Payment security system based on AI artificial intelligence and use method

Through the payment security system based on AI artificial intelligence, using transaction behavior analysis and support vector machine classifier to identify abnormal payment patterns, a multi-layer monitoring network is built, which solves the shortcomings of the existing technology in identifying and preventing new fraud methods, and achieves the following steps Efficient payment security and responsiveness.

CN118967134AInactive Publication Date: 2024-11-15ZHONGRONG PAYMENT CO LTD

Patent Information

Application Number
CN202410998235.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-24
Publication Date
2024-11-15
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

When existing payment security technologies process large-scale real-time data, it is difficult to effectively identify and prevent new and complex fraud methods. Traditional methods rely on fixed rules and patterns, lack flexibility and accuracy, and cannot update defense mechanisms in a timely manner, resulting in the system being unable to Effectively identify and prevent well-designed fraud.

Method used

Adopt payment security system based on AI artificial intelligence, identify abnormal payment patterns through transaction behavior analysis module, configure support vector machine classifiers for model training module, build a multi-layer monitoring network with multi-model integration module, and generate real-time alarm signals for real-time monitoring module to verify strength The adjustment module adjusts verification measures according to the alert urgency, and the risk feedback module optimizes the monitoring model to match new fraudulent means.

Benefits of technology

It significantly improves payment security and response speed, can identify and prevent potential fraud in real time, and forms powerful and comprehensive security protection through multi-layer monitoring networks, enhancing the overall security and trust of the payment ecosystem.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118967134A_ABST
    Figure CN118967134A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of payment security, in particular to a payment security system based on AI artificial intelligence and a use method, and the system comprises a transaction behavior analysis module, a model training module, a multi-model integration module, a real-time monitoring module, a verification intensity adjustment module and a risk feedback module. According to the invention, by applying the artificial intelligence technology, deeper analysis and abnormal mode identification are carried out on transaction data, the payment security is significantly improved, and by automatically identifying the deviation between the abnormal payment frequency and the payment mode, the potential fraudulent behavior can be reflected in real time and prevented in advance; a support vector machine is used for mode training and risk assessment, the effectiveness of a model is optimized through continuous loop testing, the system is allowed to process and analyze data layer by layer progressively, a more powerful and comprehensive payment security protection network is formed, the response time is remarkably shortened, and the overall security and credibility of the payment ecosystem are enhanced.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of payment security technology, and in particular to a payment security system based on AI artificial intelligence and a method of using the same. Background Art

[0002] The field of payment security technology is an important technical field that focuses on ensuring the security of funds and the confidentiality of personal information during transactions. It includes various technologies and methods to prevent fraud, identity theft, phishing attacks and other forms of cybercrime. Payment security technology enhances transaction security through encryption algorithms, two-factor authentication, risk assessment models and behavior monitoring systems. It also covers monitoring and response technologies to detect and respond to suspicious activities in real time to protect consumers and merchants from financial losses.

[0003] Among them, the payment security system based on AI involves the use of artificial intelligence technologies, such as machine learning and deep learning, to enhance the security of the payment system. AI can automatically detect and prevent fraud, and identify abnormal patterns and suspicious transactions by analyzing batches of transaction data. AI can also optimize the authentication process and improve the accuracy of user identity authentication through intelligent behavior analysis. Its main purpose is to reduce the incidence of fraud, shorten response time, and enhance the security and trust of the entire payment ecosystem.

[0004] Although existing payment security technologies include encryption algorithms, two-factor authentication and other security measures, they face challenges in processing large-scale real-time data, especially in identifying and responding to new and complex fraud methods. Traditional methods rely on fixed rules and patterns, which limit the ability to adapt to new changes. Therefore, when faced with changing attack strategies, they cannot update their defense mechanisms in a timely manner. Traditional systems lack sufficient flexibility and accuracy in the user verification process, rely too much on historical data and static patterns, and fail to make full use of behavioral analysis to improve the accuracy of identity verification. These deficiencies result in the system being unable to effectively identify and prevent carefully designed fraudulent behaviors in actual operations, exposing consumers and merchants to the risk of financial losses. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a payment security system based on AI artificial intelligence and a method of use.

[0006] In order to achieve the above purpose, the present invention adopts the following technical solution: a payment security system based on AI artificial intelligence includes:

[0007] The transaction behavior analysis module analyzes user behavior and payment data, uses AI to identify abnormalities in payment frequency and payment patterns, evaluates deviations from normal patterns, and outputs behavioral abnormality indicators;

[0008] The model training module adopts the abnormal behavior indicators, configures the support vector machine classifier parameters, performs abnormal payment pattern training, checks the validity of the model through cyclic testing, and generates a payment monitoring model;

[0009] The multi-model integration module uses AI artificial intelligence according to the payment monitoring model, sets the hierarchical judgment logic, adjusts the input and output relationship between the payment monitoring models, and forms a multi-layer monitoring network;

[0010] The real-time monitoring module uses the multi-layer monitoring network for ongoing transactions, utilizes AI artificial intelligence, marks and notifies abnormal transactions, synchronizes and updates monitoring logs, and obtains real-time alarm signals;

[0011] The verification strength adjustment module adjusts the strength of the verification measures according to the urgency of the real-time alarm signal, quantifies the risk of payment by reviewing the identities of both parties to the payment, and outputs the adjusted verification requirements;

[0012] Based on the adjusted verification requirements and payment feedback, the risk feedback module optimizes the logic of the payment monitoring model, matches new transaction patterns and fraud methods, and generates a payment risk management strategy.

