A payment security analysis method and system based on an information security environment

Abstract

The application discloses a payment security analysis method and system based on an information security environment, and relates to the technical field of payment security.The method comprises the following steps: S1, acquiring multiple types of data and performing preprocessing; S2, presetting a strategy and constructing an optimized fuzzy division logic; S3, extracting feature parameters and dividing to obtain four types of time sequence sets; S4, analyzing the time sequence data sets and setting differentiated correlation rules; S5, constructing a matrix according to the correlation rules, deducing and generating a hierarchical research and judgment threshold; S6, building an analysis model, and calculating and generating a full-link security research and judgment sequence; and S7, verifying the optimized research and judgment sequence, and generating a payment security comprehensive research and judgment report.The application realizes panoramic and accurate description of external threat intelligence by means of multi-source fusion collection, combined with a fine data management mechanism of preprocessing purification and multi-dimensional security feature fusion extraction, and breaks the defect that the whole domain security environment cannot be comprehensively perceived.

Description

Technical Field

[0001] This invention relates to the field of payment security technology, and more specifically, to a payment security analysis method and system based on an information security environment. Background Technology

[0002] Payment security risk control, as a core and critical link in the field of financial information security, directly determines the efficiency of transaction anomaly identification, network attack prevention capabilities, and fund transaction security level through the accuracy, real-time performance, and comprehensiveness of its risk assessment. It has become a key technical tool for building a solid information security defense line, curbing the risk of financial fraud, and ensuring the compliant and stable operation of the entire payment business chain. Dynamic payment security analysis methods based on multi-source security data fusion, fuzzy feature modeling, multi-factor temporal coupling analysis, and multi-round intelligent algorithm collaborative optimization are not only a core means to solve the problems of one-sided risk assessment, lagging risk identification, and weak adaptability to complex disturbances in traditional static risk control assessment, but also provide solid technical support for comprehensive security situation awareness, multi-level risk quantitative assessment, and dynamic prevention and control of the entire transaction chain. This effectively avoids security risks such as missed or misjudged cases, fraudulent intrusions, and fund losses caused by inaccurate risk assessment, fixed thresholds, and superficial feature analysis.

[0003] However, existing payment security analysis methods and systems based on information security environments mostly adopt a prevention and control model of single transaction data analysis, fixed rule matching, or static threshold comparison in actual use. This has prominent problems such as single risk feature extraction dimensions and incomplete coverage of the overall security situation. At the same time, conventional risk control only relies on simple fuzzy logic to complete basic risk classification, lacks linkage analysis of external threat intelligence and internal system vulnerabilities, and has not established hierarchical and adaptable risk fuzzy subsets and precise quantitative membership relationships. This makes it very easy for the risk boundary to be roughly defined and the feature association representation to be vague.

[0004] No effective solutions have yet been proposed to address the problems in the relevant technologies. Summary of the Invention

[0005] In response to the problems in related technologies, this invention proposes a payment security analysis method and system based on an information security environment to overcome the aforementioned technical problems existing in the existing related technologies.

[0006] To achieve the above objectives, the specific technical solution adopted by the present invention is as follows: According to one aspect of the present invention, a payment security analysis method based on an information security environment includes the following steps: S1. Acquire full-domain information security environment status data, full-link operation data of payment business, risk control cycle configuration parameters and historical security event tracing data. The full-domain information security environment status data includes external threat intelligence data and internal system vulnerability monitoring data. The external threat intelligence data and internal system vulnerability monitoring data are preprocessed. S2. A multi-dimensional security feature fusion extraction strategy is preset, and based on the multi-dimensional security feature fusion extraction strategy, the pre-processed external threat intelligence data and internal system vulnerability monitoring data are introduced into the fuzzy algorithm to construct risk fuzzy subsets and fuzzy membership functions. Then, the fuzzy membership functions are iteratively optimized and calibrated by the particle swarm optimization algorithm to obtain the calibrated fuzzy subset partitioning logic. As a preferred embodiment, the preset multi-dimensional security feature fusion extraction strategy, and based on this strategy, the preprocessed external threat intelligence data and internal system vulnerability monitoring data are introduced into a fuzzy algorithm to construct risk fuzzy subsets and fuzzy membership functions. Then, the fuzzy membership functions are iteratively optimized and calibrated using a particle swarm optimization algorithm to obtain the calibrated fuzzy subset partitioning logic, which includes the following steps: S21. Determine the core integration dimensions as threat level, vulnerability exploitability, business relevance, and risk transmission, and set feature filtering and multi-dimensional integration rules to form a preset multi-dimensional security feature fusion extraction strategy. S22. The preprocessed external threat intelligence data and internal system vulnerability monitoring data are extracted according to the multi-dimensional security feature fusion strategy. Fuzzy algorithms are introduced to construct three-level risk fuzzy subsets of high, medium and low risk, and corresponding fuzzy membership functions are established. As a preferred embodiment, the step of fusing and extracting preprocessed external threat intelligence data and internal system vulnerability monitoring data according to a multi-dimensional security feature fusion strategy, introducing a fuzzy algorithm to construct high, medium, and low-risk fuzzy subsets, and establishing corresponding fuzzy membership functions includes the following steps: S221. Based on the preset multi-dimensional security feature fusion and extraction strategy, feature extraction is performed on the preprocessed external threat intelligence data and internal system vulnerability monitoring data to obtain effective feature data. S222. Combine effective feature data with fuzzy algorithms, set risk classification judgment criteria, and map feature data to corresponding risk intervals through fuzzification processing. At the same time, construct three-level fuzzy subsets of high, medium and low risk respectively. S223. For each risk fuzzy subset of high, medium and low risk levels, extract the core feature parameters within each subset, determine the input variables and output range of the membership function, and then, in combination with the risk impact degree of the core feature parameters, establish fuzzy membership functions that are adapted to each level of risk subset.

[0007] S23. With the goal of minimizing the fitting error of the membership function, the parameters of the fuzzy membership function are used as optimization variables. A particle swarm optimization algorithm model is constructed to perform iterative optimization and calibration. The rationality of the calibrated fuzzy membership function is verified, and the calibrated fuzzy subset partitioning logic is determined.

[0008] S3. Based on the calibrated fuzzy subset partitioning logic, the core risk feature parameters of payment and the auxiliary feature parameters of environmental disturbance in the risk fuzzy subset are extracted. Then, according to the risk control cycle configuration parameters, the core risk feature parameters of payment, the auxiliary feature parameters of environmental disturbance, the full-link operation data of payment business and the source data of historical security events are divided into time series segments to obtain the core risk time series set, the environmental disturbance time series set, the business operation time series set and the historical event time series set as four types of time series datasets. As a preferred embodiment, the process involves extracting core payment risk characteristic parameters and auxiliary environmental disturbance characteristic parameters from the risk fuzzy subset based on the calibrated fuzzy subset partitioning logic. Then, according to the risk control cycle configuration parameters, the core payment risk characteristic parameters, auxiliary environmental disturbance characteristic parameters, full-link payment business operation data, and historical security event tracing data are divided into time-series segments to obtain four types of time-series datasets: core risk time-series dataset, environmental disturbance time-series dataset, business operation time-series dataset, and historical event time-series dataset. This includes the following steps: S31. Based on the calibrated fuzzy subset partitioning logic, extract the core risk feature parameters of payment from the high- and medium-risk fuzzy subsets, and then extract the auxiliary feature parameters of environmental disturbance from the low-risk fuzzy subset. S32. Using the risk control cycle configuration parameters as a unified time benchmark, synchronize and align the core risk characteristic parameters of payment, auxiliary characteristic parameters of environmental disturbances, full-link operation data of payment business, and historical security event tracing data in time. S33. Match the time-series sampling frequency and risk control cycle of payment core risk characteristic parameters, environmental disturbance auxiliary characteristic parameters, payment business full-link operation data and historical security event tracing data to determine the time window for dividing the risk control cycle; S34. Divide the time window according to the risk control cycle, and extract the core risk characteristic parameters of payment, auxiliary characteristic parameters of environmental disturbance, full-link operation data of payment business and historical security event traceability data in a periodic segment after synchronization and alignment. Then, collect them into core risk time series set, environmental disturbance time series set, business operation time series set and historical event time series set according to data type, forming four types of time series datasets.

[0009] S4. Based on the fuzzy comprehensive evaluation algorithm, the multi-factor coupling relationship of the four types of time series datasets is analyzed, the initial fuzzy correlation degree is calculated, and then the coupling correlation weight of the four types of time series datasets is iteratively optimized by the particle swarm optimization algorithm. Based on the optimized coupling correlation weight, the differentiated correlation influence rules are set. As a preferred embodiment, the method of relying on the fuzzy comprehensive evaluation algorithm to analyze the multi-factor coupling relationship of four types of time series datasets, calculating the initial fuzzy correlation degree, iteratively optimizing the coupling correlation weights of the four types of time series datasets through the particle swarm optimization algorithm, and setting differentiated correlation influence rules based on the optimized coupling correlation weights includes the following steps: S41. Using the four types of time series datasets as the evaluation factor set, the risk correlation level is determined as the comment set, and a multi-factor coupling evaluation model is constructed based on the fuzzy comprehensive evaluation algorithm. Then, the multi-factor coupling relationship between the four types of time series datasets is analyzed through fuzzy synthesis operation to obtain the initial fuzzy correlation degree. As a preferred embodiment, the steps of using four types of time-series datasets as evaluation factor sets, determining the risk correlation level as the comment set, constructing a multi-factor coupled evaluation model based on the fuzzy comprehensive evaluation algorithm, and then analyzing the multi-factor coupling relationship between the four types of time-series datasets through fuzzy synthesis operations to obtain the initial fuzzy correlation degree include the following steps: S411. For the core risk time series set, environmental disturbance time series set, business operation time series set and historical event time series set, extract the standardized feature parameters of each time series dataset under the same time series node after synchronization and alignment, form the evaluation factor subset of the corresponding single-class dataset, and group the evaluation factor subset according to the risk impact attribute. Then, take the single-class dataset as the first-level evaluation factor and the corresponding feature parameter as the second-level evaluation factor, and construct the evaluation factor set of fuzzy comprehensive evaluation in a hierarchical manner. S412. Based on the three-level risk classification logic of high, medium and low, the degree of risk correlation is divided into four levels: strong correlation, medium correlation, weak correlation and no correlation, and this is used as the evaluation set for fuzzy comprehensive evaluation. S413. Based on the fuzzy comprehensive evaluation algorithm, establish the fuzzy mapping relationship from the evaluation factor set to the comment set, determine the fuzzy membership degree calculation rules corresponding to each evaluation factor, set up a fuzzy synthesis operator adapted to multi-factor linkage analysis, and construct a multi-factor coupling evaluation model for judging the coupling influence between datasets. S414. Input the synchronous time series feature parameters of the four types of time series datasets into the multi-factor coupled evaluation model, perform fuzzy synthesis operation, obtain the membership results of the comment set corresponding to each evaluation factor, and use this to judge the multi-factor coupled influence relationship between the pairs of the four types of time series datasets and the whole. S415. Based on the multi-factor coupling influence relationship, quantify the degree of correlation between each type of time series dataset and other datasets and payment security risks, and normalize the calculation results to obtain the initial fuzzy correlation degree.

[0010] S42. Taking the maximization of the matching degree between the initial fuzzy correlation degree and the actual payment security risk as the optimization objective, the coupling correlation weights of the four types of time series datasets are used as optimization variables. A particle swarm optimization algorithm model is constructed for iterative optimization to obtain the optimized coupling correlation weights. As a preferred approach, the optimization objective is to maximize the matching degree between the initial fuzzy correlation degree and the actual payment security risk. The optimization involves using the coupling correlation weights of the four types of time-series datasets as optimization variables, constructing a particle swarm optimization algorithm model for iterative optimization, and obtaining the optimized coupling correlation weights. This process includes the following steps: S421. Taking the maximization of the matching degree between the initial fuzzy correlation degree and the actual payment security risk as the core optimization objective, and taking the coupling correlation weights corresponding to the four types of time series datasets as variables to be optimized, and combining the historical security event tracing data to determine the optimization constraints, the optimization framework of the particle swarm optimization algorithm is built. S422. Based on the completed optimization framework, construct a particle swarm optimization algorithm model, initialize the particle population, set the population iteration rules and convergence judgment conditions, and assign the coupling and correlation weights to the particle position vectors to form the initial configuration of the particle swarm optimization algorithm model. S423. Input the initial fuzzy correlation degree into the initialized particle swarm optimization algorithm model, perform iterative optimization operation, and obtain the optimized coupling correlation weight.

