An information risk analysis method for cross-border e-commerce and a cloud server
Patent Information
- Application Number
- CN202411383851.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-30
- Publication Date
- 2026-10-27
AI Technical Summary
传统的风险识别方法往往依赖于历史数据和简单的规则设定,难以适应瞬息万变的交易环境
(1)该方法通过用户客户端和云服务器的协同工作,实现了对用户登录、交易活动及售后行为的全面监控。这一过程通过自动采集用户行为数据与交易数据,确保了数据的实时性和准确性,为后续的风险分析提供了丰富的基础数据。此外,利用HTTPS协议保障数据传输的安全性,进一步提升了信息系统的可靠性。
Smart Images

Figure CN122989105A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of information risk detection technology, specifically to an information risk analysis method and cloud server for cross-border e-commerce. Background Technology
[0002] With the rapid development of cross-border e-commerce, the frequency and complexity of online transactions have increased significantly, leading to increasingly prominent information risks. Many businesses face challenges such as user identity theft, payment fraud, and malicious returns, and traditional risk assessment methods are unable to effectively address these emerging threats. To solve this problem, developing a risk assessment system based on real-time monitoring and data analysis has become an urgent priority. This can not only improve security but also enhance user experience and promote the sustainable development of cross-border e-commerce.
[0003] Currently, many cross-border e-commerce platforms still have significant shortcomings in analyzing abnormal user behavior. Traditional risk identification methods often rely on historical data and simple rule settings, making it difficult to adapt to the rapidly changing trading environment. Furthermore, existing systems monitor user behavior data in a static manner, failing to capture subtle changes in user behavior in a timely manner, resulting in delayed responses in risk management and a lack of dynamic real-time analysis capabilities. In such cases, the abnormal behavior of malicious users may go undetected for extended periods, leading to increased potential losses. Failure to promptly identify and handle abnormal behavior can turn the platform into a breeding ground for cybercriminals, further exacerbating information security issues. Moreover, existing risk assessment mechanisms have relatively limited depth and breadth of data analysis, making it difficult to comprehensively identify complex abnormal behavior patterns. Summary of the Invention
[0004] To address the shortcomings of existing technologies, this invention provides an information risk analysis method and cloud server for cross-border e-commerce, solving the problems mentioned in the background section.
[0005] To achieve the above objectives, the present invention provides the following technical solution: an information risk analysis method for cross-border e-commerce, comprising the following steps: S1. Real-time monitoring of user behavior data and user transaction data through cross-border e-commerce client and cloud server respectively, automatic collection and recording of relevant user information, and real-time transmission of user behavior data collected by cross-border e-commerce client to cloud server via Internet. S2. The cloud server receives user behavior data collected from the cross-border e-commerce client in real time, stores the user behavior data and user transaction data through sharding and partitioning storage, and then preprocesses the data through the SQL script of the storage cloud server to obtain the transaction behavior dataset. S3. The cloud server performs aggregate calculations based on the acquired transaction behavior dataset to obtain the user login anomaly index DLY and the user order anomaly index DDY. S4. The cloud server summarizes and calculates the user login anomaly index DLY and user order anomaly index DDY to obtain the comprehensive abnormal behavior risk coefficient YCF, and conducts a preliminary comparison and evaluation with the preset abnormal behavior risk threshold A. Based on the evaluation results, the second evaluation mechanism is activated. S5, cloud server uses PCA to reduce the dimensionality of risk data stored in categories, and then uses the visualization tool Excel to create a scatter plot, and plots the coordinate concentration of high-risk users to obtain the location of cluster center JL. Normal users are distinguished and monitored by preset area with cluster center JL as the center point and radius R, and the number of times users enter this area is recorded to obtain the frequency of entering and leaving the cluster center pl. S6. The cloud server calculates the distance YHJ between the user and the cluster center based on the obtained distance to the cluster center JL and the frequency pl of the cluster center. When the user does not have abnormal behavior in the initial assessment, a secondary assessment mechanism is triggered. The secondary assessment mechanism compares the obtained distance YHJ between the user and the cluster center with a preset area with the cluster center JL as the center point and a radius of R, and takes corresponding measures based on the assessment results.