[0013] As a further solution of the present invention, the abnormal behavior indicators include abnormal payment frequency, the degree of deviation of payment mode and abnormal transaction time; the payment monitoring model includes model parameter configuration, training data set and test verification effect; the multi-layer monitoring network includes multi-level monitoring models, input-output relationships of monitoring models and settings of hierarchical judgment logic; the real-time alarm signal includes marking, notification and update status of transaction monitoring logs for abnormal transactions; the adjusted verification requirements include review requirements for the identities of both parties to the payment, risk assessment indicators of payment and the strength of applied verification measures; the payment risk management strategy includes optimized payment monitoring model logic, matching new transaction patterns and identifying new fraud methods.

[0014] As a further solution of the present invention, the transaction behavior analysis module includes:

[0015] The payment data analysis submodule analyzes user behavior and payment data, extracts user payment frequency, amount and time data, sorts and identifies patterns of user payment habits according to payment time series, and obtains user payment behavior characteristics;

[0016] The correlation analysis submodule uses the user payment behavior characteristics and AI artificial intelligence to analyze the correlation between payment methods and locations, compare user behavior with normal payment behavior, calculate the deviation between the payment mode and the normal mode, and obtain the payment behavior analysis results;

[0017] The abnormality identification submodule sets an identification threshold for abnormal payment behavior according to the payment behavior analysis result, performs real-time scanning on the payment data, identifies transaction behaviors that exceed the normal deviation range, and outputs abnormal behavior indicators.

[0018] As a further solution of the present invention, the model training module includes:

[0019] The abnormality indicator configuration submodule sets the initial parameters of the support vector machine classifier according to the abnormal behavior indicator, including error tolerance and kernel function type, to obtain a classifier benchmark configuration;

[0020] The parameter adjustment submodule adopts the classifier benchmark configuration and the gradient descent algorithm to optimize the parameters of the support vector machine, and obtains the optimized classifier parameters by optimizing the classification boundaries and matching the differentiated payment behavior data;

[0021] The payment test submodule utilizes the optimized classifier parameters to perform cyclic testing on the reserved payment test data, analyzes the efficiency and reaction time of payment identification, and constructs a payment monitoring model.

[0022] As a further solution of the present invention, the formula of the gradient descent algorithm is as follows:

[0023]

[0024] Among them, θ is the model parameter, α is the learning rate, is the gradient of the cost function with respect to the parameter, λ is the regularization coefficient, is the gradient of the regularization term, β and φ are the amplitude and phase adjustment coefficients respectively, γ is the attenuation coefficient, σ is the attenuation rate, and t is the number of iterations.

[0025] As a further solution of the present invention, the multi-model integration module includes:

[0026] The logic setting submodule uses the payment monitoring model to define the hierarchical positions and decision logics of multiple payment monitoring models, including priority sorting and decision threshold values, to obtain hierarchical decision logic;

[0027] The model relationship adjustment submodule uses AI artificial intelligence according to the hierarchical decision logic to configure the data flow and processing priority between payment monitoring models, check the flow of payment data and process it according to the predetermined logic to obtain the adjusted model relationship;

[0028] The network structure integration submodule adopts the adjusted model relationship, integrates the multi-layer payment monitoring model, analyzes the transmission path of payment data between models, optimizes payment security, and outputs a multi-layer monitoring network.

[0029] As a further solution of the present invention, the real-time monitoring module includes:

[0030] The transaction capture submodule uses AI artificial intelligence based on the multi-layer monitoring network to receive ongoing payment data in real time, analyze the characteristics of each payment, capture payments that deviate from the normal pattern, and obtain real-time payment records;

[0031] The abnormal classification submodule uses a preset abnormal identification logic according to the real-time payment record to mark the abnormal payment, check the response and processing of the abnormal payment, and obtain the marked abnormal payment;

[0032] The log update submodule updates the payment monitoring log according to the marked abnormal payment, records the abnormal payment detection results and payment responses, tracks and records the abnormal payment events, and outputs a real-time alarm signal.

[0033] As a further solution of the present invention, the verification strength adjustment module includes:

[0034] The alarm classification submodule classifies the real-time alarm signals according to their urgency, analyzes the priority and processing requirements of each alarm, sets associated response measures, matches payment security requirements of differentiated levels, and obtains alarm level assessment results;

[0035] The identity verification submodule verifies the identities of both parties to the payment based on the alarm level assessment result, and obtains the identity verification result by checking the identity documents and payment history, evaluating the credit and security information associated with the real-time payment;

[0036] The risk quantification submodule uses the identity verification results to perform quantitative analysis on the potential risks of the payment, evaluate the potential risks and risk levels of the payment, adjust the strength of the verification measures according to the risk level, and output the adjusted verification requirements.

[0037] As a further solution of the present invention, the risk feedback module includes:

[0038] The verification requirement analysis submodule evaluates the impact of the adjusted verification requirements on the payment monitoring model, identifies the logic parts that need to be adjusted, and obtains the model parameters to be optimized for the new payment fraud situation;

[0039] The model optimization submodule readjusts the parameters and logic of the payment monitoring model according to the model parameters to be optimized, updates the records according to the real-time payment data and fraud detection technology, optimizes the adaptability of the payment monitoring model, and obtains the optimized payment monitoring model;

[0040] The strategy building submodule uses the optimized payment monitoring model to design a management plan for potential payment risks, optimizes payment security and user trust by identifying and avoiding fraudulent behavior, and outputs a payment risk management strategy.