[0011] S43. Normalize the optimized coupling correlation weights, and define the priority of the payment security risk impact of the four types of time series datasets according to the weight values. Then, combine the multi-factor coupling relationship characteristics to set differentiated correlation impact rules for the four types of time series datasets.

[0012] S5. Based on the differential correlation influence rule, the correlation of the four types of time series datasets is adjusted, a fuzzy coupling judgment matrix is ​​constructed, and a fuzzy dynamic time series inference is performed in combination with a sliding time window to calculate the benchmark threshold. Then, the benchmark threshold is converged and calibrated by the particle swarm optimization algorithm to obtain the hierarchical security judgment threshold. As a preferred embodiment, the steps of adjusting the correlation of four types of time-series datasets according to the differential correlation influence rule, constructing a fuzzy coupling judgment matrix, performing fuzzy dynamic time-series extrapolation using a sliding time window, calculating the baseline threshold, and then using a particle swarm optimization algorithm to perform baseline threshold convergence calibration to obtain the hierarchical security judgment threshold include the following steps: S51. Based on the differential correlation influence rule, the correlation of the core feature parameters of the four types of time series datasets is adjusted to strengthen the coupling mapping relationship of strongly correlated parameters, weaken the influence weight of weakly correlated parameters, and remove unrelated redundant features to correct the logical consistency of the risk transmission path. S52. Construct a fuzzy coupled judgment matrix, where the row dimension corresponds to the future preset risk control step size, and the column dimension corresponds to the core risk feature parameters and key environmental disturbance parameters of the four types of time series datasets. The feature parameters after correlation adjustment are aligned and filled one by one according to the risk control step size and parameter type to ensure that each element of the matrix corresponds to a unique time series node and feature parameter. S53. Configure the sliding time window parameters, use the fuzzy coupling judgment matrix as the data carrier, combine the fuzzy inference algorithm to perform fuzzy dynamic time series deduction, predict the changing trend of core feature parameters, and calculate to obtain the benchmark threshold for payment security judgment. S54. The optimization objectives are to achieve the highest risk identification accuracy and the lowest false alarm and false alarm rates for the baseline threshold. The baseline threshold is used as the optimization variable to construct a particle swarm optimization algorithm model. Convergence calibration operation is performed to obtain the graded security assessment threshold.

[0013] S6. A dynamic risk matrix quantification algorithm is used to construct a payment security closed-loop analysis model. The hierarchical security assessment threshold is transformed into standardized risk control judgment parameters and input into the payment security closed-loop analysis model. Real-time risk rating parameters are calculated, and a full-link payment security assessment sequence is generated. As a preferred embodiment, the method of constructing a payment security closed-loop analysis model using a dynamic risk matrix quantification algorithm, converting the hierarchical security assessment thresholds into standardized risk control judgment parameters and inputting them into the payment security closed-loop analysis model, calculating real-time risk rating parameters, and generating a full-link payment security assessment sequence includes the following steps: S61. Based on the dynamic risk matrix quantification algorithm, a dynamic risk matrix is ​​constructed with the probability of risk occurrence as the horizontal axis and the degree of risk impact as the vertical axis. High, medium and low risk rating areas are divided, a payment security closed-loop analysis model is constructed, and the judgment logic of each risk rating area is clarified. The hierarchical security judgment threshold is transformed into standardized risk control judgment parameters. S62. Input the standardized risk control judgment parameters into the payment security closed-loop analysis model, and then use the real-time feature parameters of four types of time series datasets as auxiliary inputs. Through the model quantification and accounting module, the real-time risk rating parameters are obtained by combining the probability of risk occurrence and the degree of impact. The real-time risk rating parameters are then integrated according to the risk control time sequence to generate a full-link payment security judgment sequence.

[0014] S7. Conduct real-time verification of the entire payment security assessment sequence, perform dynamic optimization based on fuzzy adaptive correction logic, and generate a comprehensive payment security assessment report under an information security environment.

[0015] According to another aspect of the present invention, a payment security analysis system based on an information security environment is provided, the system comprising: The data acquisition module is used to acquire information security environment status data, payment business full-link operation data, risk control cycle configuration parameters, and historical security event tracing data. The information security environment status data includes external threat intelligence data and internal system vulnerability monitoring data, and preprocesses the external threat intelligence data and internal system vulnerability monitoring data. The fuzzy subset module is used to preset a multi-dimensional security feature fusion extraction strategy. Based on the multi-dimensional security feature fusion extraction strategy, the preprocessed external threat intelligence data and internal system vulnerability monitoring data are introduced into the fuzzy algorithm to construct risk fuzzy subsets and fuzzy membership functions. Then, the fuzzy membership functions are iteratively optimized and calibrated by the particle swarm optimization algorithm to obtain the calibrated fuzzy subset partitioning logic. The time-series data module is used to extract the core risk feature parameters of payment and the auxiliary feature parameters of environmental disturbance from the risk fuzzy subset based on the calibrated fuzzy subset partitioning logic. It then performs time-series segmentation on the core risk feature parameters of payment, the auxiliary feature parameters of environmental disturbance, the full-link operation data of payment business, and the source data of historical security events according to the risk control cycle configuration parameters, resulting in four types of time-series datasets: core risk time-series dataset, environmental disturbance time-series dataset, business operation time-series dataset, and historical event time-series dataset. The influence rules module is used to judge the multi-factor coupling relationship of four types of time series datasets based on the fuzzy comprehensive evaluation algorithm, calculate the initial fuzzy correlation degree, and then iteratively optimize the coupling correlation weight of the four types of time series datasets through the particle swarm optimization algorithm. Based on the optimized coupling correlation weight, differentiated correlation influence rules are set. The threshold assessment module is used to adjust the correlation of four types of time series datasets according to the differential correlation influence rules, construct a fuzzy coupling assessment matrix, and perform fuzzy dynamic time series extrapolation in combination with a sliding time window to calculate the baseline threshold. Then, the baseline threshold is converged and calibrated by the particle swarm optimization algorithm to obtain the hierarchical security assessment threshold. The analysis sequence module is used to construct a payment security closed-loop analysis model using a dynamic risk matrix quantification algorithm. It converts the hierarchical security analysis threshold into standardized risk control judgment parameters and inputs them into the payment security closed-loop analysis model. It calculates real-time risk rating parameters and generates a full-link payment security analysis sequence. The assessment report module is used to conduct real-time verification of the entire payment security assessment sequence, and to perform dynamic optimization based on fuzzy adaptive correction logic to generate a comprehensive assessment report on payment security under the information security environment.

[0016] The beneficial effects of this invention are as follows: 1. This invention achieves a panoramic and accurate depiction of external threat intelligence, internal system vulnerabilities, business operation status, and historical risk patterns through multi-source fusion collection of comprehensive information security environment status data, full-link payment business operation data, risk control configuration parameters, and historical security event tracing data. Combined with a refined data governance mechanism of preprocessing purification and multi-dimensional security feature fusion extraction, it overcomes the shortcomings of payment risk control, such as single data source judgment, static feature extraction, incomplete risk dimension coverage, inability to fully perceive the comprehensive security environment, and fuzzy feature association representation. At the same time, relying on the construction of a three-level risk fuzzy subset and the hierarchical membership function adaptation design, combined with the particle swarm algorithm for the first round of iteration calibration and optimization of membership parameters, it achieves accurate risk feature hierarchy division and quantitative and standardized membership relationship. This replaces the shortcomings of conventional risk control's simple fuzzy modeling and fixed parameter assignment, which result in poor feature adaptability and coarse risk boundary division, thus improving the pertinence and robustness of security feature screening and basic risk modeling.

[0017] 2. Based on the calibrated fuzzy subset partitioning logic, this invention constructs four types of time-series datasets: core risks, environmental disturbances, business operations, and historical events. Combined with time synchronization alignment, sampling frequency matching, and time-series standardization processing through segmented risk control time windows, it achieves unified time-series organization and classification of all risk control data. This solves the problem of distorted risk time-series analysis caused by messy, misaligned, and mixed dynamic and static features in payment risk control time-series data. Simultaneously, relying on a two-layer hierarchical evaluation factor set and a multi-gradient correlation comment set, a fuzzy comprehensive evaluation system is built. This completes in-depth analysis of the coupling relationship between multiple datasets and initial correlation quantification, breaking the limitations of simply judging factor correlations in risk control and ignoring the cross-coupling effects of multi-source data. This enables a scientific and refined analysis of the risk linkage patterns of the four types of time-series data, solidifying the analytical foundation for multi-factor collaborative risk control.

[0018] 3. This invention optimizes the coupling and correlation weights of four types of time-series datasets through a second iteration of the particle swarm optimization algorithm. Combined with weight normalization and risk impact priority definition, it forms differentiated correlation and impact rules adaptable to multiple scenarios. Then, it completes the benchmark threshold calculation by building a fuzzy coupling judgment matrix and performing time-series extrapolation through a sliding time window. Finally, it achieves hierarchical judgment threshold convergence calibration through a third round of particle swarm optimization algorithm, constructing a threshold generation mechanism with multiple rounds of optimization and dynamic extrapolation. This solves the pain points of fixed threshold judgment, single algorithm optimization, weak adaptability of static thresholds in payment risk control, which are prone to missed judgments, misjudgments, and delayed risk identification. At the same time, it breaks away from empirical threshold setting and single standard risk control mode, and achieves accurate classification of high-risk, medium-risk, low-risk, and safe multi-gradient dynamic thresholds, ensuring the real-time adaptability and accuracy of risk judgment thresholds in complex information security environments.

[0019] 4. This invention relies on a dynamic risk matrix quantification algorithm to build a closed-loop analysis model for payment security. It transforms the graded judgment threshold into standardized risk control parameters embedded in the model, and combines probability and impact dual-dimensional calculations to generate a full-link payment security judgment sequence. This achieves seamless transformation from graded thresholds to real-time risk ratings and presents the entire process in a time-series manner. It makes up for the shortcomings of traditional risk control thresholds being disconnected from practical judgment, having a single risk rating dimension, and being unable to form a continuous risk control judgment chain. At the same time, it is equipped with fuzzy adaptive correction logic to complete the real-time verification and dynamic optimization of the judgment sequence. It constructs a complete risk control closed-loop link from data collection, feature modeling, coupling analysis, threshold calibration to rating judgment and dynamic correction, and continuously optimizes the accuracy and adaptability of risk judgment. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a flowchart of a payment security analysis method based on an information security environment according to an embodiment of the present invention; Figure 2 This is a system block diagram of a payment security analysis system based on an information security environment according to an embodiment of the present invention.

[0022] In the picture: 1. Data acquisition module; 2. Fuzzy subset module; 3. Time series data module; 4. Influence rule module; 5. Judgment threshold module; 6. Judgment sequence module; 7. Judgment report module. Detailed Implementation

[0023] The specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples. The following examples are for illustrative purposes only and are not intended to limit the scope of the invention.

[0024] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0025] The present invention will now be further described in conjunction with the accompanying drawings and specific embodiments, such as... Figure 1 As shown, according to one aspect of the present invention, a payment security analysis method based on an information security environment includes the following steps: S1. Acquire full-domain information security environment status data, full-link operation data of payment business, risk control cycle configuration parameters and historical security event tracing data. The full-domain information security environment status data includes external threat intelligence data and internal system vulnerability monitoring data. The external threat intelligence data and internal system vulnerability monitoring data are preprocessed. Specifically, the first step is to conduct standardized collection of multi-source core data across the entire domain, acquiring four types of basic support data: overall information security environment status data, full-link operation data of payment business, risk control cycle configuration parameters, and historical security event tracing data.

[0026] The overall information security environment situation data is divided into external threat intelligence data and internal system vulnerability monitoring data. External threat intelligence includes network attack characteristics, vulnerability warnings, malicious program records, and cross-platform intrusion information, while internal system vulnerability monitoring data integrates system port inspection records, program vulnerability scans, abnormal permission logs, and background operation monitoring information.

