A method for managing online payment security

Abstract

The present application discloses a method for online payment security management, which includes: obtaining a payment data set generated by online payments of a target user within a preset historical time period, dividing all payment data in the payment data set into corresponding time attribute groups based on their generation time nodes, obtaining a first sensitive value for each time attribute group based on a first sensitive indicator algorithm; obtaining a second sensitive value for each data type in the payment data set based on a coefficient of variation sensitivity evaluation algorithm; outputting M time-sensitive classes and N type-sensitive classes based on the first sensitive value of the time attribute group and the second sensitive value of the data type; utilizing a preset cross-class decryption mechanism to generate multiple classes of decryption information, and matching the class decryption information as the target decryption information based on the authorization level of the target user and the payment service bank. Thus, the sensitivity of the payment data is cross-evaluated in multiple dimensions, further improving the security and reliability of payment data sharing.

Description

Technical Field

[0001] The present invention relates to the technical field of online payment, and in particular to an online payment security management method. Background Art

[0002] Online payment, also known as electronic payment or digital payment, is a payment method completed through electronic channels such as the internet. Online payment can significantly improve consumer shopping experience and efficiency, saving time and effort. However, since online payments are susceptible to cybersecurity issues such as hacker attacks, data leaks, and transaction fraud, user transaction information is often leaked. Therefore, protecting user privacy and information security is crucial. Existing methods for managing user transaction data suffer from low security in data storage and access.

[0003] Chinese invention patent application number 202311354951.4 discloses a security monitoring method and system for online payments. The system sends a target user's transaction bill dataset to a security monitoring platform for encryption. Information sensitivity analysis is then performed on the transaction bill dataset to obtain a sensitivity index. Based on the sensitivity index, the level and quantity of multi-level decryption information are configured. Each level of decryption displays different transaction bill information. Based on the target user's authorization of each payroll bank, the system determines the first decryption information level and sends it to the payroll bank. The payroll bank receives the decryption information containing the specified level. This ensures the security and privacy of transaction bill data.

[0004] However, focusing solely on the sensitivity analysis of different data items ignores the time when transaction bills are generated. Users' transaction habits may change dynamically over time, and sensitivity analysis based solely on data items may not fully reflect the sensitivity of transaction bills. For example, a user may frequently make large payments during one period, while making primarily small payments during another. Ignoring the time dimension and sharing all data items that can be displayed in the decrypted information with the bank could expose the user's sensitive information at inappropriate times. Summary of the Invention

[0005] This application provides an online payment security management method to cross-evaluate the sensitivity of payment data in multiple dimensions, thereby further improving the security and reliability of payment data sharing.

[0006] This application provides a method for online payment security management, including:

[0007] S101: Obtain a payment dataset generated by online payments by a target user within a preset historical time period. The payment dataset includes payment data generated at different time points. The payment data at each time point includes data corresponding to several data types, including payment amount, product type, and payment duration.

[0008] S102: Based on pre-set time attribute groups, all payment data in the payment data set are divided into corresponding time attribute groups based on their generation time nodes, each time attribute group includes at least one payment data, and each payment data corresponds to a payment time node. A first sensitivity value is obtained for each time attribute group based on a first sensitivity indicator algorithm;

[0009] S103, obtaining a second sensitivity value for each data type in the payment data set according to a preset coefficient of variation sensitivity assessment algorithm;

[0010] S104: Based on the first sensitive values of all time attribute groups and the second sensitive values of all data types, a preset clustering algorithm is used to analyze and classify all time attribute groups and all data types, and M time-sensitive classes and N type-sensitive classes are output respectively;

[0011] S105: Upload the target user's payment data set to the security management platform to generate encryption information, encrypt the payment ciphertext using the encryption information, and use a preset cross-class decryption mechanism to generate multiple types of decryption information. Each type of decryption information is used to decrypt the payment ciphertext to obtain payment data corresponding to at least one data type.

[0012] S106, obtain the payment service bank end that has established a binding relationship with the target user, and according to the authorization level of the target user and the payment service bank end, match one or more types of decryption information among multiple types of decryption information as target decryption information and send them to the corresponding payment service bank end.

[0013] Preferably, the preset time attribute group is used to represent different payment time background types for target users, and all payment data are classified according to the time attribute group to which each payment data is generated based on the time node.