[0041] A payment security method based on AI artificial intelligence, the payment security method based on AI artificial intelligence is executed based on the above-mentioned payment security system based on AI artificial intelligence, and comprises the following steps:

[0042] S1: Based on the user’s payment history and payment data, analyze the payment frequency and payment pattern, identify deviations from the normal pattern, and generate abnormal payment behavior characteristics;

[0043] S2: According to the abnormal payment behavior characteristics, configure the parameters of the support vector machine classifier, train the abnormal payment pattern, verify the efficiency of the model through loop testing, and obtain the payment monitoring model;

[0044] S3: adopt the payment monitoring model, set up multi-level decision logic, adjust the input and output relationship between the payment monitoring models, form a multi-layer monitoring network, and establish a payment security monitoring network;

[0045] S4: Monitor ongoing payments in real time through the payment security monitoring network, mark abnormal payments and update monitoring logs, and generate real-time abnormal transaction alerts;

[0046] S5: Using the real-time abnormal transaction alert, adjust the strength of the verification measures, and output the payment risk management strategy by quantifying the transaction risk and reviewing the identities of the two payment parties, referring to new transaction patterns and fraud methods.

[0047] Compared with the prior art, the advantages and positive effects of the present invention are:

[0048] In the present invention, by applying artificial intelligence technology, especially machine learning and deep learning, it is possible to conduct a more in-depth analysis of transaction data and identify abnormal patterns, significantly improving payment security. By automatically identifying abnormal payment frequencies and deviations in payment patterns, the system can respond in real time and prevent potential fraud in advance. The use of support vector machines for pattern training and risk assessment not only enhances the accuracy of the model, but also optimizes the effectiveness of the model through continuous cycle testing. The construction of a multi-layer monitoring network allows the system to process and analyze data layer by layer, and the output of each layer provides more accurate input for the next layer, forming a more powerful and comprehensive security protection network. The innovative means are combined to significantly shorten the response time and enhance the overall security and trust of the payment ecosystem. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 is a system flow chart of the present invention;

[0050] Figure 2 It is a schematic diagram of the system framework of the present invention;

[0051] Figure 3 This is a flow chart of the transaction behavior analysis module of the present invention;

[0052] Figure 4 This is a flow chart of the model training module of the present invention;

[0053] Figure 5 It is a flow chart of the multi-model integration module of the present invention;

[0054] Figure 6 It is a flow chart of the real-time monitoring module of the present invention;

[0055] Figure 7 This is a flow chart of the verification strength adjustment module of the present invention;

[0056] Figure 8 is a flow chart of the risk feedback module of the present invention;

[0057] Fig. 9 It is a schematic diagram of the steps of the method for using the present invention. DETAILED DESCRIPTION

[0058] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0059] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, "multiple" means two or more, unless otherwise clearly and specifically defined.

[0060] Embodiment 1

[0061] See also Figure 1 to Figure 2 The present invention provides a technical solution, a payment security system based on AI artificial intelligence, comprising:

[0062] The transaction behavior analysis module analyzes user behavior and payment data, uses AI to identify anomalies in payment frequency and payment patterns, evaluates deviations from normal patterns, marks them as abnormal activities, and outputs behavior anomaly indicators;

[0063] The model training module uses behavioral anomaly indicators, configures support vector machine classifier parameters, conducts abnormal payment pattern training, verifies the effectiveness of the model through cyclic testing, and generates a payment monitoring model;

[0064] The multi-model integration module uses AI artificial intelligence to set hierarchical judgment logic based on the payment monitoring model, adjust the input and output relationship between payment monitoring models, and form a multi-layer monitoring network by real-time monitoring of payment time and location;

[0065] The real-time monitoring module uses a multi-layer monitoring network for ongoing transactions, uses AI artificial intelligence to implement real-time monitoring, marks and notifies abnormal transactions, synchronizes and updates monitoring logs, and obtains real-time alarm signals;

[0066] The verification intensity adjustment module adjusts the intensity of verification measures according to the urgency of real-time alarm signals, and outputs the adjusted verification requirements by reviewing the identities of both parties to the payment and quantifying the risks and risk levels of the payment;

[0067] Based on the adjusted verification requirements, the risk feedback module collects and analyzes verification results and payment feedback, adjusts the payment monitoring model, optimizes the monitoring algorithm and logic, matches new transaction patterns and fraud methods, and generates payment risk management strategies.

[0068] Abnormal behavior indicators include abnormal payment frequency, degree of deviation of payment patterns and abnormal transaction time. The payment monitoring model includes the model's parameter configuration, training data set and test verification results. The multi-layer monitoring network includes multi-level monitoring models, input-output relationships of monitoring models and settings of hierarchical judgment logic. Real-time alarm signals include marking, notification and update status of transaction monitoring logs for abnormal transactions. The adjusted verification requirements include requirements for review of the identities of both payment parties, payment risk assessment indicators and the strength of applied verification measures. The payment risk management strategy includes optimized payment monitoring model logic, matching of new transaction patterns and identification of new fraud methods.