[0027] The system synchronously collects payment business operation logs and real-time parameters covering the entire process of transaction initiation, transmission verification, and clearing and settlement. It sets periodic intervals and sampling frequencies to suit different scenarios according to risk control standards, compiles complete files of past abnormal transactions, network attacks, and vulnerability risks, and organizes them into historical security event tracing data. Preprocessing is performed on external threat intelligence data and internal system vulnerability monitoring data, sequentially removing redundant data, completing missing fields, filtering invalid data content, unifying the presentation format of heterogeneous data, identifying and correcting abnormal data entries, standardizing the feature classification standards of the two types of data, aligning multiple data dimensions, and unifying data annotation formats and access points. Simultaneously, the original state of messy, inconsistent, and incomplete data arrangement is streamlined, and the inherent attributes of the original data no longer interfere with subsequent feature fusion, fuzzy modeling, time-series division, and coupling analysis, providing a unified and standardized underlying data foundation for the entire payment security analysis process.

[0028] S2. A multi-dimensional security feature fusion extraction strategy is preset, and based on the multi-dimensional security feature fusion extraction strategy, the pre-processed external threat intelligence data and internal system vulnerability monitoring data are introduced into the fuzzy algorithm to construct risk fuzzy subsets and fuzzy membership functions. Then, the fuzzy membership functions are iteratively optimized and calibrated by the particle swarm optimization algorithm to obtain the calibrated fuzzy subset partitioning logic. In this embodiment of the invention, the preset multi-dimensional security feature fusion extraction strategy, and based on this strategy, the preprocessed external threat intelligence data and internal system vulnerability monitoring data are introduced into a fuzzy algorithm to construct risk fuzzy subsets and fuzzy membership functions. Then, the fuzzy membership functions are iteratively optimized and calibrated using a particle swarm optimization algorithm to obtain the calibrated fuzzy subset partitioning logic, which includes the following steps: S21. Determine the core integration dimensions as threat level, vulnerability exploitability, business relevance, and risk transmission, and set feature filtering and multi-dimensional integration rules to form a preset multi-dimensional security feature fusion extraction strategy. Specifically, based on the inherent collection rules and hierarchical classification logic of risk characteristics in the payment security field, we will carry out the work of defining core integration dimensions, and establish threat level, vulnerability exploitability, business relevance and risk transmission as the basic core dimensions of feature integration, and build a multi-dimensional security feature analysis framework.

[0029] Among them, the threat level is divided into gradients based on the attack categories and risk levels covered by external threat intelligence; the vulnerability exploitability is defined by referring to the vulnerability records and triggering conditions retained by the internal system vulnerability monitoring; the business relevance is matched by the transaction nodes and circulation links covered by the entire payment business operation; and the risk transmission is sorted out by relying on the diffusion path and radiation range recorded by the historical security incident tracing.

[0030] Around the four core fusion dimensions, a unified feature selection standard was simultaneously formulated, aligning with the data source categories and feature item ranges corresponding to each dimension, defining fixed boundaries for feature selection, collecting original feature content that fits the dimension definition, eliminating scattered feature information that does not correspond to the four core dimensions, unifying the collection criteria and classification format of all features to be processed, and establishing multi-dimensional feature fusion arrangement rules based on this, building feature association mapping relationships between the core dimensions, clarifying the combination form, arrangement logic and integration method of cross-dimensional features, aligning the annotation format and archiving specifications of features from different dimensions, forming a fusion execution paradigm of inter-dimensional communication and feature linkage, and integrating the core dimension definition standards, feature selection limitation clauses and cross-dimensional fusion arrangement specifications to solidify a preset multi-dimensional security feature fusion extraction strategy.

[0031] S22. The preprocessed external threat intelligence data and internal system vulnerability monitoring data are extracted according to the multi-dimensional security feature fusion strategy. Fuzzy algorithms are introduced to construct three-level risk fuzzy subsets of high, medium and low risk, and corresponding fuzzy membership functions are established. In this embodiment of the invention, the steps of fusing and extracting preprocessed external threat intelligence data and internal system vulnerability monitoring data according to a multi-dimensional security feature fusion strategy, introducing a fuzzy algorithm to construct high, medium, and low-risk fuzzy subsets, and establishing corresponding fuzzy membership functions include the following steps: S221. Based on the preset multi-dimensional security feature fusion and extraction strategy, feature extraction is performed on the preprocessed external threat intelligence data and internal system vulnerability monitoring data to obtain effective feature data. Specifically, based on the multi-dimensional security feature fusion and extraction strategy established in the early stage, standardized and process-oriented feature extraction work is carried out on the pre-processed external threat intelligence data and internal system vulnerability monitoring data. Throughout the process, the two types of pre-processed data are split layer by layer and identified one by one according to the four core fusion dimensions of threat level, vulnerability exploitability, business relevance and risk transmission, and in accordance with the established feature categories, defined scope and collection standards within the strategy.

[0032] For external threat intelligence data, according to the standard categories corresponding to the threat level, relevant fields such as attack identifiers, risk level labels, and external intrusion records are extracted from the data. For internal system vulnerability monitoring data, referring to the vulnerability exploitability classification items, relevant feature information such as associated vulnerability locations, system vulnerability archives, and abnormal permission labels are collected from the data.

[0033] Simultaneously, based on the matching rules of business relevance, all feature items in the two types of data that correspond to the payment business link nodes are screened. Then, according to the sorting logic of risk transmission, feature content with cross-link relevance and path extension attributes is retained. During the process, scattered fields that exceed the scope defined by the strategy dimensions are screened out, and information content that does not belong to the four core dimensions is stripped away. The record format, labeling form and classification of all extracted features are unified, the arrangement logic and archiving specifications of the two types of data features are aligned, and all feature items that fit the dimension division standards and match the strategy extraction rules are integrated and collected to form effective feature data that is well arranged, clearly categorized and uniformly defined.

[0034] S222. Combine effective feature data with fuzzy algorithms, set risk classification judgment criteria, and map feature data to corresponding risk intervals through fuzzification processing. At the same time, construct three-level fuzzy subsets of high, medium and low risk respectively. Specifically, based on the integrated and aggregated effective feature data, fuzzy algorithms are introduced to advance the risk level labeling and feature classification. First, relying on the feature distribution intervals corresponding to the four core dimensions of threat level, vulnerability exploitability, business relevance, and risk transmission, a unified risk classification judgment standard is formulated for the entire domain. Referring to the inherent data distribution patterns of various features, gradient level boundaries are delineated, the basic division intervals corresponding to high, medium, and low risk levels are clarified, and a fixed reference basis for feature classification is established.

[0035] Based on this, data fuzzification processing is carried out. According to the established risk classification judgment standard, all valid feature data are decomposed and attribute matched one by one. The structured feature parameters are mapped to the corresponding gradient risk intervals according to the dimensional relationship. The fuzzification process follows a unified mapping rule to align the belonging relationship between individual feature parameters and risk intervals, and complete the interval classification and arrangement of all feature data.

[0036] Based on the mapping results of feature data, the data is classified and grouped according to risk level, and three levels of risk fuzzy subsets are constructed in sequence: high, medium, and low. The high-risk fuzzy subset includes strong threat features, highly exploitable vulnerability features, core business related features, and transmission features with extension attributes mapped to the high-risk range. The medium-risk fuzzy subset includes moderate impact features, routine business related features, and ordinary transmission features mapped to the medium-risk range. The low-risk fuzzy subset integrates slight disturbance features, edge related features, and routine features without extension paths mapped to the low-risk range. The feature division boundaries of the three-level subsets are defined throughout the process to maintain the regularity and uniformity of feature dimensions within the subsets, avoid cross-level feature cross-mixing, solidify the feature composition architecture of each subset, and form a hierarchical and orderly three-level risk fuzzy subset system.

[0037] S223. For each risk fuzzy subset of high, medium and low risk levels, extract the core feature parameters within each subset, determine the input variables and output range of the membership function, and then, in combination with the risk impact degree of the core feature parameters, establish fuzzy membership functions that are adapted to each level of risk subset.

[0038] Specifically, based on the established high, medium, and low-risk fuzzy subsets, targeted screening and extraction of core feature parameters within each subset are carried out. Combining four integration dimensions—threat level, vulnerability exploitability, business relevance, and risk transmission—key feature parameters with strong correlation and high impact are selected from the high-risk fuzzy subset, main feature parameters with moderate correlation and conventional impact are collected from the medium-risk fuzzy subset, and basic feature parameters with weak correlation and slight impact are organized from the low-risk fuzzy subset, thus completing the standardized collection of core parameters specific to each subset.

[0039] After the core feature parameters are extracted, the input variables and output range of the fuzzy membership function are uniformly defined. The core feature parameters extracted from each subset are set as fixed input variables of the function. A uniform numerical range is defined as the function output boundary according to the global risk classification standard. The function port specifications of each subset are kept consistent. The differentiated function construction logic is matched with the risk impact gradient corresponding to the core feature parameters of different levels, which conforms to the sensitive distribution law of high-risk features, the smooth transition law of medium-risk features, and the gentle convergence law of low-risk features.

[0040] Following the corresponding construction logic, independent and adaptable fuzzy membership functions are built for the three-level risk fuzzy subsets. The access rules of the input variables and the boundary restrictions of the output range are anchored to complete the standardized mapping configuration from core feature parameters to membership values. This allows each membership function to match the feature composition and risk attributes of the corresponding subset. Through hierarchical customization, a complete membership function architecture with clear levels and accurate adaptation is formed.

[0041] S23. With the goal of minimizing the fitting error of the membership function, the parameters of the fuzzy membership function are used as optimization variables. A particle swarm optimization algorithm model is constructed to perform iterative optimization and calibration. The rationality of the calibrated fuzzy membership function is verified, and the calibrated fuzzy subset partitioning logic is determined.

[0042] Specifically, based on the fuzzy membership functions that have been built at each level, the minimum fitting error of the membership function is established as the core optimization target, and the adjustable configuration parameters inside each fuzzy membership function are uniformly defined as optimization variables, thus completing the standardization and calibration of the algorithm optimization object.

[0043] Based on this, and combined with the data distribution patterns of payment security risk characteristics, a particle swarm optimization algorithm model adapted to fuzzy quantization scenarios is built. The model iterative operation architecture, parameter update rules, and numerical constraint range are set. The basic configuration of the model is completed by matching the hierarchical features of the three-level risk subset. Then, the iterative optimization and calibration process is started. Relying on the particle swarm position inference and dynamic update, the fitting error fluctuation is continuously compared, and the various parameter configurations of the membership function are gradually adjusted to reduce the deviation between the actual feature membership relationship and the function calculation result. The overall fitting error range is continuously narrowed until the algorithm reaches the preset convergence condition, and the parameter iterative optimization process ends.

[0044] After the optimization operation is completed, a systematic rationality verification is carried out on the fuzzy membership function with adjusted parameters. By comparing it with the original feature composition standards of high, medium and low risk fuzzy subsets, the function parameter adaptation, numerical mapping rules and risk classification boundaries are checked. The degree of fit between the feature attribution judgment rules and the previously preset fusion standards is examined. Abnormal situations such as interval misalignment, parameter offset, and judgment deviation are investigated. After the verification process is completed and all indicators meet the established specifications, the division boundaries and feature classification criteria of each level of risk fuzzy subset are re-examined based on the calibrated and improved membership function mapping relationship. The attribution judgment standards of core feature parameters are solidified, the execution framework of the whole domain risk classification is unified, and the fuzzy subset division logic after calibration and finalization is determined, forming a classification basis with regular hierarchical boundaries, unified parameter matching and fixed judgment standards.