[0014] Preferably, the first sensitive indicator algorithm is set to:

[0015]

[0016] Among them, SI is the first sensitive value of the time attribute group, SI i is the sensitive value of the i-th payment data in the time attribute group, w i is the weight factor of the i-th payment data, and n is the total number of payment time nodes in this time attribute group.

[0017] Preferably, the method for setting the weight factor of the i-th payment data in the time attribute group includes:

[0018] A1. Obtain all payment time nodes of the target user in a preset historical time period. Use the K-means clustering algorithm to perform cluster analysis on all payment time nodes to obtain K clusters. The center point of each cluster is used as the cluster label. Each cluster includes several payment time nodes.

[0019] A2. For each cluster, calculate the difference between its center and the current time, and then calculate the proportion of this difference in the sum of the differences between all cluster centers and the current time as the time value ratio of the cluster.

[0020] A3. Obtain the payment time node of the i-th payment data, obtain the cluster to which it belongs, and use the time value ratio of the cluster as the weight factor of the i-th payment data.

[0021] Preferably, the method for determining the sensitive value of the i-th payment data in the time attribute group includes:

[0022] B1. Obtain the payment amount, payment duration, and product type from the payment data to form a payment feature vector;

[0023] B2. Input the payment feature vector into the pre-trained payment sensitivity prediction model and output the sensitivity value corresponding to the payment feature vector;

[0024] Methods for obtaining the pre-trained payment-sensitive prediction model include:

[0025] Collect payment feature vectors of a large number of payment time nodes belonging to the time attribute group in the history of the target user, and label each payment feature vector; the label content is set as: sensitive value;

[0026] The labeled payment feature vectors are used as training sets, and the pre-selected neural network structure is trained using the training sets to optimize the model parameters and obtain the final payment sensitivity prediction model.

[0027] Preferably, labeling each payment feature vector includes:

[0028] Collect statistics on all push content received by the target user from the payment service bank within a preset historical time period, calculate the hit probability value of the product type in the payment feature vector in the push content, and calculate the sensitivity value marked by the payment feature vector according to the following formula:

[0029]

[0030] Among them, S is the sensitive value marked by the payment feature vector, p is the hit probability value of the product type in the payment feature vector, T is the payment duration, α and β are pre-set weight values, which are used to represent the influence of product type and payment duration on the sensitive value respectively.

[0031] Preferably, the preset coefficient of variation sensitive assessment algorithm specifically includes:

[0032] S201, digitizing all data in each data type according to a preset coding rule to generate a numerical sequence corresponding to each data type;

[0033] S202, based on the numerical sequence of each data type, calculate its coefficient of variation according to the following formula:

[0034]

[0035] Among them, CV is the coefficient of variation of the data type, μ is the average value of all values in the numerical sequence of the data type, q is the total number of values in the numerical sequence, and x is the average value of the data type. i is the i-th value in the numerical sequence;

[0036] S203: Taking the coefficient of variation of each data type as its second sensitive value.

[0037] Preferably, in S104, each of the time-sensitive classes includes at least one time attribute group, and each of the type-sensitive classes includes at least one data type;

[0038] The preset clustering algorithm is set to the K-means clustering algorithm, which takes the center point of each time-sensitive class as the first representative value of the time-sensitive class, and takes the center point of each type-sensitive class as the second representative value of the type-sensitive class.

[0039] Preferably, the preset cross-class decryption mechanism specifically includes:

[0040] C1. Based on M time-sensitive classes and their corresponding first representative values, and N type-sensitive classes and their corresponding second representative values, generate an M×N time-type cross-class matrix. The value of each element in the matrix is set to the comprehensive sensitivity value of the corresponding time-type cross-class: the weighted sum of the corresponding first representative value and the second representative value;

[0041] C2. Obtain the payment data associated with each element in the time-type intersection matrix. The associated payment data is set to the intersection of the payment data in the time-sensitive class corresponding to the element and the payment data summarized in the corresponding type-sensitive class.

[0042] C3. Based on all elements in the time-type cross-class matrix, perform a merging process on them to obtain R elements, where R ≤ M × N;

[0043] C4. Based on the R elements and their respectively associated payment data and payment ciphertexts, generate R types of decryption information. Among them, the R elements correspond one-to-one with the R types of decryption information, and each type of decryption information is used to decrypt the payment ciphertext to obtain the payment data associated with the element corresponding to this type of decryption information.