[0069] See also Figure 2 , 3 , the transaction behavior analysis module includes:

[0070] The payment data analysis submodule analyzes user behavior and payment data, extracts user payment frequency, amount and time data, sorts and identifies patterns of user payment habits according to payment time series, and obtains the execution process of user payment behavior characteristics as follows;

[0071] The payment data analysis submodule collects the user's payment frequency, amount and time data by analyzing user behavior and payment data. Through statistical analysis methods, such as time series analysis, the user's payment habits are sorted and pattern recognition is performed. This can be achieved through clustering algorithms, such as K-means, which cluster users according to payment behavior characteristics and identify typical payment patterns. The system can use prediction models, such as the autoregressive moving average (ARMA) model, to predict the user's future payment behavior and obtain the user's payment behavior characteristics. The formula used is:

[0072] P(t)=C+φ 1 P(t-1)+…+φ p P(tp)+θ 1 ∈(t-1)+…+θ q ∈(tq)+∈(t)

[0073] Where P(t) is the payment amount at time t, c is a constant term, and φ 1 ,…,φ p and θ 1 ,…,θ q are the parameters of autoregression and moving average respectively, and ∈(t) is the error term.

[0075] The correlation analysis submodule uses the user payment behavior characteristics and AI to analyze the correlation between payment methods and locations, compare user behavior with normal payment behavior, calculate the deviation between payment mode and normal mode, and obtain the payment behavior analysis results. The execution process is as follows;

[0076] The correlation analysis submodule uses the characteristics of user payment behavior and uses artificial intelligence algorithms, such as decision trees or neural networks, to analyze the correlation between payment methods and locations. By comparing user behavior with the defined normal payment behavior model, the system calculates the deviation between the payment pattern and the normal pattern. This can be achieved through distance metrics, such as Euclidean distance or Manhattan distance, to obtain the payment behavior analysis results. The formula used is:

[0077]

[0078] Where D is the distance between two payment patterns, x i and i are the values ​​of user payment mode and regular payment mode in the i-th dimension respectively, and n is the number of dimensions.

[0079] The abnormal identification submodule sets the recognition threshold of abnormal payment behavior according to the payment behavior analysis results, performs real-time scanning on payment data, identifies transaction behaviors that exceed the normal deviation range, and outputs the execution process of abnormal behavior indicators as follows;

[0080] The abnormality identification submodule sets the identification threshold of abnormal payment behavior according to the payment behavior analysis results.

[0081] Use statistical analysis methods, such as standard deviation analysis or boxplot analysis, to determine whether a payment behavior constitutes abnormal payment behavior. Scan payment data in real time, identify transactions that exceed a preset threshold, and output abnormal behavior indicators.

[0082] See also Figure 2 , 4 , the model training module includes:

[0083] The abnormal indicator configuration submodule sets the initial parameters of the support vector machine classifier according to the abnormal behavior indicators, including error tolerance and kernel function type, and obtains the execution process of the classifier benchmark configuration as follows;

[0084] The abnormal indicator configuration submodule defines and classifies abnormal behavior indicators based on abnormal behavior indicators. Specific indicators include abnormal transaction frequency, sudden changes in amount, etc. Based on the indicators, the system sets the initial parameters of the support vector machine (SVM) classifier. For error tolerance, the initial value is set to 0.01. Considering the specificity of the data set, the Gaussian kernel function is selected as the kernel type, and the initial bandwidth parameter of the kernel function is set to 1.0. After the parameters are preliminarily set, a small amount of sample data is used for pre-classification to evaluate the preliminary classification effect. According to the results of the pre-classification, the error tolerance and the bandwidth of the kernel function are adjusted to ensure that the classifier can better adapt to the preliminary data, provide a basic configuration for the next step of detailed parameter tuning, and obtain the classifier benchmark configuration.

[0085] The parameter adjustment submodule adopts the classifier benchmark configuration and the gradient descent algorithm to optimize the parameters of the support vector machine. By optimizing the classification boundaries and matching the differentiated payment behavior data, the execution process of the optimized classifier parameters is as follows;

[0086] The parameter adjustment submodule uses the classifier benchmark configuration to further optimize the SVM classifier parameters, adjusts the bandwidth parameters of the kernel function, uses the binary method to gradually debug, finds the optimal bandwidth, dynamically adjusts the error tolerance, and uses an incremental adjustment strategy, increasing by 0.005 each time until the model performance no longer improves. Through the manual adjustment method, repeatedly test and record the classification performance after each adjustment, such as accuracy and recall, to ensure that the best parameter combination is found and the optimized classifier parameters are obtained.

[0087] The formula for the gradient descent algorithm is as follows:

[0088]

[0089] Among them, θ is the model parameter, α is the learning rate, is the gradient of the cost function with respect to the parameter, λ is the regularization coefficient, is the gradient of the regularization term, β and φ are the amplitude and phase adjustment coefficients respectively, γ is the attenuation coefficient, σ is the attenuation rate, and t is the number of iterations.

[0090] The execution flow is as follows:

[0091] Initialize the parameters θ and calculate the gradient of the cost function J(θ) and the gradient of the regularization term R(θ) The two gradients are added together and the amplitude is adjusted by βcos(φ) and the attenuation factor γe -σt To adjust the learning step size, use the comprehensive gradient and learning rate α to update θ, repeat the process until convergence, and the values ​​of β, φ, γ, and σ are optimized and determined on the training set through cross-validation. By introducing the amplitude and attenuation parameters, the flexibility and adaptability of gradient descent are improved, which is particularly suitable for the nonlinear and dynamic characteristics of payment behavior data.