[0045] S3. Based on the calibrated fuzzy subset partitioning logic, the core risk feature parameters of payment and the auxiliary feature parameters of environmental disturbance in the risk fuzzy subset are extracted. Then, according to the risk control cycle configuration parameters, the core risk feature parameters of payment, the auxiliary feature parameters of environmental disturbance, the full-link operation data of payment business and the source data of historical security events are divided into time series segments to obtain the core risk time series set, the environmental disturbance time series set, the business operation time series set and the historical event time series set as four types of time series datasets. In this embodiment of the invention, based on the calibrated fuzzy subset partitioning logic, the core payment risk feature parameters and environmental disturbance auxiliary feature parameters are extracted from the risk fuzzy subset. Then, according to the risk control cycle configuration parameters, the core payment risk feature parameters, environmental disturbance auxiliary feature parameters, payment business full-link operation data, and historical security event tracing data are divided into time-series segments to obtain core risk time-series sets, environmental disturbance time-series sets, business operation time-series sets, and historical event time-series sets, which serve as four types of time-series datasets. This includes the following steps: S31. Based on the calibrated fuzzy subset partitioning logic, extract the core risk feature parameters of payment from the high- and medium-risk fuzzy subsets, and then extract the auxiliary feature parameters of environmental disturbance from the low-risk fuzzy subset. Specifically, based on the calibrated and finalized fuzzy subset partitioning logic, following the solidified risk boundary standards and feature attribution judgment criteria, and continuing the classification criteria of the four integrated dimensions of threat level, vulnerability exploitability, business relevance, and risk transmission, a unified and standardized feature parameter extraction process is established.

[0046] For the high-risk fuzzy subset, parameter screening is conducted. Based on the high-level judgment criteria built into the subset, features with strong external threat attributes, high vulnerability exploitability, deep integration with core payment business, and obvious risk transmission paths are extracted. Key indicators that directly affect transaction security, fund flow, and core system operation are collected to form the main part of the core risk parameters. Then, from the medium-risk fuzzy subset, referring to the classification criteria of the medium level, relevant parameters with common external hidden dangers, general system vulnerability characteristics, related to secondary payment business nodes, and with slow diffusion characteristics are screened to supplement risk elements that are likely to have indirect impacts on the entire payment process. The parameters extracted from the high and medium risk subsets are integrated and summarized to form a complete payment core risk feature parameter system.

[0047] Based on this, relying on the definition rules of low-risk fuzzy subsets, relevant content that fits slight external interference, fluctuations in edge system operating conditions, non-core business associations, and no extended transmission characteristics is extracted. These parameters do not involve core security risks of payment transactions, but only reflect external environmental fluctuations, changes in the status of secondary systems, and minor deviations in non-critical businesses. They are uniformly classified as environmental disturbance auxiliary feature parameters. The entire extraction process is strictly executed according to the calibrated subset division logic, clearly defining the distinction boundary between the two types of parameters, avoiding cross-mixing of features at different levels, maintaining clear parameter categories and clear attribution, and completing the classification and collection of core risk parameters and environmental disturbance auxiliary parameters.

[0048] S32. Using the risk control cycle configuration parameters as a unified time benchmark, synchronize and align the core risk characteristic parameters of payment, auxiliary characteristic parameters of environmental disturbances, full-link operation data of payment business, and historical security event tracing data in time. Specifically, the risk control cycle configuration parameters are used as a unified time benchmark across the entire domain. Core benchmark elements such as risk control cycle interval, time sampling frequency, and timestamp labeling specifications are clarified. A unified execution framework for synchronizing and aligning four types of data is established. Systematic time regularization is carried out on payment core risk characteristic parameters, environmental disturbance auxiliary characteristic parameters, payment business full-link operation data, and historical security event tracing data.

[0049] The original time attributes of the four types of data were analyzed, and the collection timestamps and recording time nodes of each type of data were extracted. The differences in time recording formats and sampling periods of different data were clarified and standardized into a time labeling standard consistent with the risk control cycle configuration parameters. This eliminated the problems of heterogeneous time formats and inconsistent labeling standards. Then, using the time nodes set in the risk control cycle as a reference, the time of each type of data was calibrated. The collection time of the core risk characteristic parameters of payment and the auxiliary characteristic parameters of environmental disturbances were accurately matched to the time window of the corresponding risk control cycle. Data with different sampling frequencies were adjusted to a unified frequency to achieve time synchronization between the two types of parameters and the risk control cycle.

[0050] For the entire payment business operation data, according to the time nodes of the entire process from transaction initiation, transmission verification, and clearing and settlement, and in accordance with the risk control cycle benchmark, missing time annotations are supplemented, and time deviations are calibrated to ensure that the time records of business data are completely aligned with the risk control cycle. For historical security incident tracing data, based on information such as the time of occurrence, duration, and handling nodes of the incident, various types of incidents are accurately mapped to the corresponding risk control cycle intervals, clarifying the time correlation between incidents and risk control cycles. Throughout the process, the risk control cycle configuration parameters are used as the sole time reference. Time verification and adjustment are performed on each of the four types of data, removing data items with time misalignment and abnormal annotations, unifying the time dimension of various types of data, and achieving complete synchronization and alignment of the four types of data at the time level.

[0051] S33. Match the time-series sampling frequency and risk control cycle of payment core risk characteristic parameters, environmental disturbance auxiliary characteristic parameters, payment business full-link operation data and historical security event tracing data to determine the time window for dividing the risk control cycle; Specifically, based on the completion of time synchronization and alignment of the four types of data, the system matches the original time-series sampling frequency of payment core risk characteristic parameters, environmental disturbance auxiliary characteristic parameters, payment business full-link operation data and historical security event tracing data with the predetermined risk control cycle, thereby defining a standardized risk control cycle time window.

[0052] Furthermore, a comprehensive review of the inherent collection intervals, recording frequencies, and time-series arrangement rules of the four types of data was conducted. This clarified the high-frequency dynamic sampling characteristics of core risk parameters, the steady-state low-frequency sampling characteristics of environmental disturbance parameters, the time-series recording characteristics of business operation data that change with transaction nodes, and the time-series retention characteristics of historical security event tracing data that is archived according to key nodes. Using the unified risk control cycle configuration parameters across the entire domain as a benchmark, the differences in the sampling intervals and risk control cycle spans of various types of data were compared. For data with a sampling frequency higher than the risk control cycle, continuous time-series records were integrated, redundant sampling nodes were compressed, and the time span of the risk control cycle was adapted.

[0053] For data with a sampling frequency lower than the risk control cycle, adjacent time-series nodes are connected to fill gaps in the time-series arrangement and maintain continuity and integrity in the time dimension. Frequency adaptation calibration is completed by combining the inherent change patterns of the four types of data, taking into account the differentiated characteristics of dynamic risk fluctuations, gradual environmental changes, real-time business processes, and fixed-point retention of historical events, forming a unified time-series frequency matching system across the entire domain. Based on the time-series arrangement structure after frequency matching, the start and end nodes and fixed interval spans of the time windows for dividing the risk control cycle are defined, ensuring that a single time window can fully cover all characteristic parameters, business operation records, and event archive information within the corresponding risk control cycle, avoiding omissions of time-series nodes and misalignment of data intervals.

[0054] S34. Divide the time window according to the risk control cycle, and extract the core risk characteristic parameters of payment, auxiliary characteristic parameters of environmental disturbance, full-link operation data of payment business and historical security event traceability data in a periodic segment after synchronization and alignment. Then, collect them into core risk time series set, environmental disturbance time series set, business operation time series set and historical event time series set according to data type, forming four types of time series datasets.

[0055] Specifically, based on the established risk control cycle, time windows are divided as unified segment boundaries. For the core risk characteristic parameters of payment, auxiliary characteristic parameters of environmental disturbances, full-link operation data of payment business, and historical security event tracing data that have achieved time synchronization and sampling frequency matching in the early stage, a full-domain, cycle-by-cycle segmentation operation is carried out. Then, with the start and end nodes of a single risk control time window as fixed extraction ranges, independent analysis intervals are defined sequentially according to time order. Within each interval, all data content falling into the corresponding time period is accurately extracted to ensure the integrity of the data time sequence and clear boundaries within a single cycle.

[0056] For the core risk characteristic parameters of payment, all characteristic items representing key risk points within the corresponding period are collected window by window. For the auxiliary characteristic parameters of environmental disturbance, the relevant parameters of environmental fluctuations and secondary disturbances recorded within each time window are collected segment by segment. For the full-link operation data of payment business, the time sequence segments generated by transaction flow, node operation and link monitoring are divided according to time windows. For the source tracing data of historical security events, the time window interval is accurately matched and the event archive records and source tracing information retained within the time period are extracted.

[0057] After the segmented data is extracted, targeted collection and integration are carried out according to the preset data classification rules. The core risk parameters extracted from all periods are arranged in chronological order and uniformly summarized to form a core risk time series set. The environmental disturbance-related parameters of each period are organized and archived to form an environmental disturbance time series set. The segmented full-link business runtime sequence fragments are coherently integrated to form a business runtime sequence set. The matched historical event time series records are centrally collected to form a historical event time series set.

[0058] S4. Based on the fuzzy comprehensive evaluation algorithm, the multi-factor coupling relationship of the four types of time series datasets is analyzed, the initial fuzzy correlation degree is calculated, and then the coupling correlation weight of the four types of time series datasets is iteratively optimized by the particle swarm optimization algorithm. Based on the optimized coupling correlation weight, the differentiated correlation influence rules are set. In this embodiment of the invention, the process of using a fuzzy comprehensive evaluation algorithm to assess the multi-factor coupling relationships of four types of time-series datasets, calculating the initial fuzzy correlation degree, iteratively optimizing the coupling correlation weights of the four types of time-series datasets using a particle swarm optimization algorithm, and setting differentiated correlation influence rules based on the optimized coupling correlation weights includes the following steps: S41. Using the four types of time series datasets as the evaluation factor set, the risk correlation level is determined as the comment set, and a multi-factor coupling evaluation model is constructed based on the fuzzy comprehensive evaluation algorithm. Then, the multi-factor coupling relationship between the four types of time series datasets is analyzed through fuzzy synthesis operation to obtain the initial fuzzy correlation degree. In this embodiment of the invention, the steps of using four types of time-series datasets as evaluation factor sets, determining the risk correlation level as a comment set, constructing a multi-factor coupled evaluation model based on a fuzzy comprehensive evaluation algorithm, and then analyzing the multi-factor coupling relationship between the four types of time-series datasets through fuzzy synthesis operations to obtain the initial fuzzy correlation degree include the following steps: S411. For the core risk time series set, environmental disturbance time series set, business operation time series set and historical event time series set, extract the standardized feature parameters of each time series dataset under the same time series node after synchronization and alignment, form the evaluation factor subset of the corresponding single-class dataset, and group the evaluation factor subset according to the risk impact attribute. Then, take the single-class dataset as the first-level evaluation factor and the corresponding feature parameter as the second-level evaluation factor, and construct the evaluation factor set of fuzzy comprehensive evaluation in a hierarchical manner. Specifically, for the core risk time series set, environmental disturbance time series set, business operation time series set, and historical event time series set, based on the time series nodes that are unified across the entire domain and have been synchronized and aligned, we will carry out standardized feature parameter extraction for each dataset one by one, select and retain the regular and unified feature entries under the same time series node, classify and organize the parameters according to the inherent attributes of the single-class dataset, and generate corresponding independent evaluation factor subsets.

[0059] Throughout the process, the temporal nodes of the feature parameters within each subset are kept consistent, and the parameter format and annotation standards are kept consistent. Based on the data composition logic of the original dataset, the parameter composition structure of a single evaluation factor subset is solidified. On this basis, the four types of evaluation factor subsets are classified and grouped according to their risk impact attributes. The four attribute dimensions are distinguished as direct risk effect, indirect environmental disturbance, business link linkage, and historical pattern tracing. The risk effect level and correlation direction of each subset are straightened out, and the regular aggregation of factors with similar impact attributes is achieved.

[0060] Subsequently, a two-tiered, progressive hierarchical structure was constructed. The core risk time series set, environmental disturbance time series set, business operation time series set, and historical event time series set were respectively defined as first-level evaluation factors in the fuzzy comprehensive evaluation system, forming the top-level classification structure of the evaluation framework. Then, the standardized feature parameters extracted from each dataset were uniformly set as second-level evaluation factors under the corresponding first-level factors, serving as the underlying refined indicators for evaluation analysis. The subordinate relationship between the first-level evaluation factors and the second-level evaluation factors was defined, the aggregation standards of previous risk impact attributes were matched, and the correspondence between levels was clarified. This avoided the situation of overlapping and mixing of factors of different levels and categories. Based on the complete hierarchical structure and the regular factor classification, a fuzzy comprehensive evaluation factor set with a clear structure, clear subordinates, and unified dimensions was constructed layer by layer.