[0044] Preferably, the authorization levels between the target user and the payment service bank side include low, medium, and high. The method of matching one or more types of decryption information as the target decryption information among multiple types of decryption information according to the authorization level between the target user and the payment service bank side specifically includes:

[0045] According to the pre-set authorization level-sensitive acceptance interval matching table, obtain the sensitive acceptance interval corresponding to the authorization level of the payment service bank side, traverse the element values corresponding to multiple types of decryption information, and determine all types of decryption information whose element values fall within the sensitive acceptance interval as the target decryption information;

[0046] Among them, the pre-set authorization level-sensitive acceptance interval matching table is specifically: low - [0, a), medium - [a, b), high - [b, 1], 0 < a < b < 1, and the values of a and b are set according to the actual situation, which are used to reflect the boundaries of different comprehensive sensitivity values.

[0047] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0048] By introducing the time attribute group, the consumption habits and privacy preferences of users in different time backgrounds are considered, and the sensitivity degree of payment data in the time dimension is more comprehensively reflected; by calculating the first sensitivity value of the time attribute group and the second sensitivity value of the data type, and using the payment sensitivity prediction model to evaluate the sensitivity degree of payment data, a dynamic and accurate sensitivity evaluation is achieved; by generating multiple types of decryption information and matching and sending the target decryption information according to the authorization level, flexible data access control is achieved, which not only protects user privacy but also meets the specific information requirements of the payment service bank side.

[0049] A cross-class decryption mechanism is introduced to generate multiple types of decryption information, among which the time-type cross-class matrix plays a core role. The matrix organizes payment data into a structured form with time attributes and data types as dimensions. Through the matrix, the distribution of payment data in different time backgrounds and of different types can be clearly seen; according to the element distribution and sensitivity assessment results in the matrix, specific decryption information can be generated for different data types and time backgrounds, which improves the pertinence of decryption information, enhances the flexibility of data access control, and improves the security and efficiency of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 Schematic diagram of the process of online payment security management method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0051] To facilitate understanding of the present invention, the present application will be described more comprehensively below with reference to the relevant drawings; the drawings show preferred embodiments of the present invention, but the present invention can be implemented in many different forms and is not limited to the embodiments described herein; on the contrary, the purpose of providing these embodiments is to enable a more thorough and comprehensive understanding of the disclosed content of the present invention.

[0052] It should be noted that the terms “vertical”, “horizontal”, “up”, “down”, “left”, “right” and similar expressions used in this document are for illustrative purposes only and do not represent the only implementation method.

[0053] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this invention pertains; the terms used herein in the specification of the present invention are for the purpose of describing specific embodiments only and are not intended to limit the present invention; the term "and / or" used herein includes any and all combinations of one or more of the associated listed items.

[0054] Example 1: Figure 1 It is a flowchart of the online payment security management method according to an embodiment of the present invention.

[0055] like Figure 1 As shown, an online payment security management method includes the following steps:

[0056] S101, obtain a payment data set generated by the target user's online payment within a preset historical time period, the payment data set includes payment data generated at different time nodes, and the payment data at each time node includes data corresponding to several data types, including but not limited to payment amount, payment channel, product type, payment duration, etc.

[0057] Among them, the preset historical time period can be set to half a year, and can be set specifically according to actual conditions. It can also be set according to the cycle of updating it to the payment security management platform. The present invention will not elaborate on this.

[0058] S102: Based on a pre-set time attribute group (which can be set to a null value in advance), all payment data in the payment data set are divided into corresponding time attribute groups based on their generation time nodes, each time attribute group includes at least one payment data, and each payment data corresponds to a payment time node, and a first sensitive value of each time attribute group is obtained based on a first sensitive indicator algorithm.

[0059] In payment data analysis, time is an important dimension. Payment behaviors at different time points (such as weekdays, weekends, and holidays) may differ significantly. These differences reflect the target user's consumption habits and privacy preferences in different time contexts. The division of time context can be set according to the target user. For example, if the target user is an employee of a company, the time context can be set to working hours, off-get off work hours, weekends, holidays, the time window after payday (used to reflect the target user's payment habits within a fixed time window after payday), and the time window before payday (used to reflect the target user's payment habits within a fixed time window before payday). Payment behavior during working hours may be work-related. After get off work and on weekends, users have more leisure time, and payment behavior may be more inclined towards entertainment and personal consumption. Payment behavior on holidays may be affected by the festive atmosphere and promotional activities.