[0092] The payment test submodule uses the optimized classifier parameters to perform cyclic testing on the reserved payment test data, analyzes the efficiency and reaction time of payment recognition, and constructs the execution process of the payment monitoring model as follows;

[0093] The payment test submodule uses the optimized classifier parameters to set the number of test cycles, and selects different data subsets for testing in each cycle. In each test, the accuracy, false alarm rate, and response time of payment behavior recognition are recorded in detail, and cluster analysis is performed on the indicators to evaluate the adaptability and efficiency of the classifier in different payment behavior scenarios. Through cyclic testing, we can fully understand the performance of the classifier in actual applications, further adjust the monitoring model to adapt to the real-world payment environment, and build a payment monitoring model.

[0094] See also Figure 2 , 5 , the multi-model integration module includes:

[0095] The logic setting submodule uses the payment monitoring model to define the hierarchical positions and decision logic of multiple payment monitoring models, including priority sorting and decision threshold values, and obtains the execution process of the hierarchical decision logic as follows;

[0096] The logic setting submodule uses the payment monitoring model to define the hierarchical position of multiple payment monitoring models, including classifying and sorting the models according to the level of detail of the payment behavior analysis and the processing speed. Set the priority sorting so that the recognition model of high-risk payment behavior has a higher priority. Set the decision threshold values ​​of each level model, such as the abnormal transaction amount threshold, the abnormal frequency threshold, etc., to ensure that each level model is processed only after the appropriate threshold is triggered. It involves a comprehensive assessment of the sensitivity and specificity of each model. Through actual data simulation testing, the threshold value is continuously adjusted to ensure the accuracy of the decision and the rationality of the hierarchy, and the hierarchical decision logic is obtained. The formula used is:

[0097]

[0098] Where D is the decision logic of the payment monitoring network, n is the number of models, and p i represents the priority of the i-th model, θ i is the decision threshold of the model.

[0099] The model relationship adjustment submodule uses AI artificial intelligence according to the hierarchical decision logic to configure the data flow and processing priority between payment monitoring models, check the flow of payment data and process it according to the predetermined logic, and obtain the execution process of the adjusted model relationship as follows;

[0100] The model relationship adjustment submodule configures the data flow direction and processing priority between payment monitoring models according to the hierarchical decision logic. Define the data flow direction to ensure that the high-priority model receives the payment data first. Use AI technology to analyze the efficiency and accuracy of payment data flow, and adjust the data flow path and processing priority according to the analysis results. Comprehensively consider the data dependency and processing capabilities between models, optimize the data flow strategy, and continuously optimize the model relationship through simulation and actual data operation to ensure that data can flow efficiently and accurately between models. The adjusted model relationship is obtained using the formula:

[0101]

[0102] Among them, M is the influence of the relationship between models, d i represents the data inflow of the i-th model, α i is the processing priority of the model, and n is the number of models.

[0103] The network structure integration submodule adopts the adjusted model relationship, integrates the multi-layer payment monitoring model, analyzes the transmission path of payment data between models, optimizes payment security, and outputs the execution process of the multi-layer monitoring network as follows;

[0104] The network structure integration submodule uses the adjusted model relationship to integrate the multi-layer payment monitoring model, analyze the transmission path of payment data between models, and ensure that each layer of the model can receive the necessary data for processing. The entire network structure is optimized to improve payment security. This includes reducing the redundancy of the transmission path and enhancing the data processing capacity of the key path. Through integration and optimization, a multi-layer monitoring network is output, using the formula:

[0105]

[0106] Where N is the structural integration effect of the payment monitoring network, f i represents the data processing capacity of the i-th layer model, n is the number of models, and γ i is the relative importance of the model in the entire network.

[0107] See also Figure 2 , 6 , the real-time monitoring module includes:

[0108] The transaction capture submodule is based on a multi-layer monitoring network and uses AI to receive ongoing payment data in real time, analyze the characteristics of each payment, capture payments that deviate from the normal pattern, and obtain real-time payment records. The execution process is as follows;

[0109] The transaction capture submodule is based on a multi-layer monitoring network and uses artificial intelligence technology to receive and process ongoing payment data in real time. This includes receiving payment data streams in real time and conducting in-depth analysis of the characteristics of each payment, such as the amount, transaction time, and transaction location. By setting anomaly detection algorithms, such as setting statistical thresholds, payment behaviors that deviate significantly from conventional payment patterns can be captured in real time. Each dimension of payment behavior is evaluated, compared with historical data models, and data points that deviate from the norm are captured in a timely manner to generate real-time payment records.

[0110] The abnormal classification submodule uses the preset abnormal identification logic based on the real-time payment records to mark the abnormal payment, check the response and processing of the abnormal payment, and obtain the execution process of the marked abnormal payment as follows;

[0111] The abnormal classification submodule uses the preset abnormal identification logic to classify each payment based on the real-time payment records, and applies the classification logic to evaluate the characteristics of the payment records, such as whether it exceeds the average transaction amount, whether the transaction frequency is abnormal, etc. Through logical judgment, the payment that meets the abnormal characteristics is marked and the payment data is marked as abnormal. The response and processing of the marked abnormal payment are further analyzed to implement the next step of the response strategy and obtain the marked abnormal payment.

