes. The real-time acquired feature data is analyzed to calculate the relevance and importance of each feature in the current environment, and the relevance analysis method is used to automatically identify the feature most relevant to the current application environment. The feature data is dynamically weighted, and the weight of each feature is calculated. The weight is adjusted according to the degree of influence of the feature on the security assessment in the current environment, so as to ensure that the model pays attention to key features in different environments. The weighted feature data is input into the improved XGBoost algorithm model. In each training round, the splitting nodes of the decision tree are optimized based on the real-time adjusted feature weights, and features that can maximize information gain or reduce error are automatically selected for data partitioning. This invention addresses the problem of existing evaluation methods neglecting environmental changes by dynamically monitoring real-time environmental changes, automatically identifying the features most relevant to the current environment, and adjusting the weights of the feature data. By weighting each feature based on relevance and environmental dependence, this invention can provide optimal security assessments in different application scenarios. The dynamism and flexibility of this method enable cryptographic schemes to respond to new security threats and cyberattacks in real time, thereby significantly improving the security protection capabilities of cryptographic systems in practical applications.

[0030] In this embodiment, the settings of decision tree parameters, split nodes, and leaf nodes in the XGBoost algorithm model optimized through iterative training specifically include: An initial decision tree structure is established by initializing the improved XGBoost algorithm model, where each decision tree is trained using the gradient boosting method. In each round of iterative training, based on the residual of the current model, each decision tree is optimized, the gradient of each feature is calculated, and the best split point of the split node in each tree is determined to reduce the prediction error of the model. Optimize the leaf node settings of each decision tree, and adjust the weight of the leaf node based on the prediction output of each leaf node, so that the output of the final leaf node can match the target value of the training data as accurately as possible, thereby improving the overall prediction performance of the model. During each training round, the depth and number of splits of the decision tree are dynamically adjusted by calculating the gradient information of the loss function to avoid overfitting or underfitting and ensure that the model has strong generalization ability on different datasets. In iterative training, regularization techniques are introduced to constrain each decision tree, control the complexity of the tree, prevent over-growing trees from overfitting the training data, and improve the stability of the model. Setting the learning rate controls the contribution of the new decision tree to the improved XGBoost algorithm model in each training round, avoiding the excessive influence of a single tree on the model. By adjusting the balance between the learning rate and the number of training rounds, the overall training process is optimized to ensure that the model converges at an appropriate speed. Based on the optimization results of each round of training, the decision tree structure in the improved XGBoost algorithm model is adjusted, including the depth of the decision tree, the setting of split nodes, leaf nodes, and training parameters, to gradually improve the model's performance in cryptographic scheme security evaluation, and finally obtain the optimized decision tree set.

[0031] This invention optimizes the decision tree parameters, split nodes, and leaf node settings in the XGBoost algorithm through iterative training, enabling the decision tree to better adapt to data changes after each training round. By dynamically adjusting the structural parameters of each tree, overfitting or underfitting is avoided, ensuring accurate fitting of the model to the training data and good generalization ability to unknown data. Optimizing the settings of nodes and leaf nodes during each training process improves the model's accuracy and stability, enhances the anti-interference capability of cryptographic applications under different attack modes, and ensures long-term security of the cryptographic system.

[0032] In this embodiment, the step of evaluating the security of the cryptographic scheme in a real-world environment based on its feature data and outputting an optimized security assessment score specifically includes: The model evaluates cryptographic schemes based on the weight of each feature and the decision tree structure; Based on the prediction results output by the improved XGBoost algorithm model, and considering the changes in feature data and the current environment, the security of cryptographic schemes in different application environments is evaluated, and a security score is calculated. The security score represents the overall security level of the cryptographic scheme in the current environment. Based on the security score of the cryptographic scheme and related security risk factors, including attack threats and vulnerability severity, a separate evaluation score is generated for each security dimension, including encryption strength, key management and protocol stability. The evaluation scores of each dimension are comprehensively processed to generate the final optimized security evaluation score. The security evaluation score is a weighted average based on the weight of each evaluation dimension and the evaluation results.