[0060] Therefore, the preset time attribute group is used to represent different payment time background types for target users. Based on the generation time node of each payment data, all payment data are classified according to the time attribute group to which it belongs.

[0061] In some embodiments, the first sensitive indicator algorithm is set as:

[0062]

[0063] Among them, SI is the first sensitive value of the time attribute group, SI i is the sensitive value of the i-th payment data in the time attribute group, w i is the weight factor of the i-th payment data, and n is the total number of payment time nodes in the time attribute group (it can also be understood as the total number of payment data, one payment time node corresponds to one payment data).

[0064] The method for setting the weight factor of the i-th payment data in the time attribute group includes:

[0065] A1. Obtain all payment time nodes of the target user in a preset historical time period. Perform cluster analysis on all payment time nodes using an existing clustering algorithm (K-means clustering algorithm). Clustering is based on the difference between any two payment time nodes, which can be used to divide the payment-intensive time periods within the preset historical time period. K clusters are obtained. The center point (central payment time node) of each cluster is used as the label for the cluster. Each cluster includes several payment time nodes.

[0066] For example, K payment time nodes are randomly selected as initial cluster centers. These cluster centers represent the centers of K potential payment-intensive time periods. For each payment time node, its distance to the K cluster centers is calculated. Each payment time node is assigned to the cluster represented by the cluster center closest to it. Once all payment time nodes are assigned to clusters, the center of each cluster is recalculated by taking the average of all payment time nodes in the cluster. The above assignment and updating steps are repeated until the cluster center no longer changes significantly or the preset number of iterations is reached, indicating that the algorithm has converged and the clustering results have stabilized.

[0067] It should be noted that the working principle of the K-means clustering algorithm is not described in detail in the present invention, and reference may be made to the relevant prior art.

[0068] A2. For each cluster, calculate the difference between its center point and the current time, and calculate the proportion of this difference in the sum of the differences between all cluster centers and the current time as the time value ratio of the cluster.

[0069] A3. Obtain the payment time node of the i-th payment data, determine the cluster to which it belongs, and use the time value ratio of the cluster as the weight factor for the i-th payment data. Therefore, the closer the payment data is to the current time, the more it reflects the user's latest payment behavior habits, and the higher the reference value and sensitivity of the payment data in terms of time.

[0070] The method for determining the sensitive value of the i-th payment data in the time attribute group includes:

[0071] B1. Obtain the payment amount, payment duration, and product type from the payment data to form a payment feature vector;

[0072] B2. Input the payment feature vector into the pre-trained payment sensitivity prediction model and output the sensitivity value corresponding to the payment feature vector.

[0073] Methods for obtaining the pre-trained payment-sensitive prediction model include:

[0074] Collect payment feature vectors of a large number of payment time nodes belonging to the time attribute group in the target user's history, label each payment feature vector, and set the label content to: sensitive value; wherein, the payment feature vector labeling method is: count all push content received by the target user from the payment service bank within a preset historical time period, calculate the hit probability value of the product type in the payment feature vector in the push content (the proportion of push content related to the product type and all push content can be calculated), and calculate the sensitive value labeled by the payment feature vector according to the following formula:

[0075]

[0076] Where S is the sensitive value marked by the payment feature vector, p is the hit probability value of the product type in the payment feature vector, T is the payment duration, and α and β are pre-set weights, respectively used to represent the degree of influence of product type and payment duration on the sensitive value;

[0077] The labeled payment feature vectors are used as training sets, and the pre-selected neural network structure is trained using the training sets to optimize the model parameters and obtain the final payment sensitivity prediction model.

[0078] Therefore, by introducing the time dimension and setting time attribute groups to reflect users' consumption habits and privacy preferences under different time background types, the first sensitive value of each time attribute group is obtained for subsequent analysis and classification, thereby enabling a more comprehensive assessment of the sensitivity of payment data in the time dimension.

[0079] S103: Obtain a second sensitive value for each data type in the payment data set according to a preset coefficient of variation sensitivity evaluation algorithm.