[0061] S412. Based on the three-level risk classification logic of high, medium and low, the degree of risk correlation is divided into four levels: strong correlation, medium correlation, weak correlation and no correlation, and this is used as the evaluation set for fuzzy comprehensive evaluation. Specifically, based on the existing risk levels of high, medium, and low, the framework is expanded by combining the linkage characteristics between four types of time-series data: core risks, environmental disturbances, business operations, and historical events. This expands the subdivided judgment dimensions of correlation and improves the data linkage evaluation system under a single risk level. Based on the underlying logic of the three-level risk, the strong correlation level is first defined, benchmarked against the characteristics of the high-risk level, and the correlation forms between time-series data that have direct risk transmission, deep coupling and linkage, and can trigger core payment security fluctuations are identified.

[0062] Secondly, a medium-level correlation is established to align with the attributes of the medium-risk level, standardizing the regular interactions between data to form a linkage relationship with indirect risk impact. Then, a weak-level correlation is established to match the patterns of the low-risk level, defining the loose connection between data with only slight operational interference and no risk transmission path. Finally, an uncorrelated supplementary level is added to clarify the independent state where there is no interaction or risk impact between data. The boundaries of the four correlation levels are uniformly standardized and defined, clarifying the scope of correlation, linkage manifestation, and risk transmission characteristics of each level, ensuring that the gradient between levels is consistent and clearly defined, and forming a close connection with the original three-level risk classification logic. At the same time, the four levels of strong correlation, medium correlation, weak correlation, and no correlation are established as a standard evaluation set for the fuzzy comprehensive evaluation system, forming an adaptive architecture with the previously established hierarchical evaluation factor set, and unifying the classification criteria for the overall correlation analysis.

[0063] S413. Based on the fuzzy comprehensive evaluation algorithm, establish the fuzzy mapping relationship from the evaluation factor set to the comment set, determine the fuzzy membership degree calculation rules corresponding to each evaluation factor, set up a fuzzy synthesis operator adapted to multi-factor linkage analysis, and construct a multi-factor coupling evaluation model for judging the coupling influence between datasets. Specifically, firstly, a complete fuzzy mapping relationship is established from the set of evaluation factors to the set of comments. The four types of time-series datasets corresponding to the primary evaluation factors, as well as the subordinate secondary evaluation factors, are individually bound to the four comment levels of strong correlation, medium correlation, weak correlation, and no correlation. The correlation interval to which each feature factor can belong is clarified, and the mapping boundaries across levels and dimensions are defined to ensure that all evaluation content can complete the standardized level correspondence.

[0064] Based on this, for all secondary evaluation factors, and in combination with the data category and time series characteristics of each factor, specific fuzzy membership degree calculation rules are determined. The membership degree parameter standards that were previously calibrated and finalized are used to distinguish the calculation methods of core risk factors, environmental disturbance factors, business operation factors and historical event factors, unify the quantitative accounting basis of all factors, and maintain the adaptability and consistency of the calculation logic of different categories of factors.

[0065] Simultaneously, considering the practical needs of multi-dimensional linkage analysis of four types of time series data, a dedicated fuzzy synthesis operator adapted to multi-factor coupling operations is set up. This operator balances the independent weight contribution of single factors with the cross-linking influence of multiple factors, aligns with the characteristics of risk transmission and synergistic effects in payment security scenarios, and avoids the judgment bias caused by a single operation mode. Furthermore, by integrating the established fuzzy mapping system, standardized membership degree calculation rules, and customized fuzzy synthesis operators, a multi-factor coupling evaluation model for multi-data interaction analysis is constructed in an integrated manner. The model's data access port, internal operation process, and result output format are clearly defined, and a complete operation link from evaluation factor input to coupling relationship judgment is established, enabling the model to adapt to the synchronous analysis needs of four types of time series datasets.

[0066] S414. Input the synchronous time series feature parameters of the four types of time series datasets into the multi-factor coupled evaluation model, perform fuzzy synthesis operation, obtain the membership results of the comment set corresponding to each evaluation factor, and use this to judge the multi-factor coupled influence relationship between the pairs of the four types of time series datasets and the whole. Specifically, the time-series feature parameters that have been synchronously aligned and calibrated within the four types of time-series datasets—core risks, environmental disturbances, business operations, and historical events—are uniformly imported into the established multi-factor coupled evaluation model. Data matching and access are strictly completed according to the unified time-series node standard to ensure that the input parameters are accurately adapted to the model's internal evaluation factors and fuzzy mapping architecture.

[0067] Then, relying on the model's pre-set exclusive fuzzy synthesis operator and standardized membership degree calculation rules, the full-domain fuzzy synthesis operation process is initiated. Following the progressive logic of step-by-step calculation of secondary evaluation factors and aggregation of primary evaluation factors, the imported full-volume time-series feature parameters are subjected to hierarchical calculation. During the operation stage, according to the established factor-comment mapping relationship, the membership values ​​of each evaluation factor corresponding to strong correlation, medium correlation, weak correlation, and no correlation levels are calculated item by item. Then, the results are integrated and summarized at each level, and finally, the complete membership degree results of each primary evaluation factor matching the entire set of comments are output.

[0068] Simultaneously, relying on the quantitatively derived membership data, multi-level coupling relationship analysis was conducted. First, pairwise cross-analysis of the four types of time-series datasets was completed. The interaction patterns and correlation strengths between core risks and environmental disturbances, core risks and business operations, core risks and historical events, as well as between environmental disturbances and business operations, environmental disturbances and historical events, and business operations and historical events were analyzed. Based on the pairwise analysis, the linkage characteristics of all datasets were integrated to analyze the global linkage network formed by the synergistic superposition of the four types of data. The overall risk coupling trend and mutual influence path under the joint action of multiple factors were defined. The objective membership results output by the model calculation were used as the judgment benchmark throughout the process to clarify the coupling characteristics of individual interactions and overall linkage of various types of time-series data, and to form clear and implementable multi-factor coupling analysis conclusions.

[0069] S415. Based on the multi-factor coupling influence relationship, quantify the degree of correlation between each type of time series dataset and other datasets and payment security risks, and normalize the calculation results to obtain the initial fuzzy correlation degree.

[0070] Specifically, based on the multi-factor coupling influence relationship of pairwise interactions and global collaboration, a unified quantitative accounting scale is established. Combined with the membership degree numerical system generated by strong correlation, medium correlation, weak correlation, and no correlation, a refined measurement of the degree of correlation influence of various time series datasets is carried out.

[0071] For the core risk time series set, environmental disturbance time series set, business operation time series set, and historical event time series set, the bidirectional interaction impact score of a single dataset on the other three datasets is calculated one by one. At the same time, the vertical transmission impact weight of each dataset pointing to the overall payment security risk system is calculated. It takes into account both horizontal data linkage and vertical risk penetration measurement dimensions. The established assignment gradient standard of the fuzzy evaluation process is used throughout the process to transform the qualitatively presented coupling relationship into standardized measurement values, distinguish the score levels of direct risk transmission, indirect linkage interference, and marginal slight impact, and accurately anchor the impact ratio of each type of time series data in the interaction scenario.

[0072] Then, the internal interaction scores and external risk transmission scores of the single-class datasets are integrated and summarized to form a complete and independent original measurement result of the correlation impact. On this basis, a full-domain normalization and regularization process is carried out to unify the numerical range and accounting calibrator of all measurement results, offset the measurement deviation caused by the differences in data volume and feature dimensions of the four types of datasets, calibrate extreme values, balance the overall score ratio, and ensure that all correlation impact values ​​are within a unified reference system. After normalization and calibration, the entire set of measurement results is finally defined as the initial fuzzy correlation degree, which objectively reflects the strength of the interaction between the four types of time series datasets, clearly defines the correlation contribution ratio of each dataset to the overall payment security risk, and forms a quantitative basis that fits the actual coupling law.

[0073] S42. Taking the maximization of the matching degree between the initial fuzzy correlation degree and the actual payment security risk as the optimization objective, the coupling correlation weights of the four types of time series datasets are used as optimization variables. A particle swarm optimization algorithm model is constructed for iterative optimization to obtain the optimized coupling correlation weights. In this embodiment of the invention, the optimization objective is to maximize the matching degree between the initial fuzzy correlation degree and the actual payment security risk. The optimization involves using the coupling correlation weights of four types of time-series datasets as optimization variables, constructing a particle swarm optimization algorithm model for iterative optimization, and obtaining the optimized coupling correlation weights. This process includes the following steps: S421. Taking the maximization of the matching degree between the initial fuzzy correlation degree and the actual payment security risk as the core optimization objective, and taking the coupling correlation weights corresponding to the four types of time series datasets as variables to be optimized, and combining the historical security event tracing data to determine the optimization constraints, the optimization framework of the particle swarm optimization algorithm is built. Specifically, the core optimization objective is to maximize the match between the initial fuzzy correlation degree and the actual payment security risk, ensuring a precise fit between the two. At the same time, by combining historical security event tracing data, existing risk control standards, and actual application scenarios, the constraints of the particle swarm optimization algorithm are determined, an optimization framework adapted to payment security analysis is built, and the core optimization direction is clarified. The match between the initial fuzzy correlation degree and the actual payment security risk is taken as the core consideration to avoid the optimization process deviating from the actual risk control requirements and to ensure that the optimized correlation can truly reflect the payment security status.

[0074] The constraints are set based on historical security incident traceability data. The occurrence patterns, risk transmission paths and correlation characteristics of past payment security incidents are analyzed. Combined with the explicit requirements of existing risk control standards regarding risk level, data format and correlation rules, the value boundaries of the variables to be optimized are defined, and the adjustment range of coupling correlation weights and data access specifications are clarified.

[0075] Simultaneously, in line with the security control requirements of the entire payment business process, the association rules of various data are integrated, and historical security event records and risk control standards are incorporated into the optimization framework to ensure that the constraints not only conform to the actual application scenarios, but also build a complete particle swarm optimization algorithm model by clarifying the optimization goals and refining the constraints.

[0076] S422. Based on the completed optimization framework, construct a particle swarm optimization algorithm model, initialize the particle population, set the population iteration rules and convergence judgment conditions, and assign the coupling and correlation weights to the particle position vectors to form the initial configuration of the particle swarm optimization algorithm model. Specifically, the process begins with initializing the particle population, defining the number of particles, their initial positions, and motion rules. The coupling and correlation weights of various time-series data are then assigned as the position vectors of the particles to ensure that each particle accurately corresponds to the core optimization objective.

[0077] Secondly, the population iteration rules are set to clarify the logic of particle position updates. In combination with the actual needs of payment security analysis, the speed range and update frequency of particle movement are determined to ensure the orderly progress of the iteration process. At the same time, clear convergence judgment conditions are formulated to clarify the core criteria for iteration termination, including the correlation error control range and weight adjustment threshold. When the position vector of the particle population tends to be stable and the correlation error reaches the preset requirements, it is determined that the convergence state has been reached. Finally, the coupling correlation weights are fully assigned to the particle position vectors to complete the initial configuration of the particle swarm optimization algorithm model.

[0078] S423. Input the initial fuzzy correlation degree into the initialized particle swarm optimization algorithm model, perform iterative optimization operation, and obtain the optimized coupling correlation weight.

[0079] Specifically, the initial fuzzy correlation data is fully input into the particle swarm optimization algorithm model that has been initialized and configured, and the iterative optimization operation is started.

[0080] Based on the initial fuzzy correlation degree as input, and combined with the core logic of the particle swarm optimization algorithm, the coupling correlation weight is used as the particle position vector. A reasonable iteration step size and update frequency are set, and the weight parameters are gradually adjusted. During the iteration process, the initial fuzzy correlation degree is used as a benchmark to continuously compare the deviation between the current coupling weight and the ideal weight. Through particle position updates and velocity adjustments, the weight values ​​are continuously corrected, and the deviation range is gradually reduced.

[0081] Furthermore, each iteration records the weight adjustment data and simultaneously optimizes the particle motion trajectory to ensure that the weight adjustment aligns with the actual needs of payment security analysis and avoids parameter deviation. During the iteration process, the weight change trend is monitored in real time. When the weight adjustment reaches a preset standard and the deviation is within a reasonable range, the iteration calculation stops, ultimately yielding the optimized coupling and correlation weights.