[0033] This invention employs a comprehensive evaluation method, weighting and averaging the scores of each security dimension to generate a final optimized security assessment score. This method combines multiple security dimensions, including encryption strength, key management, and protocol stability, comprehensively considering potential risks and vulnerabilities to provide a holistic security assessment for cryptographic systems. Furthermore, by incorporating the security assessment score output by the XGBoost algorithm, it can offer users specific improvement suggestions, such as enhancing key strength and improving encryption protocols, providing practical and feasible security optimization solutions for cryptographic schemes.

[0034] In this embodiment, automatically adjusting the relevant security features in the cryptographic scheme according to security improvement suggestions includes increasing the key length and selecting encryption algorithms or optimizing encryption protocol configurations, and further optimizing the cryptographic scheme according to new security improvement suggestions, specifically including: Based on the security improvement suggestions in the security assessment report, the system automatically identifies security weaknesses in the cryptographic scheme and generates relevant optimization strategies based on the characteristic that the security assessment score is lower than a predetermined threshold. Based on the generated optimization strategies, the key length in the cryptographic scheme is automatically increased to enhance the key's resistance to cracking. The increase in key length is automatically selected based on the evaluation results to ensure that the security of the cryptographic scheme meets the latest security standards. The encryption protocol is optimized and configured. If the assessment report shows that the current encryption protocol has vulnerabilities or is not suitable for new attack methods, the encryption protocol settings are automatically optimized to enhance the protocol's resistance to attacks, adopt a stronger authentication mechanism, and increase data integrity protection. By using automated tools to update key management and protocol configuration in cryptographic schemes, the optimized cryptographic schemes can be automatically applied to the cryptographic system, ensuring that new security improvements can be quickly deployed and implemented in real-world environments. After updating the configuration, the security assessment was re-performed using the improved XGBoost algorithm model, and an optimized security score was generated. Based on the new assessment results and security improvement recommendations, we will continue to iterate and optimize the security features of the cryptographic scheme to achieve continuous improvement and long-term optimization, ensuring that the cryptographic scheme is always in the best state of protection against potential future security threats.

[0035] This invention automates the adjustment of key security features in a cryptographic scheme, such as increasing key length and selecting stronger encryption algorithms, based on feedback from security assessment reports. This process allows the invention to rapidly respond to and optimize the cryptographic scheme based on real-time assessment results, avoiding the inefficiency of traditional manual configuration modifications. This automated optimization not only enhances the security of the cryptographic system but also significantly improves implementation efficiency. Furthermore, as new security assessment results emerge, the system can continuously optimize and adjust, ensuring the cryptographic scheme remains in optimal protection against ever-changing attack environments.

[0036] Example 1: To verify the feasibility of this invention in practice, it was applied to an internet finance platform. This platform employs encryption protocols to protect the security of user data transmission, ensuring data confidentiality and integrity during large-scale user data processing and high-frequency data exchange. With the growth of the platform's user base and the continuous evolution of network attack methods, traditional static security assessment methods can no longer meet the needs of real-time assessment and rapid response. To enhance the system's adaptability to different security threats, this invention provides a dynamic security assessment method based on an improved XGBoost algorithm. This method can automatically evaluate and optimize cryptographic schemes in a constantly changing network environment, and promptly detect potential security vulnerabilities.

[0037] In the platform's production environment, a multimodal data acquisition system enables the platform to acquire real-time feature data on cryptographic application schemes, including encryption algorithm type, key length, network traffic, device hardware resource usage, system logs, user behavior data, and operating environment characteristics. The collected raw data undergoes data cleaning, standardization, and missing value imputation to form a standardized feature dataset. This processed data is then fed into a security assessment model based on an improved XGBoost algorithm. This model not only trains on the existing feature data but also optimizes the decision tree structure in each training iteration and adjusts the feature weights in each iteration to respond in real-time to changes in network traffic, system load fluctuations, and potential security threats.