[0080] In some embodiments, the preset coefficient of variation sensitive evaluation algorithm specifically includes:

[0081] S201: digitize all data in each data type according to a preset coding rule to generate a numerical sequence corresponding to each data type.

[0082] Among them, when the data in the data type is numerical, there is no need to digitize it, and a numerical sequence can be directly generated; when the data in the data type is not numerical, all data in the data type are digitized using preset encoding rules.

[0083] For example, when the data type is payment amount, the amount value is used directly; when the data type is payment channel, the payment channel is converted to the numerical representation of the categorical variable, for example, bank card is 1, Alipay is 2, and WeChat Pay is 3; when the data type is product type, the product type is converted to the numerical representation of the categorical variable, for example, product is 1, service is 2, and subscription is 3.

[0084] S202, based on the numerical sequence of each data type, calculate its coefficient of variation according to the following formula:

[0085]

[0086] Among them, CV is the coefficient of variation of the data type, μ is the average value of all values in the numerical sequence of the data type, q is the total number of values in the numerical sequence, and x is the average value of the data type. i is the i-th value in the numerical sequence.

[0087] S203: Taking the coefficient of variation of each data type as its second sensitive value. The larger the coefficient of variation, the higher the degree of discreteness of the data type and the higher the sensitivity.

[0088] S104: Based on the first sensitive values of all time attribute groups and the second sensitive values of all data types, a preset clustering algorithm is used to analyze and classify all time attribute groups and all data types, and M time-sensitive classes and N type-sensitive classes are output respectively. Each time-sensitive class includes at least one time attribute group, and each type-sensitive class includes at least one data type.

[0089] Specifically, the preset clustering algorithm is set to the K-means clustering algorithm, which takes the center point of each time-sensitive class (center first sensitive value) as the first representative level value of the time-sensitive class, and takes the center point of each type-sensitive class (center second sensitive value) as the second representative level value of the type-sensitive class.

[0090] For example, for all time attribute groups, randomly select M first sensitive values as the initial cluster centers of the time sensitive class; for all data types, randomly select N second sensitive values as the initial cluster centers of the type sensitive class; for the time attribute group, calculate the distance from the first sensitive value of each time attribute group to the M time sensitive class cluster centers, and assign it to the cluster with the closest distance; for the data type, calculate the distance from the second sensitive value of each data type to the N type sensitive class cluster centers, and assign it to the cluster with the closest distance; for each time sensitive class, recalculate the mean of the first sensitive values of all time attribute groups it contains as the new cluster center; for each type sensitive class, recalculate the mean of the second sensitive values of all data types it contains as the new cluster center; repeat the above allocation and update steps until the cluster center no longer changes significantly or reaches the preset number of iterations. The specific cluster analysis process can refer to the relevant existing technology, and the present invention will not elaborate on this.

[0091] S105, upload the target user's payment data set to the security management platform to generate encrypted information, use the encrypted information to encrypt to obtain the payment ciphertext, and use the preset cross-level decryption mechanism to generate multiple types of decryption information. After each type of decryption information decrypts the payment ciphertext, it obtains payment data corresponding to at least one data type.

[0092] In some embodiments, the preset cross-class decryption mechanism specifically includes:

[0093] C1. Based on the M time-sensitive classes and their corresponding first representative values, and the N type-sensitive classes and their corresponding second representative values, generate an M×N time-type intersection class matrix. The value of each element in the matrix (a total of M×N elements) is set to the comprehensive sensitivity value of the corresponding time-type intersection class (the weighted sum of the corresponding first representative value and the second representative value);

[0094] C2. Obtain the payment data associated with each element in the time-type intersection matrix. The associated payment data is set to the intersection of the payment data in the time-sensitive class corresponding to the element and the payment data summarized in the corresponding type-sensitive class.

[0095] C3. Based on all elements in the time-type intersection matrix, merge them (merge elements with the same value and merge their associated payment data) to obtain R elements, where R ≤ M × N;

[0096] C4. Generate R types of decryption information based on the R elements and their associated payment data and payment ciphertext. The R elements correspond one-to-one to the R types of decryption information. Each type of decryption information is used to decrypt the payment ciphertext to obtain the payment data associated with the element corresponding to the type of decryption information.

[0097] S106, obtain the payment service bank end that has established a binding relationship with the target user, and according to the authorization level of the target user and the payment service bank end, match one or more types of decryption information among multiple types of decryption information as target decryption information and send them to the corresponding payment service bank end.