[0112] The log update submodule updates the payment monitoring log according to the marked abnormal payment, records the abnormal payment detection results and payment responses, tracks and records the abnormal payment events, and outputs the real-time alarm signal. The execution process is as follows;

[0113] The log update submodule updates the payment monitoring log based on the marked abnormal payments, including receiving abnormal payment data and recording the abnormal detection results and payment response status in the monitoring log. By tracking and recording the development of abnormal payment events, relevant departments can take timely measures. Ensure that all relevant payment data is recorded and tracked to facilitate subsequent data analysis and decision support, and output real-time alarm signals.

[0114] See also Figure 2 , 7 , the verification strength adjustment module includes:

[0115] The alarm classification submodule classifies the real-time alarm signals according to their urgency, analyzes the priority and processing requirements of each alarm, sets the associated response measures, matches the payment security requirements of the differentiated levels, and obtains the execution process of the alarm level assessment results as follows;

[0116] The alarm classification submodule is classified according to the urgency of real-time alarm signals. Each alarm signal contains an urgency mark, and the alarm is classified according to the mark. By analyzing the alarm content, the transaction amount involved, the transaction frequency and other factors, the priority and processing requirements of each alarm are determined. According to different alarm levels, corresponding response measures are set, such as immediately freezing transactions, notifying users for verification, etc., to ensure that the payment security requirements of differentiated levels can be matched. Through the setting of classification and response measures, various emergency situations can be effectively managed and responded to, the security of the payment system can be improved, and the alarm level assessment results can be obtained. The formula used is:

[0117]

[0118] Among them, A level is the alarm level assessment result, n is the number of alarms, u i Score the urgency of the i-th alarm, v i is the processing demand weight corresponding to the i-th alarm.

[0119] The identity verification submodule verifies the identities of both parties to the payment based on the alarm level assessment results, and evaluates the credit and security information associated with the real-time payment by checking the identity documents and payment history. The execution process of obtaining the identity verification result is as follows;

[0120] Based on the alarm level assessment results, the identity verification submodule verifies the identities of both parties to the transaction, mainly by checking identity documents such as ID cards and passports, and combining payment history to evaluate credit and security information related to real-time payment. The system can verify the authenticity and credit status of both parties to the transaction and ensure the security of payment. It not only strengthens the security of the payment system, but also helps reduce risks such as payment fraud and identity theft. The identity verification result is obtained using the formula:

[0121]

[0122] Among them, I result is the authentication result, n is the number of alarms, w i is the credit score of the ith document, x i is the safety factor of the ith document.

[0123] The risk quantification submodule uses the identity authentication results to conduct a quantitative analysis of the potential risks of the payment, evaluate the potential risks and risk levels of the payment, adjust the strength of the verification measures according to the risk level, and output the execution process of the adjusted verification requirements as follows.

[0124] The risk quantification submodule uses the identity verification results to conduct a quantitative analysis of the potential risks of the payment, including analyzing any abnormal indicators found in the identity verification results and evaluating the degree of risk caused by the indicators, such as credit risk, fraud risk, etc. Based on the analysis results, the system will classify the risks by level and adjust the intensity of the verification measures according to the risk level, such as strengthening identity verification requirements, increasing transaction monitoring, etc., to ensure that more stringent control measures can be taken for high-risk payments, and output the adjusted verification requirements, using the formula:

[0125]

[0126] Among them, R quantify is the adjusted verification requirement, n is the number of alarms, y i is the score of the ith risk indicator, z i It is the processing priority weight of the risk indicator.

[0127] See also Figure 2 , 8 , the risk feedback module includes:

[0128] The verification requirement analysis submodule evaluates the impact of the adjusted verification requirements on the payment monitoring model, identifies the logical parts that need to be adjusted, and obtains the execution process of the model parameters to be optimized for the new payment fraud situation as follows;

[0129] The verification requirement analysis submodule evaluates the impact of the adjusted verification requirements on the payment monitoring model. It involves a comprehensive evaluation of the current parameters of the model and identifies the logical parts that need to be adjusted due to new payment fraud situations. By comparing the established model parameters with the emerging fraud patterns, it identifies the parameters that fail to effectively respond to the latest payment fraud threats. It determines that the model parameters need to be optimized to improve the ability to identify new fraud strategies. Ensure that the payment monitoring model can continuously adapt to new security threats and maintain effectiveness and accuracy. The model parameters to be optimized are obtained using the formula:

[0130]

[0131] Among them, V analysis is the model parameter to be optimized, n is the number of models, p i is the current setting value of the i-th parameter, δ i The change in demand.

[0132] The model optimization submodule readjusts the parameters and logic of the payment monitoring model according to the model parameters to be optimized, updates the records according to the real-time payment data and fraud detection technology, optimizes the adaptability of the payment monitoring model, and obtains the execution process of the optimized payment monitoring model as follows;

[0133] The model optimization submodule records and updates the model parameters to be optimized based on real-time payment data and the latest fraud detection technology, and readjusts the parameters and logic of the payment monitoring model. This includes modifying parameter settings, introducing new data processing logic, and replacing or upgrading algorithms to ensure the adaptability of the model. Through regular optimization, the payment monitoring model can more accurately identify and prevent fraudulent behavior, improve the security of the entire payment system, and obtain an optimized payment monitoring model using the formula:

[0134]

[0135] Among them, M optimize is the optimized payment monitoring model, n is the number of models, q i is the new value of the i-th optimization parameter, β i is the importance weight of the parameter in the model.