[0038] During training, the model uses gradient boosting to weight each feature data point and optimizes the parameters, split nodes, and leaf nodes of the decision tree in each iteration, ensuring that the model continuously improves the accuracy of security assessments in dynamic environments. Once training is complete, the platform can evaluate the security of the encryption protocol in real time based on the model's output, generating a detailed security assessment report. The report not only lists the performance across various security dimensions (such as encryption strength, key management, and encryption protocol configuration) but also provides targeted security improvement suggestions, such as increasing key length, switching encryption algorithms, and optimizing protocol configuration.

[0039] To verify the effectiveness of the method of this invention, the platform underwent a three-month test involving 50 million records. During the test, the evaluation method based on the improved XGBoost algorithm was compared with the traditional static evaluation method using manual inspection and fixed rules. The test results show that the security evaluation method based on the improved XGBoost algorithm has significantly higher evaluation accuracy than the traditional method in terms of encryption strength, key management, and encryption protocol stability. For example, in the encryption strength evaluation, the method of this invention scored 88, compared to 75 for the traditional method, an improvement of 17.33%; in the key management evaluation, the method of this invention scored 85, compared to 70 for the traditional method, an improvement of 21.43%; and in the encryption protocol stability evaluation, the method of this invention scored 78, compared to 60 for the traditional method, an improvement of 30.00%. A comparison of the cryptographic scheme security evaluations is shown in Table 1.

[0040] Table 1 Comparison of Cryptographic Scheme Security Assessments

[0041] As shown in Table 1, the evaluation method based on the improved XGBoost algorithm outperforms the traditional evaluation method across all security evaluation dimensions, particularly in cryptographic protocol stability evaluation, key management evaluation, and system load impact evaluation, with improvements of 30.00%, 21.43%, and 26.15%, respectively. These significant improvements indicate that the improved XGBoost algorithm can more accurately identify and evaluate the performance of cryptographic schemes under different security threats. Especially when facing complex attack patterns and dynamic network environments, it can adjust the evaluation criteria in a timely manner, thereby providing more accurate and real-time evaluation results.

[0042] Traditional evaluation methods have significant shortcomings in key dimensions such as encryption protocol stability and system load impact. Because traditional methods typically rely on manual checks or fixed rules, they cannot respond in real-time to changes in system load and the evolution of network attack patterns, thus failing to effectively reflect potential risks in the current environment. In contrast, the evaluation method based on the improved XGBoost algorithm of this invention, by dynamically adjusting feature weights, can accurately capture these changes and provide more precise security evaluation results. The introduction of this method not only enhances the ability of cryptographic systems to cope with complex attack scenarios but also ensures the long-term security of cryptographic schemes in constantly changing network environments.

[0043] As shown in Table 1, the evaluation method based on the improved XGBoost algorithm performs excellently across multiple security dimensions, particularly in encryption protocol stability, key management, and encryption strength evaluation, where significant improvements are achieved. This comparative analysis verifies the effectiveness of this invention in practical applications, demonstrating that the method has significant advantages. It can provide real-time and accurate security assessments for cryptographic systems and automatically generate targeted security improvement suggestions, thereby greatly enhancing the security protection capabilities of cryptographic schemes and ensuring long-term security during data transmission.

[0044] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for verifying and evaluating cryptographic application schemes based on artificial intelligence, characterized in that, Includes the following steps: After obtaining the original feature dataset of the cryptographic application scheme, preprocessing is performed to obtain the feature dataset; The feature dataset is input into the improved XGBoost algorithm model, each feature data is weighted based on the gradient boosting method, and the decision tree structure of the improved XGBoost algorithm model is optimized through each round of training. Based on real-time environmental changes, the weights of feature data are dynamically adjusted, the features most relevant to the current application environment are automatically identified, and the split nodes of each decision tree are adjusted according to the weights of these feature data. The decision tree parameters, split nodes, and leaf node settings in the XGBoost algorithm model were optimized through iterative training. The improved XGBoost algorithm model after training is applied to real-time security assessment. Based on the feature data of the cryptographic scheme, the security of the cryptographic scheme in the real environment is evaluated, and the optimized security assessment score is output. A security assessment report is generated based on the optimized security assessment score. The security assessment report details the performance of the cryptographic scheme under various security dimensions, including security risks, vulnerability analysis and optimization suggestions, and generates specific security improvement suggestions. Based on the security improvement recommendations, the relevant security features in the cryptographic scheme are automatically adjusted, including increasing the key length and optimizing the encryption protocol configuration, and the cryptographic scheme is further optimized based on the new security improvement recommendations.