[0098] The payment service bank with which the target user establishes a binding relationship communicates with the security management platform. The security management platform is a system for securely managing transaction data and is embedded with a data encryption module. The data encryption module is used to generate encrypted information based on the target user's characteristics. The module stores data encryption algorithms, such as the AES algorithm, the DSA algorithm, and the SHA2 algorithm. Those skilled in the art can select an appropriate encryption algorithm based on actual needs. Multiple types of decryption information are configured based on the target user's encrypted information. Each type of decryption information is used to decrypt the payment ciphertext, and the decrypted payment data of each type of decryption information is different.

[0099] In some embodiments, the authorization levels between the target user and the payment service bank end include low, medium, and high, which can be determined according to a pre-signed authorization agreement. According to the authorization level between the target user and the payment service bank end, one or more types of decryption information are matched as the target decryption information from multiple types of decryption information. Specifically, it includes:

[0100] According to a pre-set authorization level - sensitive acceptance interval matching table, obtain the sensitive acceptance interval corresponding to the authorization level of the payment service bank end, traverse the element values corresponding to multiple types of decryption information, and determine all types of decryption information whose element values fall within the sensitive acceptance interval as the target decryption information.

[0101] Among them, the pre-set authorization level - sensitive acceptance interval matching table is specifically:

[0102] Low - [0, a), Medium - [a, b), Medium - [b, 1], where 0 < a < b < 1, and the values of a and b are set according to the actual situation, which are used to reflect the boundaries of different comprehensive sensitivity values.

[0103] Therefore, according to the authorization level of the target user, the target decryption information is matched from multiple types of decryption information and sent to the payment service bank end that meets the authorization level, ensuring that only authorized entities can access specific-level payment data. The payment service bank end receives the target decryption information that matches its authorization level, realizing secure payment data sharing, protecting user privacy, and at the same time meeting the specific information needs of the payment service bank end.

[0104] The technical solutions in the embodiments of the present application described above have at least the following technical effects or advantages:

[0105] By introducing the time attribute group, the consumption habits and privacy preferences of users in different time backgrounds are considered, and the sensitivity degree of payment data in the time dimension is more comprehensively reflected; by calculating the first sensitivity value of the time attribute group and the second sensitivity value of the data type, and using the payment sensitivity prediction model to evaluate the sensitivity degree of payment data, dynamic and accurate sensitivity evaluation is achieved; by generating multiple types of decryption information and matching and sending the target decryption information according to the authorization level, flexible data access control is achieved, which not only protects user privacy but also meets the specific information needs of the payment service bank end.

[0106] A cross-class decryption mechanism is introduced to generate multiple types of decryption information, among which the time-type cross-class matrix plays a core role. The matrix organizes payment data into a structured form with time attributes and data types as dimensions. Through the matrix, the distribution of payment data in different time backgrounds (such as working hours, off-get off work hours, weekends, etc.) and different types (such as payment amount, payment duration, product type, etc.) can be clearly seen; according to the element distribution and sensitivity assessment results in the matrix, specific decryption information can be generated for different data types and time backgrounds, which improves the pertinence of decryption information, enhances the flexibility of data access control, and improves the security and efficiency of the system.

[0107] Example 2: In traditional methods, the authorization levels of the target user and the payment service bank are fixed or determined in a pre-defined agreement. However, due to the complexity and variability of the online payment environment and the dynamic changes in the payment habits of the target user, their service experience with different payment service banks is also constantly changing. Therefore, the traditional fixed authorization level marking method has certain limitations.

[0108] Therefore, the embodiments of the present application are optimized based on the above embodiments.

[0109] In some embodiments, the method further comprises:

[0110] S107, periodically executing steps S101 to S106 according to a preset time interval, to periodically share payment data for the payment service bank bound to the target user, so that the bank can send promotional content in a targeted manner based on the payment data of the target user, so that the bank can send promotional content in a targeted manner based on the latest payment data.

[0111] In some embodiments, in step S106, the method for determining the authorization level between the target user and the payment service bank further includes:

[0112] S301, when entering the next cycle, obtain all payment data shared to the payment service bank in the previous cycle, extract all product types and their corresponding payment amounts in the payment data, and form a product feature vector for each product type and corresponding payment amount.