[0136] The strategy building submodule uses the optimized payment monitoring model to design a management plan for potential payment risks. By identifying and avoiding fraudulent behaviors, it optimizes payment security and user trust, and outputs the execution process of the payment risk management strategy as follows;

[0137] The strategy building submodule uses the optimized payment monitoring model to design a management plan for potential payment risks. This module analyzes the data provided by the optimization model to identify fraudulent behavior and risks, including setting stricter verification steps, adjusting user authentication processes, and improving transaction monitoring mechanisms. It aims to improve payment security and enhance user trust. The payment system can more effectively avoid fraudulent behavior, ensure transaction security and user satisfaction, and output payment risk management strategies. The formula used is:

[0138]

[0139] Among them, S construct is the output payment risk management strategy, n is the number of models, r i is the management measure for the ith risk point, s i Score the potential impact of risk points.

[0140] See also Fig. 9 A payment security method based on AI artificial intelligence is implemented based on the above-mentioned payment security system based on AI artificial intelligence, and includes the following steps:

[0141] S1: Based on the user’s payment history and payment data, analyze the payment frequency and payment pattern, identify deviations from the normal pattern, and generate abnormal payment behavior characteristics;

[0142] S2: According to the characteristics of abnormal payment behavior, configure the parameters of the support vector machine classifier, train the abnormal payment pattern, verify the efficiency of the model through loop testing, and obtain the payment monitoring model;

[0143] S3: Adopt payment monitoring models, set up multi-level decision logic, adjust the input and output relationship between payment monitoring models, form a multi-layer monitoring network, and establish a payment security monitoring network;

[0144] S4: Monitor ongoing payments in real time through the payment security monitoring network, mark abnormal payments and update monitoring logs, and generate real-time abnormal transaction alerts;

[0145] S5: Use real-time abnormal transaction alerts to adjust the strength of verification measures, quantify transaction risks, review the identities of both parties to the payment, and refer to new transaction patterns and fraud methods to output payment risk management strategies.

[0146] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A payment security system based on AI artificial intelligence, characterized in that: The system comprises: The transaction behavior analysis module analyzes user behavior and payment data, uses AI to identify abnormalities in payment frequency and payment patterns, evaluates deviations from normal patterns, and outputs behavioral abnormality indicators; The model training module adopts the abnormal behavior indicators, configures the support vector machine classifier parameters, performs abnormal payment pattern training, checks the validity of the model through cyclic testing, and generates a payment monitoring model; The multi-model integration module uses AI artificial intelligence according to the payment monitoring model, sets the hierarchical judgment logic, adjusts the input and output relationship between the payment monitoring models, and forms a multi-layer monitoring network; The real-time monitoring module uses the multi-layer monitoring network for ongoing transactions, utilizes AI artificial intelligence, marks and notifies abnormal transactions, synchronizes and updates monitoring logs, and obtains real-time alarm signals; The verification strength adjustment module adjusts the strength of the verification measures according to the urgency of the real-time alarm signal, quantifies the risk of payment by reviewing the identities of both parties to the payment, and outputs the adjusted verification requirements; Based on the adjusted verification requirements and payment feedback, the risk feedback module optimizes the logic of the payment monitoring model, matches new transaction patterns and fraud methods, and generates a payment risk management strategy.

2. The payment security system based on AI artificial intelligence according to claim 1 is characterized in that: The abnormal behavior indicators include abnormal payment frequency, the degree of deviation of payment mode and abnormal transaction time. The payment monitoring model includes the parameter configuration of the model, the training data set and the test verification effect. The multi-layer monitoring network includes multi-level monitoring models, the input-output relationship of the monitoring model and the setting of hierarchical judgment logic. The real-time alarm signal includes the marking, notification and update status of the transaction monitoring log for abnormal transactions. The adjusted verification requirements include the review requirements for the identities of the two payment parties, the risk assessment indicators of the payment and the strength of the verification measures applied. The payment risk management strategy includes the optimized payment monitoring model logic, matching new transaction patterns and identifying new fraud methods.

3. The payment security system based on AI artificial intelligence according to claim 1 is characterized in that: The transaction behavior analysis module includes: The payment data analysis submodule analyzes user behavior and payment data, extracts user payment frequency, amount and time data, sorts and identifies patterns of user payment habits according to payment time series, and obtains user payment behavior characteristics; The correlation analysis submodule uses the user payment behavior characteristics and AI artificial intelligence to analyze the correlation between payment methods and locations, compare user behavior with normal payment behavior, calculate the deviation between the payment mode and the normal mode, and obtain the payment behavior analysis results; The abnormality identification submodule sets an identification threshold for abnormal payment behavior according to the payment behavior analysis result, performs real-time scanning on the payment data, identifies transaction behaviors that exceed the normal deviation range, and outputs abnormal behavior indicators.