2. The method for verifying and evaluating cryptographic application schemes based on artificial intelligence according to claim 1, characterized in that, The process of obtaining the original feature dataset of the cryptographic application scheme and then preprocessing it to obtain the feature dataset includes the following steps: The original feature dataset is obtained from the cryptographic application scheme. The original feature dataset includes encryption algorithm type, key length, network traffic, device hardware resource usage, system logs, user behavior data, and operating environment characteristics. The original feature dataset is cleaned to remove missing values, outliers, and duplicate data. The cleaned original feature dataset is then normalized. Missing values ​​are imputed in the normalized original feature dataset by using the mean imputation method. By applying data standardization techniques, each original feature data, after missing values ​​are filled, is adjusted to the standard normal distribution range; Based on correlation analysis, the original feature data most relevant to the security assessment is selected, and redundant or irrelevant original feature data is removed. The original feature data is formatted and transformed into a format suitable for inputting the improved XGBoost algorithm model, forming the final feature dataset for cryptographic applications.

3. The method for verifying and evaluating cryptographic application schemes based on artificial intelligence according to claim 1, characterized in that, The improved XGBoost algorithm model includes a decision tree construction module, a dynamic feature selection module, a regularization module, an adaptive learning rate module, a multi-objective optimization module, and an output layer module. The decision tree construction module constructs a decision tree in each round of training, and iteratively optimizes the structure of each decision tree based on the gradient boosting method, and calculates the gradient of the loss function. The dynamic feature selection module calculates the information gain of each feature based on real-time environmental changes, selects the feature most relevant to the current environment, and automatically adjusts the feature weights. The regularization module introduces L1 and L2 regularization methods and adjusts the complexity of the improved XGBoost algorithm model during each training process. The adaptive learning rate module dynamically adjusts the learning rate based on the performance of the improved XGBoost algorithm model in the current training. The multi-objective optimization module is responsible for handling multi-dimensional evaluation objectives. When optimizing the parameters of the decision tree, it takes into account multiple objectives, including prediction accuracy, tree depth, feature selection quality, and regularization, and minimizes the loss function during the optimization process. The output layer module combines the optimization results of all modules to output the final security assessment score, and generates a detailed security assessment report based on the security assessment score. The security assessment report provides the assessment results of each security dimension and specific security improvement suggestions.

4. The method for verifying and evaluating cryptographic application schemes based on artificial intelligence according to claim 1, characterized in that, The decision tree structure of the XGBoost algorithm model, which weights each feature data using the gradient boosting method and optimizes it through each round of training, specifically includes: The feature data is weighted using the gradient boosting method, and the contribution of each feature to the prediction error of the improved XGBoost algorithm model is calculated. The contribution is then used to adjust the weights of the feature data. The decision tree is trained using weighted feature data. The decision tree is optimized for each round based on the residuals. The prediction error of the improved XGBoost algorithm model in the previous round is gradually corrected and iterative training is performed. In each round of training, the splitting parameters of each node in the decision tree are updated by calculating the gradient of the loss function, and a greedy algorithm is used to find the optimal splitting point. For each new decision tree, it is trained based on the residuals of the current model and then weighted and combined with the decision tree from the previous round. In multiple iterations, the contribution of the decision tree in each round is controlled by adjusting the learning rate.