[0113] S302: Collect all product feature vectors to obtain a product feature vector set.

[0114] S303: Count all product types and their corresponding value amounts that the target user has interacted with on the push content of the payment service bank in the previous cycle, and generate a set of valid product feature vectors (each valid product feature vector is generated by a product type with which there is an interactive behavior and its corresponding value amount).

[0115] Among them, interactive behaviors can be set to click, collect, place an order, etc.

[0116] S304: Based on the product feature vector set, the valid product feature vector set is screened out (if the valid product feature vector does not exist in the product feature vector set, it is screened out) to obtain a new valid product feature vector set.

[0117] S305, obtain the pushed content of the payment service bank in the previous cycle, calculate the hit probability value of the new valid product feature vector set in the pushed content, and redefine its authorization level based on the hit probability value. The larger the hit probability value, the less invalid content the payment service bank pushes to the target user, and the higher its authorization level.

[0118] Specifically, when the hit probability value is greater than the upper limit of the preset hit probability threshold range, the authorization level is set to high; when the hit probability value is less than the lower limit of the preset hit probability threshold range, the authorization level is set to low; otherwise, the authorization level is set to medium. The hit probability threshold range is set based on actual conditions and expert experience.

[0119] The technical solutions in the above embodiments of the present application have at least the following technical effects or advantages:

[0120] By periodically redefining the authorization level, the authorization level can be dynamically adjusted as the payment habits of target users and the content pushed by the payment service bank change. This improves the flexibility and adaptability of the authorization level, making it more in line with actual conditions.

[0121] By redefining authorization levels based on target users' payment data and interactions with push content over the previous cycle, payment service banks can more accurately understand target users' preferences and needs, helping them deliver more targeted promotional content, improving promotion effectiveness and user satisfaction.

[0122] For target users with higher authorization levels, the bank can invest more resources in precision marketing; for users with lower authorization levels, it can reduce resource investment to avoid waste; in addition, the bank's authorization level can be dynamically adjusted based on historical interaction data between the target user and the bank, which can further improve the security of user payment data management, share specific levels of payment data with the bank, and avoid the leakage of sensitive data, which may cause security risks to the target user's payment.

[0123] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Various modifications and variations are readily apparent to those skilled in the art. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

Claims

1. An online payment security management method, characterized in that: include: S101: Obtain a payment dataset generated by online payments by a target user within a preset historical time period. The payment dataset includes payment data generated at different time points. The payment data at each time point includes data corresponding to several data types, including payment amount, product type, and payment duration. S102: Based on pre-set time attribute groups, all payment data in the payment data set are divided into corresponding time attribute groups based on their generation time nodes, each time attribute group includes at least one payment data, and each payment data corresponds to a payment time node. A first sensitivity value is obtained for each time attribute group based on a first sensitivity indicator algorithm; S103, obtaining a second sensitivity value for each data type in the payment data set according to a preset coefficient of variation sensitivity assessment algorithm; S104: Based on the first sensitive values of all time attribute groups and the second sensitive values of all data types, a preset clustering algorithm is used to analyze and classify all time attribute groups and all data types, and M time-sensitive classes and N type-sensitive classes are output respectively; the center point of each time-sensitive class is used as the first representative value of the time-sensitive class, and the center point of each type-sensitive class is used as the second representative value of the type-sensitive class; S105: Upload the target user's payment data set to the security management platform to generate encryption information. The encryption information is used to encrypt the payment ciphertext. Then, a preset cross-class decryption mechanism is used to generate multiple types of decryption information. Each type of decryption information decrypts the payment ciphertext to obtain payment data corresponding to at least one data type. The cross-class decryption mechanism is as follows: Based on M time-sensitive classes and their corresponding first representative values, and N type-sensitive classes and their corresponding second representative values, an M×N time-type cross-class matrix is generated. The value of each element in the matrix is set to the comprehensive sensitivity value of the corresponding time-type cross-class: the weighted sum of the corresponding first representative value and the second representative value; Obtain the payment data associated with each element in the time-type intersection matrix: the intersection of the payment data in the time-sensitive class corresponding to that element and the payment data summarized in the corresponding type-sensitive class; merge all elements in the time-type intersection matrix to obtain R elements, where R ≤ M × N; generate R decryption information based on the R elements and their associated payment data and payment ciphertext, where each R element corresponds to the R decryption information; S106, obtain the payment service bank end that has established a binding relationship with the target user, and according to the authorization level of the target user and the payment service bank end, match one or more types of decryption information among multiple types of decryption information as target decryption information and send them to the corresponding payment service bank end.