[0006] Preferably, S1 specifically includes the following steps: S11. Monitor user transaction activities in real time through the cross-border e-commerce client, record each user's login status, analyze abnormal login frequency, and detect login time zone, login IP address, and login device machine code in real time during login; when users log in and perform transaction operations, monitor order and shopping-related operation data in real time through the cross-border e-commerce client, track user address changes in different orders, and determine if there is frequent address changes; monitor order amount fluctuations and analyze if there are abnormally large orders or abnormal price deviations; record the time required for users to add goods to their shopping cart, determine if it is an abnormally fast or slow operation, identify automated machine operation behavior, and obtain user behavior data; S12. The cloud server is integrated and connected to e-commerce websites and cross-border e-commerce clients via the Internet. Then, through tracking code, it tracks the after-sales requests of users in e-commerce websites and cross-border e-commerce clients in real time, tracks the return behavior of each user, obtains the return rate based on the return behavior, and identifies potential malicious returns and fraudulent behavior by analyzing the return rate, thereby obtaining user transaction data. S13. The cloud server connects to the cross-border e-commerce client via the Internet, establishes communication with the cloud server through the HTTPS protocol, and transmits the user behavior data and user transaction data collected by the cross-border e-commerce client to the cloud server in real time via the Internet.
[0007] Preferably, S2 specifically includes the following steps: S21. The cloud server receives user behavior data collected by the cross-border e-commerce client in real time, stores the user behavior data and user transaction data monitored by the cloud server in real time according to the storage method of sharding and partitioning, and then accesses the data through the indexing function. Sharding: Cloud servers horizontally shard data, distributing large-scale data across multiple physical cloud servers for storage; Partitioning: Data is stored in partitions based on time, region, transaction type, payment method, order amount, product category, and supplier. S22. The cloud server preprocesses user behavior data and user transaction data using SQL scripts. The preprocessing includes missing value handling, outlier handling, duplicate data handling, standardization, and normalization to obtain a transaction behavior dataset. The transaction behavior dataset includes user login dataset and user order dataset; The user login dataset includes login frequency dl, time zone change sq, IP address change ip, device change sb, and address change count dz; The user order dataset includes order amount deviation je, time required to add items to cart sj, and return rate th.
[0008] Preferably, S3 specifically includes the following steps: S31. The cloud server performs summary calculations based on the obtained user login dataset to obtain the user login anomaly index DLY. The user login anomaly index DLY is calculated using the following formula; ; In the formula, log represents the logarithmic function, and e represents the exponential function of the time zone variation sq. S32. The cloud server performs summary calculations based on the acquired user order dataset to obtain the user order anomaly index DDY. The user order anomaly index DDY is calculated using the following formula; ; In the formula, e represents the negative exponential function of the time sj required to add the item to the shopping cart, and sin represents the sine function. Representing pi, it can be combined with a sine function to help construct the periodic variation characteristics of user behavior.
[0009] Preferably, S4 specifically includes the following steps: S41. The cloud server calculates the comprehensive abnormal behavior risk coefficient YCF by summarizing and calculating the obtained user login anomaly index DLY and user order anomaly index DDY. The comprehensive abnormal behavior risk coefficient YCF is calculated using the following formula; ; In the formula, w1 and w2 represent the weighting coefficients of the user login anomaly index DLY and the user order anomaly index DDY, respectively, and w1+w2=1. The specific values are set by the user according to the situation.
[0010] Preferably, S42, based on the average value of historical user risk data stored in the cloud server, a preset abnormal behavior risk threshold A is set, and a preliminary comparison and evaluation is performed with the obtained comprehensive abnormal behavior risk coefficient YCF. Based on the evaluation results, corresponding measures are implemented. The specific evaluation scheme is as follows. When the comprehensive abnormal behavior risk coefficient YCF ≥ the preset abnormal behavior risk threshold A, it indicates that the user has abnormal behavior. At this time, the cloud server will verify the user's identity through the payment platform. If the identity verification fails, the user's account will be frozen, the user will be marked as a risk user, and the risk data of the marked risk users will be classified and stored in one category. When the comprehensive abnormal behavior risk coefficient YCF is less than the preset abnormal behavior risk threshold A, it means that the user does not have abnormal behavior. At this time, monitoring will continue and the second assessment mechanism will be triggered.