5. The method for verifying and evaluating cryptographic application schemes based on artificial intelligence according to claim 1, characterized in that, The process of dynamically adjusting the weights of feature data based on real-time environmental changes, automatically identifying the features most relevant to the current application environment, and adjusting the splitting nodes of each decision tree according to the weights of these feature data specifically includes: The system acquires characteristic data of the current application environment through real-time data monitoring, including network traffic, hardware resource usage, system load and user behavior data, as well as security characteristic data related to cryptographic application schemes. The real-time acquired feature data is analyzed to calculate the relevance and importance of each feature in the current environment, and the relevance analysis method is used to automatically identify the feature most relevant to the current application environment. The feature data is dynamically weighted, and the weight of each feature is calculated. The weight is adjusted according to the degree of influence of the feature on the security assessment in the current environment. The weighted feature data is input into the improved XGBoost algorithm model. In each training round, the splitting nodes of the decision tree are optimized based on the real-time adjusted feature weights, and features that can maximize information gain or reduce error are automatically selected for data partitioning.

6. The method for verifying and evaluating cryptographic application schemes based on artificial intelligence according to claim 1, characterized in that, The settings for decision tree parameters, split nodes, and leaf nodes in the XGBoost algorithm model, which is optimized and improved through iterative training, specifically include: An initial decision tree structure is established by initializing the improved XGBoost algorithm model, where each decision tree is trained using the gradient boosting method. In each round of iterative training, based on the residual of the current model, each decision tree is optimized, the gradient of each feature is calculated, and the optimal split point of the split node in each tree is determined. Optimize the leaf node settings of each decision tree, and adjust the weight of the leaf nodes based on the predicted output of each leaf node, so that the output of the final leaf node can match the target value of the training data as accurately as possible. During each training round, the depth and number of splits of the decision tree are dynamically adjusted by calculating the gradient information of the loss function. In iterative training, regularization techniques are introduced to constrain each decision tree and control the complexity of the tree. The learning rate is set to control the contribution of the new decision tree to the improved XGBoost algorithm model in each training round, and the overall training process is optimized by adjusting the balance between the learning rate and the number of training rounds. Based on the optimization results of each training round, the decision tree structure in the improved XGBoost algorithm model is adjusted, including the depth of the decision tree, the setting of split nodes, leaf nodes, and training parameters.

7. The method for verifying and evaluating cryptographic application schemes based on artificial intelligence according to claim 1, characterized in that, The process of evaluating the security of a cryptographic scheme in a real-world environment based on its characteristic data and outputting an optimized security assessment score specifically includes: The cryptographic scheme is evaluated based on the weight of each feature and the decision tree structure; Based on the prediction results output by the improved XGBoost algorithm model, and considering the changes in feature data and the current environment, the security of cryptographic schemes in different application environments is evaluated, and a security score is calculated. The security score represents the overall security level of the cryptographic scheme in the current environment. Based on the security score of the cryptographic scheme and related security risk factors, including attack threats and vulnerability severity, a separate evaluation score is generated for each security dimension, including encryption strength, key management and protocol stability. The evaluation scores of each dimension are comprehensively processed to generate the final optimized security evaluation score. The security evaluation score is a weighted average based on the weight of each evaluation dimension and the evaluation results.

8. The method for verifying and evaluating cryptographic application schemes based on artificial intelligence according to claim 1, characterized in that, The automatic adjustment of relevant security features in the cryptographic scheme based on security improvement recommendations includes increasing the key length and selecting encryption algorithms or optimizing encryption protocol configurations, and further optimizing the cryptographic scheme based on the new security improvement recommendations, specifically including: Based on the security improvement suggestions in the security assessment report, the system automatically identifies security weaknesses in the cryptographic scheme and generates relevant optimization strategies based on the characteristic that the security assessment score is lower than a predetermined threshold. Based on the generated optimization strategy, the key length in the cryptographic scheme is automatically increased, and the increase in key length is automatically selected based on the evaluation results; The encryption protocol is optimized and configured. If the assessment report shows that the current encryption protocol has vulnerabilities or is not suitable for new attack methods, the encryption protocol settings are automatically optimized, and a stronger authentication mechanism and increased data integrity protection are adopted. By using automated tools to update key management and protocol configuration in cryptographic schemes, the optimized cryptographic schemes can be automatically applied to the cryptographic system. After updating the configuration, the security assessment was re-performed using the improved XGBoost algorithm model, and an optimized security score was generated. Based on the new assessment results and security improvement recommendations, we will continue to iterate and optimize the security features of the cryptographic scheme.

Images