[0082] S43. Normalize the optimized coupling correlation weights, and define the priority of the payment security risk impact of the four types of time series datasets according to the weight values. Then, combine the multi-factor coupling relationship characteristics to set differentiated correlation impact rules for the four types of time series datasets.

[0083] Specifically, the optimized coupling and correlation weights are normalized. The core of this process is to eliminate the differences between weights of different dimensions and magnitudes, making all types of weights comparable. First, all optimized coupling and correlation weight data are collected, and the sum of all weights is calculated. Then, each weight is divided by the sum of weights using a formula to obtain a standardized weight percentage. All weight values ​​are uniformly mapped to the range of 0-1, ensuring that weights of different types and magnitudes are on the same evaluation dimension and avoiding judgment bias caused by differences in weight magnitude.

[0084] After normalization, the payment security risk impact priority of the four types of time series datasets (core risk, environment, business, and history) is defined according to the normalized weight values. The time series dataset with the highest weight value corresponds to the core risk time series dataset and has the highest priority. The time series datasets with the second highest weight value correspond to the business and historical time series datasets and have the middle priority. The time series dataset with the lowest weight value corresponds to the environment time series dataset and has the lowest priority.

[0085] Based on this, and combining the coupling relationship characteristics of the four types of time series datasets with the multi-factor coupling correlation rules, differentiated correlation influence rules are set for each type of dataset. The core risk dataset focuses on strengthening risk transmission, the business dataset focuses on process linkage, the historical dataset focuses on pattern reference, and the environmental dataset focuses on auxiliary support. By clarifying the correlation influence scope and function of each type of dataset, targeted correlation rules are formulated to achieve differentiated management of the four types of time series datasets.

[0086] S5. Based on the differential correlation influence rule, the correlation of the four types of time series datasets is adjusted, a fuzzy coupling judgment matrix is ​​constructed, and a fuzzy dynamic time series inference is performed in combination with a sliding time window to calculate the benchmark threshold. Then, the benchmark threshold is converged and calibrated by the particle swarm optimization algorithm to obtain the hierarchical security judgment threshold. In this embodiment of the invention, the steps of adjusting the correlation of four types of time-series datasets according to the differential correlation influence rule, constructing a fuzzy coupling judgment matrix, performing fuzzy dynamic time-series extrapolation using a sliding time window, calculating the baseline threshold, and then using a particle swarm optimization algorithm to perform baseline threshold convergence calibration to obtain the hierarchical security judgment threshold include the following steps: S51. Based on the differential correlation influence rule, the correlation of the core feature parameters of the four types of time series datasets is adjusted to strengthen the coupling mapping relationship of strongly correlated parameters, weaken the influence weight of weakly correlated parameters, and remove unrelated redundant features to correct the logical consistency of the risk transmission path. Specifically, based on the differentiated correlation and impact rules determined in the early stage, correlation adjustment work was carried out for the core feature parameters of four types of time series datasets (core risks, business operations, historical traceability, and environmental disturbances) to ensure the accuracy and logical coherence of coupling mapping. The core feature parameters of each type of dataset were sorted out, and parameters with strong correlation, medium correlation, weak correlation and no correlation were distinguished, and the correlation weight and impact range of each type of parameter were clarified.

[0087] Based on the differentiated association influence rules, the core feature parameters of the four types of time series datasets are adjusted in a targeted manner to strengthen the coupling mapping of strongly correlated parameters, increase their weight in the overall judgment, deepen the association with other datasets, further consolidate the coupling mapping relationship of strongly correlated parameters, reduce the weight of weakly correlated parameters to reduce their impact on the overall judgment and weaken their transmission effect, conduct a comprehensive investigation of unrelated redundant features and invalid parameters, resolutely eliminate unrelated redundant information and invalid features, and avoid interfering with the judgment logic.

[0088] Simultaneously, by combining the correlation patterns of various datasets, logical loopholes in the risk transmission path are corrected, the connections between different datasets are streamlined, the coupling mapping of strongly correlated parameters is strengthened, the influence weight of weakly correlated parameters is weakened, and irrelevant redundant features and invalid information are completely eliminated. Through parameter adjustment, weight optimization, and redundancy elimination, logical deviations in the risk transmission path are corrected, ensuring that the correlation mapping of various parameters conforms to the differentiation rules, strengthening the coupling binding of strongly correlated parameters, weakening the influence of irrelevant parameters, and ultimately achieving a reasonable correlation of core feature parameters.

[0089] S52. Construct a fuzzy coupled judgment matrix, where the row dimension corresponds to the future preset risk control step size, and the column dimension corresponds to the core risk feature parameters and key environmental disturbance parameters of the four types of time series datasets. The feature parameters after correlation adjustment are aligned and filled one by one according to the risk control step size and parameter type to ensure that each element of the matrix corresponds to a unique time series node and feature parameter. Specifically, when constructing the fuzzy coupled judgment matrix, the row dimension should be the future preset risk control step size, and the column dimension should be the core feature parameters of the four types of time series datasets. A judgment matrix with corresponding row and column should be built. The row dimension should be defined as the future preset risk control step size. Each risk control time window should be used as a row of the matrix to ensure that the row dimension corresponds completely with the risk control step size and covers all preset risk control cycles. The column dimension corresponds to the four types of time series datasets: core risk, environmental disturbance, business operation, and historical data. The core feature parameters of each dataset and the key parameters of environmental disturbance should be extracted as column indicators of the matrix.

[0090] Subsequently, the core feature parameters, environmental disturbance parameters, business operation parameters, and historical data parameters of the four types of time series datasets are respectively used as columns of the matrix to ensure that the parameters in each column can accurately match the risk control step size of the corresponding row. When filling the matrix, the feature parameters of each time series node are filled row by row according to the risk control step size to ensure that the parameters in each row correspond to a unique risk control period and the parameters in each column correspond to a unique time series node, thereby achieving precise alignment between rows and columns.

[0091] At the same time, in strict accordance with the correspondence between risk control step size and parameter type, the feature parameters of different datasets are classified and filled into the corresponding rows and columns to ensure that each element in the matrix corresponds to a unique risk control cycle and time sequence node, avoiding parameter misalignment and cycle confusion. Through this matching of rows (risk control step size) and columns (feature parameters), a fuzzy coupled judgment matrix is ​​constructed.

[0092] S53. Configure the sliding time window parameters, use the fuzzy coupling judgment matrix as the data carrier, combine the fuzzy inference algorithm to perform fuzzy dynamic time series deduction, predict the changing trend of core feature parameters, and calculate to obtain the benchmark threshold for payment security judgment. Specifically, when configuring sliding time window parameters, it is necessary to combine the previously established risk control cycle, data time series characteristics, and payment security assessment requirements to clarify the window size, sliding step size, and data collection range. This ensures that the window can fully cover each risk control cycle and accurately match the collection frequency and update rhythm of various time series data to avoid data omissions or duplicate collections. This provides a unified time framework to support time series extrapolation. The completed fuzzy coupled judgment matrix serves as the core data carrier. This matrix integrates various time series characteristic parameters, correlations, and weight configurations, covering all dimensions of data such as core risks, environmental disturbances, and business operations, providing complete data support for fuzzy dynamic time series extrapolation.

[0093] Based on fuzzy inference algorithms and using the temporal feature parameters in the matrix as a foundation, combined with pre-defined membership calculation rules, association mapping relationships, and composition operators, dynamic temporal inference calculations are performed. By analyzing the changing patterns of feature parameters and the strength of association relationships at each temporal node, the future trends of core feature parameters are accurately predicted, including the fluctuation range of risk parameters, the stable state of environmental parameters, and the dynamic changes of business data, ensuring that the inference results closely match the actual needs of payment security scenarios.

[0094] Based on the time-series simulation, and combined with the payment security risk level standards and the influence weights of characteristic parameters, the various types of data obtained from the simulation are quantitatively calculated to clarify the threshold range of characteristic parameters corresponding to different risk levels. Finally, the benchmark threshold for payment security assessment is determined, and this benchmark threshold needs to cover the assessment needs of different risk levels and different data types to ensure that subsequent risk identification and parameter optimization have clear judgment standards.

[0095] S54. The optimization objectives are to achieve the highest risk identification accuracy and the lowest false alarm and false alarm rates for the baseline threshold. The baseline threshold is used as the optimization variable to construct a particle swarm optimization algorithm model. Convergence calibration operation is performed to obtain the graded security assessment threshold.

[0096] Specifically, with improving the accuracy of risk control assessment as the core guiding principle, the dual core optimization objectives are to maximize the accuracy of risk identification of the benchmark threshold and simultaneously minimize the false alarm rate and false alarm rate. This establishes a solid evaluation benchmark for threshold optimization. On this basis, the payment security assessment benchmark threshold obtained from previous simulations is set as the core optimization variable. A dedicated computational model is built based on the particle swarm optimization algorithm adapted to risk quantification scenarios. At the same time, a large number of historical security event samples and actual transaction risk control data are imported as the basis for model verification and iteration.

[0097] Based on the actual rules of payment business risk control, a reasonable fluctuation range and numerical constraint boundary for the benchmark threshold are defined to avoid the optimized threshold deviating from business reality and exceeding the normal judgment range. Adopting a unified and standardized particle initialization configuration, speed and position iteration update logic, and stringent population convergence judgment conditions, the algorithm completes the full-dimensional parameter settings for the model. Subsequently, it initiates a global convergence calibration iteration operation. The algorithm continuously fine-tunes the benchmark threshold parameters by dynamically adjusting particle positions. Each iteration is benchmarked against real risk samples to verify the identification effectiveness, calculates the core indicators of missed and false alarms in real time, continuously selects the best threshold combinations, continuously reduces judgment bias, and optimizes risk identification efficiency.

[0098] Furthermore, the convergence status of the objective function is monitored in real time throughout the process. When the risk identification accuracy reaches its peak, the false alarm rate and false alarm rate are stably controlled within the preset critical range, and the threshold parameters corresponding to the particle population tend to be stable without significant fluctuations, the iteration is terminated and convergence calibration is completed. Finally, based on the optimal threshold determined by calibration, combined with the original three-level risk classification architecture of high, medium and low, the numerical range and gradient boundary corresponding to different risk levels are refined and decomposed, and the hierarchical judgment criteria are divided to form a hierarchical safety assessment threshold.

[0099] S6. A dynamic risk matrix quantification algorithm is used to construct a payment security closed-loop analysis model. The hierarchical security assessment threshold is transformed into standardized risk control judgment parameters and input into the payment security closed-loop analysis model. Real-time risk rating parameters are calculated, and a full-link payment security assessment sequence is generated. In this embodiment of the invention, the step of constructing a payment security closed-loop analysis model using a dynamic risk matrix quantification algorithm, converting the hierarchical security assessment thresholds into standardized risk control judgment parameters and inputting them into the payment security closed-loop analysis model, calculating real-time risk rating parameters, and generating a full-link payment security assessment sequence includes the following steps: S61. Based on the dynamic risk matrix quantification algorithm, a dynamic risk matrix is ​​constructed with the probability of risk occurrence as the horizontal axis and the degree of risk impact as the vertical axis. High, medium and low risk rating areas are divided, a payment security closed-loop analysis model is constructed, and the judgment logic of each risk rating area is clarified. The hierarchical security judgment threshold is transformed into standardized risk control judgment parameters. Specifically, relying on the dynamic risk matrix quantification algorithm, the probability of risk occurrence is strictly set as the horizontal axis and the degree of risk impact as the vertical axis. A two-dimensional dynamic risk matrix infrastructure adapted to payment security scenarios is built. Combined with the previously calibrated graded security assessment thresholds, and referring to the probability gradient distribution and the hierarchical rules of harm impact, three independent risk rating areas of high, medium and low are accurately divided. The numerical boundaries and ranges corresponding to the horizontal and vertical axes of each area are clearly defined, which fits the actual characteristics of the evolution of payment transaction risks, event triggering and harm spread.