[0011] Preferably, S5 specifically includes the following steps: S51: Based on the risk data stored in the cloud server, the PCA dimensionality reduction technology is used to project the data into a low-dimensional space while retaining the characteristics of the data. S52. After data dimensionality reduction, a scatter plot is created using the visualization tool Excel. Scatter points are drawn on the plot based on the dimensionality reduction coordinates of each user. Each point represents a high-risk user. The location of the cluster center JL is obtained by calculating the coordinate concentration of all high-risk users and is marked with specific markers. Based on the psychological behavior of high-risk users who usually stay away from the normal behavior cluster center JL, normal users in a preset area with a radius of R centered on the cluster center JL are distinguished and monitored. The number of times a user enters this area is recorded to obtain the frequency pl of entering and leaving the cluster center.
[0012] Preferably, S6 specifically includes the following steps: S61. The cloud server calculates the distance between a user's data point and its cluster center based on the obtained distance to the cluster center JL and the frequency of the cluster center pl, and obtains the distance YHJ between the user and the cluster center. The distance YHJ between a user and a cluster center is calculated using the following formula; ; In the formula, YHJ(u, JL) j ) represents the distance between user u and the j-th cluster center JL, u i JL represents the value of user u on the i-th feature. j,i表示第j个聚类中心JL在第i个特征上的值,n表示特征的数量。
[0013] Preferably, in step S62, when the user does not exhibit any abnormal behavior during the initial assessment, a secondary assessment mechanism is triggered. The secondary assessment mechanism compares the distance YHJ between the user and the cluster center with a preset area centered at the cluster center JL and with a radius of R. The specific assessment scheme is as follows. When a user's distance YHJ from the cluster center is less than the preset radius R, and the user frequently enters and exits the cluster center JL, a first key attention signal is generated. This first key attention signal is then transmitted to the cloud server via the Internet for in-depth analysis of the user's historical behavior data. When abnormal data and behaviors are detected, an alert signal is generated by the cloud server to notify relevant personnel to pay attention to the user. When a user's distance YHJ from the cluster center is less than the preset radius R, and the user enters or leaves the cluster center JL for the first time, the user is marked and their behavioral characteristic data in this area is recorded for continued monitoring. When a user's distance YHJ from the cluster center is greater than the preset radius R, and the user frequently enters and exits the cluster center JL, a second key attention signal is generated. Monitoring is maintained, and the second key attention signal is transmitted to the cloud server via the Internet. The cloud server then generates an alert signal to notify relevant personnel to pay attention to this user. When the distance YHJ between a user and the cluster center is greater than the preset radius R, and the user does not enter the cluster center JL, no monitoring is required.
[0014] A cloud server includes a real-time user monitoring module, a storage processing module, a cloud computing analytics module, and a communication module. The real-time user monitoring module monitors user behavior data and user transaction data in real time through the cross-border e-commerce client and cloud server, and transmits the collected user behavior data and user transaction data to the storage and processing module in real time via the Internet. The storage processing module is used to receive user behavior data and user transaction data in real time, store user behavior data and transaction data using sharding and partitioning, and preprocess them using SQL scripts. The cloud computing analysis module uses anomaly index calculation and cluster analysis to assess the risk of abnormal user behavior, and analyzes data through PCA dimensionality reduction and visualization tools to determine the user's risk. The communication module is used for communication between the cloud server and the cross-border e-commerce client. Through the HTTPS protocol, the cloud server can securely obtain user behavior data and user transaction data collected by the user client and transmit them to the storage and processing module in real time via the Internet. It is also responsible for sending evaluation results and risk alerts to the monitoring module and analysis module so that corresponding measures can be taken in a timely manner.
[0015] This invention provides a method for information risk analysis in cross-border e-commerce and a cloud server. It has the following beneficial effects: (1) This method achieves comprehensive monitoring of user login, transaction activities, and after-sales behavior through the collaborative work of the user client and the cloud server. This process ensures the real-time nature and accuracy of the data by automatically collecting user behavior data and transaction data, providing rich basic data for subsequent risk analysis. In addition, the use of the HTTPS protocol to ensure the security of data transmission further enhances the reliability of the information system.
[0016] (2) This method employs a variety of advanced algorithms, such as PCA dimensionality reduction and cluster analysis, to help identify high-risk users and their behavioral patterns. In the process of calculating the user login anomaly index DLY and the user order anomaly index DDY, a variety of uncommon parameters are comprehensively considered, which makes the risk assessment more accurate, enables the timely detection of potential fraudulent behavior, and performs comprehensive calculations to obtain the comprehensive abnormal behavior risk coefficient YCF. By setting a preset abnormal behavior risk threshold A, the system can quickly identify suspicious accounts and take corresponding measures, thereby reducing the risk of economic losses and decreased user trust.