2. The online payment security management method according to claim 1, characterized in that: The preset time attribute group is used to represent different payment time background types for target users. Based on the generation time node of each payment data, all payment data are classified according to the time attribute group to which it belongs.

3. The online payment security management method according to claim 1, wherein: The first sensitive indicator algorithm is set to: in, is the first sensitive value of the time attribute group, is the sensitive value of the i-th payment data in the time attribute group, is the weight factor of the i-th payment data, and n is the total number of payment time nodes in this time attribute group.

4. The online payment security management method according to claim 3, wherein: The method for setting the weight factor of the i-th payment data in the time attribute group includes: A1. Obtain all payment time nodes of the target user in a preset historical time period. Use the K-means clustering algorithm to perform cluster analysis on all payment time nodes to obtain K clusters. The center point of each cluster is used as the cluster label. Each cluster includes several payment time nodes. A2. For each cluster, calculate the difference between its center and the current time, and then calculate the proportion of this difference in the sum of the differences between all cluster centers and the current time as the time value ratio of the cluster. A3. Obtain the payment time node of the i-th payment data, obtain the cluster to which it belongs, and use the time value ratio of the cluster as the weight factor of the i-th payment data.

5. The online payment security management method according to claim 3, characterized in that: The method for determining the sensitive value of the i-th payment data in the time attribute group includes: B1. Obtain the payment amount, payment duration, and product type from the payment data to form a payment feature vector; B2. Input the payment feature vector into the pre-trained payment sensitivity prediction model and output the sensitivity value corresponding to the payment feature vector; Methods for obtaining the pre-trained payment-sensitive prediction model include: Collect payment feature vectors of a large number of payment time nodes belonging to the time attribute group in the history of the target user, and label each payment feature vector; the label content is set as: sensitive value; The labeled payment feature vectors are used as training sets, and the pre-selected neural network structure is trained using the training sets to optimize the model parameters and obtain the final payment sensitivity prediction model.

6. The online payment security management method according to claim 5, characterized in that: The labeling of each payment feature vector includes: Collect statistics on all push content received by the target user from the payment service bank within a preset historical time period, calculate the hit probability value of the product type in the payment feature vector in the push content, and calculate the sensitivity value marked by the payment feature vector according to the following formula: in, S is the sensitive value marked by the payment feature vector, p is the hit probability value of the product type in the payment feature vector, T is the payment duration, and These are pre-set weight values, which are used to indicate the impact of product type and payment duration on sensitive values.

7. The online payment security management method according to claim 1, wherein: The preset coefficient of variation sensitivity assessment algorithm specifically includes: S201, digitizing all data in each data type according to a preset coding rule to generate a numerical sequence corresponding to each data type; S202, based on the numerical sequence of each data type, calculate its coefficient of variation according to the following formula: in, is the coefficient of variation of the data type, is the average value of all values in the numerical sequence of this data type, and q is the total number of values in the numerical sequence. is the i-th value in the numerical sequence; S203: Taking the coefficient of variation of each data type as its second sensitive value.

8. The online payment security management method according to claim 1, wherein: In S104, each of the time-sensitive classes includes at least one time attribute group, and each of the type-sensitive classes includes at least one data type; The preset clustering algorithm is set to the K-means clustering algorithm.

9. The online payment security management method according to claim 8, characterized in that: The authorization levels of the target user and the payment service bank include low, medium, and high. According to the authorization levels of the target user and the payment service bank, one or more types of decryption information are matched as target decryption information among multiple types of decryption information, specifically including: According to the pre-set authorization level-sensitive acceptance interval matching table, the sensitive acceptance interval corresponding to the authorization level of the payment service bank is obtained, and the element values corresponding to multiple types of decryption information are traversed, and all types of decryption information whose element values fall into the sensitive acceptance interval are determined as target decryption information; Among them, the pre-set authorization level-sensitive acceptance interval matching table is specifically: low - [0, a), medium - [a, b), high - [b, 1], where 0 < a < b < 1, and the values of a and b are set according to the actual situation, and are used to reflect the boundaries of different comprehensive sensitivity values.

Images