[0100] Building upon this foundation, the system integrates core processes such as full-process time-series data analysis, multi-factor coupled judgment, fuzzy dynamic deduction, and threshold optimization calibration. It also integrates the functions of data collection, feature screening, correlation analysis, and threshold determination to build a closed-loop payment security analysis model. This model enables closed-loop control of risks from source identification and dynamic judgment to risk classification. Furthermore, it refines the exclusive judgment logic for the three risk rating regions, clarifying that high-risk regions correspond to high probability of occurrence, strong financial harm, and impact on core businesses; medium-risk regions correspond to medium trigger probability and impact on routine businesses; and low-risk regions are suitable for low-probability fluctuations and only minor environmental disturbances. The system clarifies the detailed rules for regional division and risk classification criteria, eliminating issues of ambiguous boundaries and overlapping classifications.

[0101] Finally, based on the coordinate rules and regional judgment standards of the dynamic risk matrix, the optimized hierarchical security judgment thresholds are standardized and decomposed to adapt to the matrix rating logic, and converted into fixed judgment parameters that can be directly applied to real-time risk control. The parameter matching rules and early warning trigger conditions are solidified, so that the dynamic matrix rating, closed-loop model judgment and risk control parameter judgment are deeply linked, providing visualized matrix support and standardized and implementable risk control judgment basis for real-time monitoring of payment security, hierarchical risk early warning and precise policy implementation.

[0102] S62. Input the standardized risk control judgment parameters into the payment security closed-loop analysis model, and then use the real-time feature parameters of four types of time series datasets as auxiliary inputs. Through the model quantification and accounting module, the real-time risk rating parameters are obtained by combining the probability of risk occurrence and the degree of impact. The real-time risk rating parameters are then integrated according to the risk control time sequence to generate a full-link payment security judgment sequence.

[0103] Specifically, the standardized risk control judgment parameters that have been finalized and implemented will be fully imported into the completed payment security closed-loop analysis model as the core benchmark for real-time judgment throughout the entire process, fixing the underlying judgment logic and threshold reference system of the model. On this basis, real-time dynamic feature parameters of four types of time-series datasets, namely core risks, environmental disturbances, business operations, and historical events, will be simultaneously accessed. With real-time updated multi-dimensional data as auxiliary input, the deep integration and adaptation of fixed risk control standards and dynamic business data will be achieved.

[0104] Relying on the professional quantitative accounting module built into the model, and combining the two core evaluation dimensions of risk occurrence probability and risk impact degree in the dynamic risk matrix, the model performs superimposed calculations. It compares the grading threshold range defined by standardized risk control judgment parameters with the fluctuation patterns of real-time feature parameters and the coupling correlation of multiple factors, and performs refined quantitative accounting for each time-series node. The accounting process takes into account the dynamic changes in risk trigger probability and the actual level of harm to capital security and business links after the risk occurs. It accurately calculates and generates real-time risk rating parameters corresponding to each risk control time window, ensuring that the risk rating of a single node is objective and consistent with the current trading operation and data monitoring situation.

[0105] Subsequently, following the unified risk control time sorting rules across the entire domain, real-time risk rating parameters scattered across different periods and time nodes are systematically linked, organized, and integrated. Invalid data nodes with misaligned time sequences or abnormal values ​​are eliminated, and the progressive relationship of the rating data throughout the entire process is straightened out. This forms a continuous data chain covering the entire process of transaction initiation, data monitoring, fund transfer, and risk tracing, generating a complete and standardized end-to-end payment security assessment sequence. This clearly presents the entire process of risk dynamic evolution and fluctuation over time, providing the platform with a core assessment basis that is time-coherent, quantitatively accurate, and fully traceable for conducting real-time risk monitoring, tiered early warning push, differentiated risk control measures, and dynamic strategy optimization.

[0106] S7. Conduct real-time verification of the entire payment security assessment sequence, perform dynamic optimization based on fuzzy adaptive correction logic, and generate a comprehensive payment security assessment report under an information security environment.

[0107] Specifically, a full-dimensional real-time verification process is conducted on the generated end-to-end payment security assessment sequence. This process involves connecting with real-time transaction monitoring data, on-site risk control feedback information, and archived historical security benchmark data. The focus is on verifying the temporal continuity of the assessment sequence, the accuracy of feature parameter matching, and the logical rationality of risk rating. This process accurately identifies issues such as numerical drift, node anomalies, temporal discontinuities, and rating misalignments in the sequence, pinpointing key points that require optimization and correction.

[0108] Based on this, dynamic optimization is carried out by relying on the preset fuzzy adaptive correction logic. Combined with the membership rules established in the early stage, the optimized coupling association weights and hierarchical judgment thresholds, the feature parameter association mapping relationship, risk rating calculation weights and dynamic judgment standards are finely adjusted in real time for the deviation content found in the verification, so as to adapt to the data anomalies brought about by complex information security scenarios such as network fluctuations, transaction peaks and environmental interference.

[0109] The correction process takes into account both immediate real-time risk anomaly signals and long-term historical risk evolution patterns, balances short-term data deviation calibration with normalized security baseline control, and optimizes unreasonable ratings and lagging data within the judgment sequence node by node to ensure that the entire judgment timeline aligns with the current real risk control situation. After completing adaptive dynamic optimization, the system integrates and calibrates the complete payment security judgment sequence, collects the characteristic fluctuation patterns, multi-factor coupling and linkage changes, and dynamic evolution trends of risk levels from four types of time-series datasets, clearly marks high-risk warning points, medium-risk key monitoring intervals, and low-risk normalized security ranges, sorts out the auxiliary impact of environmental disturbances and historical event reference comparison information, and finally summarizes the real-time security situation, risk causes, transmission paths, hierarchical judgment conclusions, and targeted risk control optimization suggestions according to the standardized analysis architecture, comprehensively sorts out the details of security across the entire domain, and finally generates a comprehensive payment security judgment report that is adapted to the overall information security environment, with detailed data and clear conclusions.

[0110] According to another aspect of the invention, such as Figure 2 As shown, a payment security analysis system based on an information security environment is provided. The system includes: Data acquisition module 1 is used to acquire information security environment status data, payment business full-link operation data, risk control cycle configuration parameters and historical security event tracing data. The information security environment status data includes external threat intelligence data and internal system vulnerability monitoring data, and preprocesses the external threat intelligence data and internal system vulnerability monitoring data. Fuzzy subset module 2 is used to preset a multi-dimensional security feature fusion extraction strategy, and based on the multi-dimensional security feature fusion extraction strategy, it introduces the preprocessed external threat intelligence data and internal system vulnerability monitoring data into the fuzzy algorithm to construct risk fuzzy subsets and fuzzy membership functions. Then, the particle swarm optimization algorithm is used to iteratively optimize and calibrate the fuzzy membership functions to obtain the calibrated fuzzy subset partitioning logic. The time series data module 3 is used to extract the core risk feature parameters of payment and the auxiliary feature parameters of environmental disturbance from the risk fuzzy subset based on the calibrated fuzzy subset partitioning logic, and to perform time series segmentation on the core risk feature parameters of payment, the auxiliary feature parameters of environmental disturbance, the full-link operation data of payment business and the source data of historical security events according to the risk control cycle configuration parameters, so as to obtain the core risk time series set, the environmental disturbance time series set, the business operation time series set and the historical event time series set as four types of time series datasets; The influence rule module 4 is used to judge the multi-factor coupling relationship of four types of time series datasets based on the fuzzy comprehensive evaluation algorithm, calculate the initial fuzzy correlation degree, and then iteratively optimize the coupling correlation weight of the four types of time series datasets through the particle swarm optimization algorithm. Based on the optimized coupling correlation weight, differentiated correlation influence rules are set. The threshold judgment module 5 is used to adjust the correlation of four types of time series datasets according to the differential correlation influence rules, construct a fuzzy coupling judgment matrix, and perform fuzzy dynamic time series inference in combination with a sliding time window to calculate the benchmark threshold. Then, the benchmark threshold is converged and calibrated by the particle swarm optimization algorithm to obtain the hierarchical security judgment threshold. The analysis sequence module 6 is used to construct a payment security closed-loop analysis model using a dynamic risk matrix quantification algorithm. It converts the hierarchical security analysis threshold into standardized risk control judgment parameters and inputs them into the payment security closed-loop analysis model. It calculates real-time risk rating parameters and generates a full-link payment security analysis sequence. The assessment report module 7 is used to conduct real-time verification of the entire payment security assessment sequence, and to perform dynamic optimization based on fuzzy adaptive correction logic to generate a comprehensive assessment report on payment security under the information security environment.

[0111] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A payment security analysis method based on an information security environment, characterized in that, Includes the following steps: S1. Acquire full-domain information security environment status data, full-link operation data of payment business, risk control cycle configuration parameters and historical security event tracing data. The full-domain information security environment status data includes external threat intelligence data and internal system vulnerability monitoring data. The external threat intelligence data and internal system vulnerability monitoring data are preprocessed. S2. A multi-dimensional security feature fusion extraction strategy is preset, and based on the multi-dimensional security feature fusion extraction strategy, the pre-processed external threat intelligence data and internal system vulnerability monitoring data are introduced into the fuzzy algorithm to construct risk fuzzy subsets and fuzzy membership functions. Then, the fuzzy membership functions are iteratively optimized and calibrated by the particle swarm optimization algorithm to obtain the calibrated fuzzy subset partitioning logic. S3. Based on the calibrated fuzzy subset partitioning logic, the core risk feature parameters of payment and the auxiliary feature parameters of environmental disturbance in the risk fuzzy subset are extracted. Then, according to the risk control cycle configuration parameters, the core risk feature parameters of payment, the auxiliary feature parameters of environmental disturbance, the full-link operation data of payment business and the source data of historical security events are divided into time series segments to obtain the core risk time series set, the environmental disturbance time series set, the business operation time series set and the historical event time series set as four types of time series datasets. S4. Based on the fuzzy comprehensive evaluation algorithm, the multi-factor coupling relationship of the four types of time series datasets is analyzed, the initial fuzzy correlation degree is calculated, and then the coupling correlation weight of the four types of time series datasets is iteratively optimized by the particle swarm optimization algorithm. Based on the optimized coupling correlation weight, the differentiated correlation influence rules are set. S5. Based on the differential correlation influence rule, the correlation of the four types of time series datasets is adjusted, a fuzzy coupling judgment matrix is ​​constructed, and a fuzzy dynamic time series inference is performed in combination with a sliding time window to calculate the benchmark threshold. Then, the benchmark threshold is converged and calibrated by the particle swarm optimization algorithm to obtain the hierarchical security judgment threshold. S6. A dynamic risk matrix quantification algorithm is used to construct a payment security closed-loop analysis model. The hierarchical security assessment threshold is transformed into standardized risk control judgment parameters and input into the payment security closed-loop analysis model. Real-time risk rating parameters are calculated, and a full-link payment security assessment sequence is generated. S7. Conduct real-time verification of the entire payment security assessment sequence, perform dynamic optimization based on fuzzy adaptive correction logic, and generate a comprehensive payment security assessment report under an information security environment.

2. The payment security analysis method based on an information security environment according to claim 1, characterized in that, The preset multi-dimensional security feature fusion extraction strategy, and based on this strategy, introduces preprocessed external threat intelligence data and internal system vulnerability monitoring data into a fuzzy algorithm to construct risk fuzzy subsets and fuzzy membership functions. Then, the fuzzy membership functions are iteratively optimized and calibrated using a particle swarm optimization algorithm to obtain the calibrated fuzzy subset partitioning logic, which includes the following steps: S21. Determine the core integration dimensions as threat level, vulnerability exploitability, business relevance, and risk transmission, and set feature filtering and multi-dimensional integration rules to form a preset multi-dimensional security feature fusion extraction strategy. S22. The preprocessed external threat intelligence data and internal system vulnerability monitoring data are extracted according to the multi-dimensional security feature fusion strategy. Fuzzy algorithms are introduced to construct three-level risk fuzzy subsets of high, medium and low risk, and corresponding fuzzy membership functions are established. S23. With the goal of minimizing the fitting error of the membership function, the parameters of the fuzzy membership function are used as optimization variables. A particle swarm optimization algorithm model is constructed to perform iterative optimization and calibration. The rationality of the calibrated fuzzy membership function is verified, and the calibrated fuzzy subset partitioning logic is determined.