[0017] (3) This method, based on distance assessment between users and cluster centers, achieves the goals of dynamic monitoring and precise intervention. When users frequently enter and exit cluster centers, the system can generate key attention signals and conduct in-depth analysis to detect abnormal behavior. This targeted approach ensures that cross-border e-commerce platforms can flexibly respond to various risks, not only optimizing platform security but also improving user experience, laying a solid foundation for the platform's sustainable development. In summary, this method has achieved significant improvements in risk management, user trust, and economic benefits. Attached Figure Description
[0018] Figure 1 This is a schematic diagram illustrating the steps of an information risk analysis method for cross-border e-commerce according to the present invention. Figure 2 This is a block diagram of an information risk analysis device for cross-border e-commerce according to the present invention. Detailed Implementation
[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0020] Example 1 Please see Figure 1 This invention provides an information risk analysis method for cross-border e-commerce. To achieve the above objectives, this invention is implemented through the following technical solution: S1. Real-time monitoring of user behavior data and user transaction data through cross-border e-commerce client and cloud server respectively, automatic collection and recording of relevant user information, and real-time transmission of user behavior data collected by cross-border e-commerce client to cloud server via Internet. S2. The cloud server receives user behavior data collected from the cross-border e-commerce client in real time, stores the user behavior data and user transaction data through sharding and partitioning storage, and then preprocesses the data through the SQL script of the storage cloud server to obtain the transaction behavior dataset. S3. The cloud server performs aggregate calculations based on the acquired transaction behavior dataset to obtain the user login anomaly index DLY and the user order anomaly index DDY. S4. The cloud server summarizes and calculates the user login anomaly index DLY and user order anomaly index DDY to obtain the comprehensive abnormal behavior risk coefficient YCF, and conducts a preliminary comparison and evaluation with the preset abnormal behavior risk threshold A. Based on the evaluation results, the second evaluation mechanism is activated. S5, cloud server uses PCA to reduce the dimensionality of risk data stored in categories, and then uses the visualization tool Excel to create a scatter plot, and plots the coordinate concentration of high-risk users to obtain the location of cluster center JL. Normal users are distinguished and monitored by preset area with cluster center JL as the center point and radius R, and the number of times users enter this area is recorded to obtain the frequency of entering and leaving the cluster center pl. S6. The cloud server calculates the distance YHJ between the user and the cluster center based on the obtained distance to the cluster center JL and the frequency pl of the cluster center. When the user does not have abnormal behavior in the initial assessment, a secondary assessment mechanism is triggered. The secondary assessment mechanism compares the obtained distance YHJ between the user and the cluster center with a preset area with the cluster center JL as the center point and a radius of R, and takes corresponding measures based on the assessment results.
[0021] In this embodiment, through the cooperation of the user client and the cloud server, the system can efficiently collect user behavior data and transaction data, ensuring the real-time nature and integrity of the information. This real-time monitoring significantly improves the timeliness of risk identification. Compared with traditional methods, it can detect abnormal behavior at the first moment, thereby taking timely intervention measures to reduce potential risks. The cloud server stores data through sharding and partitioning, and uses SQL scripts for preprocessing to construct a high-efficiency transaction behavior dataset. This process not only improves the flexibility and access speed of data storage, but also provides a reliable data foundation for subsequent anomaly index calculation. By comprehensively calculating the user login anomaly index DLY and the order anomaly index DDY, the comprehensive abnormal behavior risk coefficient YCF is obtained. The system performs a preliminary comparative assessment with a preset abnormal behavior risk threshold A. This dynamic risk assessment mechanism can promptly detect potential fraudulent activities, making risk control measures more scientific and reasonable. Based on the application of cluster analysis and PCA dimensionality reduction technology, this method can effectively identify high-risk users and their behavioral patterns, further improving the accuracy and efficiency of monitoring. By using visualization tools to draw scatter plots, the distance YHJ between users and cluster centers becomes an important evaluation indicator, enabling the system to respond flexibly in dynamic monitoring. Compared with current technologies, this method not only achieves more refined risk management but also enhances the sensitivity to changes in user behavior, ultimately improving platform security and user experience, and promoting the sustainable development of cross-border e-commerce.