4. The payment security system based on AI artificial intelligence according to claim 1 is characterized in that: The model training module includes: The abnormality indicator configuration submodule sets the initial parameters of the support vector machine classifier according to the abnormal behavior indicator, including error tolerance and kernel function type, to obtain a classifier benchmark configuration; The parameter adjustment submodule adopts the classifier benchmark configuration and the gradient descent algorithm to optimize the parameters of the support vector machine, and obtains the optimized classifier parameters by optimizing the classification boundaries and matching the differentiated payment behavior data; The payment test submodule utilizes the optimized classifier parameters to perform cyclic testing on the reserved payment test data, analyzes the efficiency and reaction time of payment identification, and constructs a payment monitoring model.

5. The payment security system based on AI artificial intelligence according to claim 4 is characterized in that: The formula for the gradient descent algorithm is as follows: Among them, θ is the model parameter, α is the learning rate, is the gradient of the cost function with respect to the parameter, λ is the regularization coefficient, is the gradient of the regularization term, β and φ are the amplitude and phase adjustment coefficients respectively, γ is the attenuation coefficient, σ is the attenuation rate, and t is the number of iterations.

6. The AI-based payment security system according to claim 1, characterized in that: The multi-model integration module includes: The logic setting submodule uses the payment monitoring model to define the hierarchical positions and decision logics of multiple payment monitoring models, including priority sorting and decision threshold values, to obtain hierarchical decision logic; The model relationship adjustment submodule uses AI artificial intelligence according to the hierarchical decision logic to configure the data flow and processing priority between payment monitoring models, check the flow of payment data and process it according to the predetermined logic to obtain the adjusted model relationship; The network structure integration submodule adopts the adjusted model relationship, integrates the multi-layer payment monitoring model, analyzes the transmission path of payment data between models, optimizes payment security, and outputs a multi-layer monitoring network.

7. The payment security system based on AI artificial intelligence according to claim 1 is characterized in that: The real-time monitoring module comprises: The transaction capture submodule uses AI artificial intelligence based on the multi-layer monitoring network to receive ongoing payment data in real time, analyze the characteristics of each payment, capture payments that deviate from the normal pattern, and obtain real-time payment records; The abnormal classification submodule uses a preset abnormal identification logic according to the real-time payment record to mark the abnormal payment, check the response and processing of the abnormal payment, and obtain the marked abnormal payment; The log update submodule updates the payment monitoring log according to the marked abnormal payment, records the abnormal payment detection results and payment responses, tracks and records the abnormal payment events, and outputs a real-time alarm signal.

8. The payment security system based on AI artificial intelligence according to claim 1 is characterized in that: The verification strength adjustment module includes: The alarm classification submodule classifies the real-time alarm signals according to their urgency, analyzes the priority and processing requirements of each alarm, sets associated response measures, matches payment security requirements of differentiated levels, and obtains alarm level assessment results; The identity verification submodule verifies the identities of both parties to the payment based on the alarm level assessment result, and obtains the identity verification result by checking the identity documents and payment history, evaluating the credit and security information associated with the real-time payment; The risk quantification submodule uses the identity verification results to perform quantitative analysis on the potential risks of the payment, evaluate the potential risks and risk levels of the payment, adjust the strength of the verification measures according to the risk level, and output the adjusted verification requirements.

9. The payment security system based on AI artificial intelligence according to claim 1 is characterized in that: The risk feedback module includes: The verification requirement analysis submodule evaluates the impact of the adjusted verification requirements on the payment monitoring model, identifies the logic parts that need to be adjusted, and obtains the model parameters to be optimized for the new payment fraud situation; The model optimization submodule readjusts the parameters and logic of the payment monitoring model according to the model parameters to be optimized, updates the records according to the real-time payment data and fraud detection technology, optimizes the adaptability of the payment monitoring model, and obtains the optimized payment monitoring model; The strategy building submodule uses the optimized payment monitoring model to design a management plan for potential payment risks, optimizes payment security and user trust by identifying and avoiding fraudulent behavior, and outputs a payment risk management strategy.

10. A payment security method based on AI artificial intelligence, characterized in that: According to any one of claims 1 to 9, the payment security system based on AI artificial intelligence is executed. The following steps are involved: Analyze payment frequency and payment patterns based on the user's payment history and payment data, identify deviations from normal patterns, and generate abnormal payment behavior characteristics; According to the abnormal payment behavior characteristics, the parameters of the support vector machine classifier are configured, the abnormal payment pattern is trained, the efficiency of the model is verified through a loop test, and a payment monitoring model is obtained; Adopting the payment monitoring model, setting up multi-level decision logic, adjusting the input and output relationship between the payment monitoring models, forming a multi-layer monitoring network, and establishing a payment security monitoring network; Through the payment security monitoring network, the ongoing payments are monitored in real time, abnormal payments are marked and the monitoring log is updated to generate real-time abnormal transaction alerts; The real-time abnormal transaction alert is used to adjust the strength of verification measures, and payment risk management strategies are output by quantifying transaction risks and reviewing the identities of both parties to the payment, with reference to new transaction patterns and fraud methods.

Citation Information

Patent Citations

  • Bank auditing method and system based on deep neural network

    CN118312909A

Cited By

  • Financial payment abnormal transaction identification method based on neural network

    CN119295083A

  • Mobile payment real-time supervision system based on behavior analysis algorithm

    CN119762065A

  • Payment security control method and system for data center

    CN119963186A

  • A payment security control method and system of a data middle platform

    CN119963186B

  • Internet-based AI voice payment system and use method thereof

    CN120013542A