3. The payment security analysis method based on an information security environment according to claim 1, characterized in that, The process, based on the calibrated fuzzy subset partitioning logic, extracts the core payment risk feature parameters and environmental disturbance auxiliary feature parameters from the risk fuzzy subset. Then, according to the risk control cycle configuration parameters, it performs time-series segmentation on the core payment risk feature parameters, environmental disturbance auxiliary feature parameters, full-link payment business operation data, and historical security event tracing data, resulting in four types of time-series datasets: core risk time-series dataset, environmental disturbance time-series dataset, business operation time-series dataset, and historical event time-series dataset. The process includes the following steps: S31. Based on the calibrated fuzzy subset partitioning logic, extract the core risk feature parameters of payment from the high- and medium-risk fuzzy subsets, and then extract the auxiliary feature parameters of environmental disturbance from the low-risk fuzzy subset. S32. Using the risk control cycle configuration parameters as a unified time benchmark, synchronize and align the core risk characteristic parameters of payment, auxiliary characteristic parameters of environmental disturbances, full-link operation data of payment business, and historical security event tracing data in time. S33. Match the time-series sampling frequency and risk control cycle of payment core risk characteristic parameters, environmental disturbance auxiliary characteristic parameters, payment business full-link operation data and historical security event tracing data to determine the time window for dividing the risk control cycle; S34. Divide the time window according to the risk control cycle, and extract the core risk characteristic parameters of payment, auxiliary characteristic parameters of environmental disturbance, full-link operation data of payment business and historical security event traceability data in a periodic segment after synchronization and alignment. Then, collect them into core risk time series set, environmental disturbance time series set, business operation time series set and historical event time series set according to data type, forming four types of time series datasets.

4. The payment security analysis method based on an information security environment according to claim 1, characterized in that, The process of using a fuzzy comprehensive evaluation algorithm to assess the multi-factor coupling relationships of four types of time-series datasets, calculating the initial fuzzy correlation degree, iteratively optimizing the coupling correlation weights of the four types of time-series datasets using a particle swarm optimization algorithm, and setting differentiated correlation influence rules based on the optimized coupling correlation weights includes the following steps: S41. Using the four types of time series datasets as the evaluation factor set, the risk correlation level is determined as the comment set, and a multi-factor coupling evaluation model is constructed based on the fuzzy comprehensive evaluation algorithm. Then, the multi-factor coupling relationship between the four types of time series datasets is analyzed through fuzzy synthesis operation to obtain the initial fuzzy correlation degree. S42. Taking the maximization of the matching degree between the initial fuzzy correlation degree and the actual payment security risk as the optimization objective, the coupling correlation weights of the four types of time series datasets are used as optimization variables. A particle swarm optimization algorithm model is constructed for iterative optimization to obtain the optimized coupling correlation weights. S43. Normalize the optimized coupling correlation weights, and define the priority of the payment security risk impact of the four types of time series datasets according to the weight values. Then, combine the multi-factor coupling relationship characteristics to set differentiated correlation impact rules for the four types of time series datasets.

5. The payment security analysis method based on an information security environment according to claim 1, characterized in that, The process of adjusting the correlation of four types of time-series datasets according to the differential correlation influence rule, constructing a fuzzy coupling judgment matrix, and performing fuzzy dynamic time-series extrapolation using a sliding time window to calculate the baseline threshold, followed by calibration of the baseline threshold convergence using a particle swarm optimization algorithm to obtain the hierarchical security judgment threshold, includes the following steps: S51. Based on the differential correlation influence rule, the correlation of the core feature parameters of the four types of time series datasets is adjusted to strengthen the coupling mapping relationship of strongly correlated parameters, weaken the influence weight of weakly correlated parameters, and remove unrelated redundant features to correct the logical consistency of the risk transmission path. S52. Construct a fuzzy coupled judgment matrix, where the row dimension corresponds to the future preset risk control step size, and the column dimension corresponds to the core risk feature parameters and key environmental disturbance parameters of the four types of time series datasets. The feature parameters after correlation adjustment are aligned and filled one by one according to the risk control step size and parameter type to ensure that each element of the matrix corresponds to a unique time series node and feature parameter. S53. Configure the sliding time window parameters, use the fuzzy coupling judgment matrix as the data carrier, combine the fuzzy inference algorithm to perform fuzzy dynamic time series deduction, predict the changing trend of core feature parameters, and calculate to obtain the benchmark threshold for payment security judgment. S54. The optimization objectives are to achieve the highest risk identification accuracy and the lowest false alarm and false alarm rates for the baseline threshold. The baseline threshold is used as the optimization variable to construct a particle swarm optimization algorithm model. Convergence calibration operation is performed to obtain the graded security assessment threshold.

6. The payment security analysis method based on an information security environment according to claim 1, characterized in that, The process of constructing a payment security closed-loop analysis model using a dynamic risk matrix quantification algorithm, converting hierarchical security assessment thresholds into standardized risk control judgment parameters and inputting them into the payment security closed-loop analysis model, calculating real-time risk rating parameters, and generating a full-link payment security assessment sequence includes the following steps: S61. Based on the dynamic risk matrix quantification algorithm, a dynamic risk matrix is ​​constructed with the probability of risk occurrence as the horizontal axis and the degree of risk impact as the vertical axis. High, medium and low risk rating areas are divided, a payment security closed-loop analysis model is constructed, and the judgment logic of each risk rating area is clarified. The hierarchical security judgment threshold is transformed into standardized risk control judgment parameters. S62. Input the standardized risk control judgment parameters into the payment security closed-loop analysis model, and then use the real-time feature parameters of four types of time series datasets as auxiliary inputs. Through the model quantification and accounting module, the real-time risk rating parameters are obtained by combining the probability of risk occurrence and the degree of impact. The real-time risk rating parameters are then integrated according to the risk control time sequence to generate a full-link payment security judgment sequence.

7. The payment security analysis method based on an information security environment according to claim 2, characterized in that, The process of fusing and extracting preprocessed external threat intelligence data and internal system vulnerability monitoring data according to a multi-dimensional security feature fusion strategy, introducing a fuzzy algorithm to construct high, medium, and low-risk fuzzy subsets, and establishing corresponding fuzzy membership functions includes the following steps: S221. Based on the preset multi-dimensional security feature fusion and extraction strategy, feature extraction is performed on the preprocessed external threat intelligence data and internal system vulnerability monitoring data to obtain effective feature data. S222. Combine effective feature data with fuzzy algorithms, set risk classification judgment criteria, and map feature data to corresponding risk intervals through fuzzification processing. At the same time, construct three-level fuzzy subsets of high, medium and low risk respectively. S223. For each risk fuzzy subset of high, medium and low risk levels, extract the core feature parameters within each subset, determine the input variables and output range of the membership function, and then, in combination with the risk impact degree of the core feature parameters, establish fuzzy membership functions that are adapted to each level of risk subset.

8. The payment security analysis method based on an information security environment according to claim 4, characterized in that, The process of using four types of time-series datasets as evaluation factor sets, determining the risk correlation level as the comment set, constructing a multi-factor coupled evaluation model based on the fuzzy comprehensive evaluation algorithm, and then analyzing the multi-factor coupling relationship among the four types of time-series datasets through fuzzy synthesis operations to obtain the initial fuzzy correlation degree includes the following steps: S411. For the core risk time series set, environmental disturbance time series set, business operation time series set and historical event time series set, extract the standardized feature parameters of each time series dataset under the same time series node after synchronization and alignment, form the evaluation factor subset of the corresponding single-class dataset, and group the evaluation factor subset according to the risk impact attribute. Then, with the single-class dataset as the first-level evaluation factor and the corresponding feature parameter as the second-level evaluation factor, construct the evaluation factor set of fuzzy comprehensive evaluation in a hierarchical manner. S412. Based on the three-level risk classification logic of high, medium and low, the degree of risk correlation is divided into four levels: strong correlation, medium correlation, weak correlation and no correlation, and this is used as the evaluation set for fuzzy comprehensive evaluation. S413. Based on the fuzzy comprehensive evaluation algorithm, establish the fuzzy mapping relationship from the evaluation factor set to the comment set, determine the fuzzy membership degree calculation rules corresponding to each evaluation factor, set up a fuzzy synthesis operator adapted to multi-factor linkage analysis, and construct a multi-factor coupling evaluation model for judging the coupling influence between datasets. S414. Input the synchronous time series feature parameters of the four types of time series datasets into the multi-factor coupled evaluation model, perform fuzzy synthesis operation, obtain the membership results of the comment set corresponding to each evaluation factor, and use this to judge the multi-factor coupled influence relationship between the pairs of the four types of time series datasets and the whole. S415. Based on the multi-factor coupling influence relationship, quantify the degree of correlation between each type of time series dataset and other datasets and payment security risks, and normalize the calculation results to obtain the initial fuzzy correlation degree.

9. A payment security analysis method based on an information security environment according to claim 4, characterized in that, The optimization objective is to maximize the matching degree between the initial fuzzy correlation degree and the actual payment security risk. The coupling correlation weights of the four types of time-series datasets are used as optimization variables. A particle swarm optimization algorithm model is constructed for iterative optimization to obtain the optimized coupling correlation weights. This process includes the following steps: S421. Taking the maximization of the matching degree between the initial fuzzy correlation degree and the actual payment security risk as the core optimization objective, and taking the coupling correlation weights corresponding to the four types of time series datasets as variables to be optimized, and combining the historical security event tracing data to determine the optimization constraints, the optimization framework of the particle swarm optimization algorithm is built. S422. Based on the completed optimization framework, construct a particle swarm optimization algorithm model, initialize the particle population, set the population iteration rules and convergence judgment conditions, and assign the coupling and correlation weights to the particle position vectors to form the initial configuration of the particle swarm optimization algorithm model. S423. Input the initial fuzzy correlation degree into the initialized particle swarm optimization algorithm model, perform iterative optimization operation, and obtain the optimized coupling correlation weight.

10. A payment security analysis system based on an information security environment, used to implement the payment security analysis method based on an information security environment as described in any one of claims 1-9, characterized in that, The system includes: The data acquisition module is used to acquire information security environment status data, payment business full-link operation data, risk control cycle configuration parameters, and historical security event tracing data. The information security environment status data includes external threat intelligence data and internal system vulnerability monitoring data, and preprocesses the external threat intelligence data and internal system vulnerability monitoring data. The fuzzy subset module is used to preset a multi-dimensional security feature fusion extraction strategy. Based on the multi-dimensional security feature fusion extraction strategy, the preprocessed external threat intelligence data and internal system vulnerability monitoring data are introduced into the fuzzy algorithm to construct risk fuzzy subsets and fuzzy membership functions. Then, the fuzzy membership functions are iteratively optimized and calibrated by the particle swarm optimization algorithm to obtain the calibrated fuzzy subset partitioning logic. The time-series data module is used to extract the core risk feature parameters of payment and the auxiliary feature parameters of environmental disturbance from the risk fuzzy subset based on the calibrated fuzzy subset partitioning logic. It then performs time-series segmentation on the core risk feature parameters of payment, the auxiliary feature parameters of environmental disturbance, the full-link operation data of payment business, and the source data of historical security events according to the risk control cycle configuration parameters, resulting in four types of time-series datasets: core risk time-series dataset, environmental disturbance time-series dataset, business operation time-series dataset, and historical event time-series dataset. The influence rules module is used to judge the multi-factor coupling relationship of four types of time series datasets based on the fuzzy comprehensive evaluation algorithm, calculate the initial fuzzy correlation degree, and then iteratively optimize the coupling correlation weight of the four types of time series datasets through the particle swarm optimization algorithm. Based on the optimized coupling correlation weight, differentiated correlation influence rules are set. The threshold assessment module is used to adjust the correlation of four types of time series datasets according to the differential correlation influence rules, construct a fuzzy coupling assessment matrix, and perform fuzzy dynamic time series extrapolation in combination with a sliding time window to calculate the baseline threshold. Then, the baseline threshold is converged and calibrated by the particle swarm optimization algorithm to obtain the hierarchical security assessment threshold. The analysis sequence module is used to construct a payment security closed-loop analysis model using a dynamic risk matrix quantification algorithm. It converts the hierarchical security analysis threshold into standardized risk control judgment parameters and inputs them into the payment security closed-loop analysis model. It calculates real-time risk rating parameters and generates a full-link payment security analysis sequence. The assessment report module is used to conduct real-time verification of the entire payment security assessment sequence, and to perform dynamic optimization based on fuzzy adaptive correction logic to generate a comprehensive assessment report on payment security under the information security environment.

Images