[0022] Example 2 This embodiment is an explanation based on Embodiment 1. Please refer to it. Figure 1 Specifically, S1 includes the following steps: S11. Monitor user transaction activities in real time through the cross-border e-commerce client, record each user's login status, analyze abnormal login frequency, and detect login time zone, login IP address, and login device machine code in real time during login; when users log in and perform transaction operations, monitor order and shopping-related operation data in real time through the cross-border e-commerce client, track address changes in different orders; monitor order amount fluctuations, analyze whether there are abnormally large orders or abnormal price deviations; record the time required for users to add goods to their shopping cart, identify automated machine operation behavior, and obtain user behavior data; S12. The cloud server is integrated and connected to e-commerce websites and cross-border e-commerce clients via the Internet. Then, through tracking code, it tracks the after-sales requests of users in e-commerce websites and cross-border e-commerce clients in real time, tracks the return behavior of each user, obtains the return rate based on the return behavior, and obtains user transaction data by analyzing the return rate. S13. The cloud server connects to the cross-border e-commerce client via the Internet, establishes communication with the cloud server through the HTTPS protocol, and transmits the user behavior data and user transaction data collected by the cross-border e-commerce client to the cloud server in real time via the Internet.
[0023] This embodiment achieves comprehensive monitoring and analysis of user behavior and transaction activities, significantly improving the ability to identify and prevent risks. Real-time monitoring of user login status, transaction behavior, and after-sales feedback enables the system to quickly detect abnormal activities, such as frequent address changes or unusual order amount fluctuations, thereby promptly identifying potential fraud risks. Furthermore, the use of HTTPS protocol ensures the security of data transmission, guaranteeing the privacy and security of user information and further enhancing user trust. This efficient monitoring mechanism not only improves the accuracy of risk management but also lays the foundation for the long-term stable development of the platform, thus achieving a dual optimization of user experience and platform security.
[0024] Example 3 This embodiment is an explanation based on Embodiment 2. Please refer to it. Figure 1 Specifically, S2 includes the following steps: S21. The cloud server receives user behavior data collected by the cross-border e-commerce client in real time, stores the user behavior data and user transaction data monitored by the cloud server in real time according to the storage method of sharding and partitioning, and then accesses the data through the indexing function. S22. The cloud server preprocesses user behavior data and user transaction data using SQL scripts. The preprocessing includes missing value handling, outlier handling, duplicate data handling, standardization, and normalization to obtain a transaction behavior dataset. The transaction behavior dataset includes user login dataset and user order dataset; The user login dataset includes login frequency dl, time zone change sq, IP address change ip, device change sb, and address change count dz; The user order dataset includes order amount deviation je, time required to add items to cart sj, and return rate th.
[0025] In this embodiment, by employing sharding and partitioning storage methods, the cloud server can efficiently manage and quickly access user behavior and transaction data, thereby significantly improving the speed and accuracy of data processing. This structured storage approach not only optimizes data read and write efficiency but also reduces the impact of single-node failures on overall system performance, resulting in stronger system scalability. Simultaneously, data preprocessing ensures high-quality datasets, providing a reliable foundation for subsequent analysis. Ultimately, this comprehensive management strategy makes the analysis of user behavior and transaction data more accurate and real-time, providing strong support for timely identification of potential risks, thereby enhancing the overall security and user trust of the e-commerce platform.
[0026] Example 4 This embodiment is an explanation based on Embodiment 3. Please refer to it. Figure 1 Specifically, S3 includes the following steps: S31. The cloud server performs summary calculations based on the obtained user login dataset to obtain the user login anomaly index DLY. The user login anomaly index DLY is calculated using the following formula; ; In the formula, log represents the logarithmic function, and e represents the exponential function of the time zone variation sq. S32. The cloud server performs summary calculations based on the acquired user order dataset to obtain the user order anomaly index DDY. The user order anomaly index DDY is calculated using the following formula; ; In the formula, e represents the negative exponential function of the time sj required to add the item to the shopping cart, and sin represents the sine function. It represents pi (π).
[0027] In this embodiment, by aggregating and calculating user login datasets and order datasets, the cloud server can generate a user login anomaly index (DLY) and a user order anomaly index (DDY). This process brings several significant benefits. First, the calculation of these anomaly indices utilizes the mathematical properties of logarithms and exponential functions, making risk assessment more flexible and adaptable, and enabling timely identification of potential abnormal behaviors. Second, this data-driven approach reduces errors from human judgment, improves the accuracy and efficiency of risk identification, and ensures that the platform can quickly respond to various security challenges in a dynamic environment. Finally, through accurate anomaly indices, the system not only enhances its risk warning capabilities but also provides a scientific basis for subsequent risk management measures, thereby effectively reducing the risk of potential economic losses and a decline in user trust.
[0028] Example 5 This embodiment is an explanation based on Embodiment 4. Please refer to it. Figure 1 Specifically, S4 includes the following steps: S41. The cloud server calculates the comprehensive abnormal behavior risk coefficient YCF by summarizing and calculating the obtained user login anomaly index DLY and user order anomaly index DDY. The comprehensive abnormal behavior risk coefficient YCF is calculated using the following formula; ; In the formula, w1 and w2 represent the weighting coefficients of the user login anomaly index DLY and the user order anomaly index DDY, respectively, and w1+w2=1. The specific values are set by the user according to the situation.
[0029] S42. Based on the average value of historical user risk data stored in the cloud server, a preset abnormal behavior risk threshold A is set, and a preliminary comparison and evaluation is conducted with the obtained comprehensive abnormal behavior risk coefficient YCF. Corresponding measures are then implemented based on the evaluation results. The specific evaluation scheme is as follows. When the comprehensive abnormal behavior risk coefficient YCF ≥ the preset abnormal behavior risk threshold A, it indicates that the user has abnormal behavior. At this time, the cloud server will verify the user's identity through the payment platform. If the identity verification fails, the user's account will be frozen, the user will be marked as a risk user, and the risk data of the marked risk users will be classified and stored in one category. When the comprehensive abnormal behavior risk coefficient YCF is less than the preset abnormal behavior risk threshold A, it means that the user does not have abnormal behavior. At this time, monitoring will continue and the second assessment mechanism will be triggered.
[0030] In this embodiment, by calculating the comprehensive abnormal behavior risk coefficient YCF and comparing it with a preset abnormal behavior risk threshold A, this method achieves accurate and efficient risk management. First, based on the weighted coefficients of the user login anomaly index DLY and the user order anomaly index DDY, the system can flexibly adapt to the needs of different e-commerce environments, thereby effectively identifying potential abnormal behaviors. This dynamic evaluation mechanism not only captures changes in user behavior in real time, improving the timeliness of risk response, but also ensures rapid identity verification and account protection for suspicious users. When the risk coefficient exceeds the preset abnormal behavior risk threshold A, emergency measures are immediately taken, reducing potential economic losses and security risks. Furthermore, continuous monitoring of users not identified as abnormal, combined with the second evaluation mechanism, further enhances the platform's security and user trust. This systematic and data-driven risk management model significantly improves the operational efficiency and security level of cross-border e-commerce, ensuring user transaction security and the healthy development of the platform.
[0031] Example 6 This embodiment is an explanation based on Embodiment 5. Please refer to it. Figure 1 Specifically, S5 includes the following steps: S51: Based on the risk data stored in the cloud server, the PCA dimensionality reduction technology is used to project the data into a low-dimensional space while retaining the characteristics of the data. S52. After data dimensionality reduction, a scatter plot is created using the visualization tool Excel. Scatter points are drawn on the plot based on the dimensionality reduction coordinates of each user. Each point represents a high-risk user. The location of the cluster center JL is obtained by calculating the coordinate concentration of all high-risk users and is marked with specific markers. Based on the psychological behavior of high-risk users who usually stay away from the normal behavior cluster center JL, normal users in a preset area with a radius of R centered on the cluster center JL are distinguished and monitored. The number of times a user enters this area is recorded to obtain the frequency pl of entering and leaving the cluster center.
[0032] In this embodiment, by employing PCA dimensionality reduction technology, the cloud server can effectively project complex high-dimensional data into a low-dimensional space while retaining important features. This dimensionality reduction process not only reduces data complexity but also improves the efficiency of subsequent analysis. Using visualization tools such as Excel to create scatter plots, user behavior patterns can be intuitively presented, and the location of cluster centers makes risk monitoring more scientific. In particular, monitoring the distance JL between users and cluster centers and the entry / exit frequency pl not only helps identify high-risk users but also promptly detects abnormal behavior. This dynamic monitoring method offers greater flexibility and accuracy compared to traditional methods, thereby improving the overall risk management responsiveness and effectively protecting platform security and user trust.
[0033] Example 7 This embodiment is an explanation based on Embodiment 6. Please refer to it. Figure 1 Specifically, S6 includes the following steps: S61. The cloud server calculates the distance between a user's data point and its cluster center based on the obtained distance to the cluster center JL and the frequency of the cluster center pl, and obtains the distance YHJ between the user and the cluster center. The distance YHJ between a user and a cluster center is calculated using the following formula; ; In the formula, YHJ(u, JL) j ) represents the distance between user u and the j-th cluster center JL, u i JL represents the value of user u on the i-th feature. j,i表示第j个聚类中心JL在第i个特征上的值,n表示特征的数量。
[0034] S62. When the user does not exhibit any abnormal behavior in the initial assessment, a secondary assessment mechanism is triggered. The secondary assessment mechanism compares the distance YHJ between the user and the cluster center with a preset area centered on the cluster center JL and with a radius of R. The specific assessment scheme is as follows. When a user's distance YHJ from the cluster center is less than the preset radius R, and the user frequently enters and exits the cluster center JL, a first key attention signal is generated. This first key attention signal is then transmitted to the cloud server via the Internet for in-depth analysis of the user's historical behavior data. When abnormal data and behaviors are detected, an alert signal is generated by the cloud server to notify relevant personnel to pay attention to the user. When a user's distance YHJ from the cluster center is less than the preset radius R, and the user enters or leaves the cluster center JL for the first time, the user is marked and their behavioral characteristic data in this area is recorded for continued monitoring. When a user's distance YHJ from the cluster center is greater than the preset radius R, and the user frequently enters and exits the cluster center JL, a second key attention signal is generated. Monitoring is maintained, and the second key attention signal is transmitted to the cloud server via the Internet. The cloud server then generates an alert signal to notify relevant personnel to pay attention to this user. When the distance YHJ between a user and the cluster center is greater than the preset radius R, and the user does not enter the cluster center JL, no monitoring is required.
[0035] In this embodiment, a dynamic and hierarchical risk assessment mechanism is established by accurately calculating the distance YHJ between the user and the cluster center, combined with the entry / exit frequency pl. First, it can promptly identify potentially high-risk users, making risk management more targeted and flexible. By generating key attention signals, the system can conduct in-depth analysis of users who frequently enter and exit cluster centers, quickly identifying and responding to abnormal behavior. This real-time feedback mechanism significantly improves response speed and decision-making efficiency. Second, for users entering and exiting for the first time, the system marks them and maintains monitoring, helping to build a long-term analysis framework for user behavior and enhancing the predictive ability for new user risks. Overall, this method not only improves the accuracy and timeliness of risk identification but also achieves optimal resource allocation through refined monitoring strategies, ultimately enhancing the security and user trust of the cross-border e-commerce platform and laying a solid foundation for the platform's sustainable development.
[0036] Example 8 A type of cloud server, please refer to Figure 2 It includes a real-time user monitoring module, a storage processing module, a cloud computing analysis module, and a communication module: The real-time user monitoring module monitors user behavior data and user transaction data in real time through the cross-border e-commerce client and cloud server, and transmits the collected user behavior data and user transaction data to the storage and processing module in real time via the Internet. The storage processing module is used to receive user behavior data and user transaction data in real time, store user behavior data and transaction data using sharding and partitioning, and preprocess them using SQL scripts. The cloud computing analysis module uses anomaly index calculation and cluster analysis to assess the risk of abnormal user behavior, and analyzes data through PCA dimensionality reduction and visualization tools to determine the user's risk. The communication module is used for communication between the cloud server and the cross-border e-commerce client. Through the HTTPS protocol, the cloud server can securely obtain user behavior data and user transaction data collected by the user client and transmit them to the storage and processing module in real time via the Internet. It is also responsible for sending evaluation results and risk alerts to the monitoring module and analysis module so that corresponding measures can be taken in a timely manner.
[0037] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.
Claims
1. A method for information risk analysis in cross-border e-commerce, characterized in that: Includes the following steps: S1. Real-time monitoring of user behavior data and user transaction data through cross-border e-commerce client and cloud server respectively, automatic collection and recording of relevant user information, and real-time transmission of user behavior data collected by cross-border e-commerce client to cloud server via Internet. S2. The cloud server receives user behavior data collected from the cross-border e-commerce client in real time, stores it through sharding and partitioning, and then preprocesses it through the SQL script of the storage cloud server to obtain the transaction behavior dataset. S3. The cloud server performs aggregate calculations based on the acquired transaction behavior dataset to obtain the user login anomaly index DLY and the user order anomaly index DDY. S4. The cloud server summarizes and calculates the user login anomaly index DLY and user order anomaly index DDY to obtain the comprehensive abnormal behavior risk coefficient YCF, and conducts a preliminary comparison and evaluation with the preset abnormal behavior risk threshold A. Based on the evaluation results, the second evaluation mechanism is activated. S5, cloud server uses PCA to reduce the dimensionality of risk data stored in categories, and then uses the visualization tool Excel to create a scatter plot, and plots the coordinate concentration of high-risk users to obtain the location of cluster center JL. Normal users are distinguished and monitored by preset area with cluster center JL as the center point and radius R, and the number of times users enter this area is recorded to obtain the frequency of entering and leaving the cluster center pl. S6. The cloud server calculates the distance YHJ between the user and the cluster center based on the obtained distance to the cluster center JL and the frequency pl of the cluster center. When the user does not have abnormal behavior in the initial assessment, a secondary assessment mechanism is triggered. The secondary assessment mechanism compares the obtained distance YHJ between the user and the cluster center with a preset area with the cluster center JL as the center point and a radius of R, and takes corresponding measures based on the assessment results.
2. The information risk analysis method for cross-border e-commerce according to claim 1, characterized in that: S1 specifically includes the following steps: S11. Monitor user transaction activities in real time through the cross-border e-commerce client, record each user's login status, analyze abnormal login frequency, and detect login time zone, login IP address, and login device machine code in real time during login; when users log in and perform transaction operations, monitor order and shopping-related operation data in real time through the cross-border e-commerce client, track address changes in different orders; monitor order amount fluctuations, analyze whether there are abnormally large orders or abnormal price deviations; record the time required for users to add goods to their shopping cart, identify automated machine operation behavior, and obtain user behavior data; S12. The cloud server is integrated and connected to e-commerce websites and cross-border e-commerce clients via the Internet. Then, through tracking code, it tracks the after-sales requests of users in e-commerce websites and cross-border e-commerce clients in real time, tracks the return behavior of each user, obtains the return rate based on the return behavior, and obtains user transaction data by analyzing the return rate. S13. The cloud server communicates with the cross-border e-commerce client via the Internet HTTPS protocol, and transmits the user behavior data collected by the cross-border e-commerce client to the cloud server in real time via the Internet.
3. The information risk analysis method for cross-border e-commerce according to claim 2, characterized in that: S2 specifically includes the following steps: S21. The cloud server receives user behavior data collected by the cross-border e-commerce client in real time, stores the user behavior data and user transaction data monitored by the cloud server in real time according to the storage method of sharding and partitioning, and then accesses the data through the indexing function. S22. The cloud server preprocesses user behavior data and user transaction data using SQL scripts. The preprocessing includes missing value handling, outlier handling, duplicate data handling, standardization, and normalization to obtain a transaction behavior dataset. The transaction behavior dataset includes user login dataset and user order dataset; The user login dataset includes login frequency dl, time zone change sq, IP address change ip, device change sb, and address change count dz; The user order dataset includes order amount deviation je, time required to add items to cart sj, and return rate th.
4. The information risk analysis method for cross-border e-commerce according to claim 3, characterized in that: S3 specifically includes the following steps: S31. The cloud server performs summary calculations based on the obtained user login dataset to obtain the user login anomaly index DLY. The user login anomaly index DLY is calculated using the following formula; ; In the formula, log represents the logarithmic function, e sq An exponential function representing time zone variation sq; S32. The cloud server performs summary calculations based on the acquired user order dataset to obtain the user order anomaly index DDY. The user order anomaly index DDY is calculated using the following formula; ; In the formula, e -sj The negative exponential function sj represents the time sj required to add an item to the shopping cart, where sin represents the sine function. It represents pi (π).
5. The information risk analysis method for cross-border e-commerce according to claim 4, characterized in that: S4 specifically includes the following steps: S41. The cloud server calculates the comprehensive abnormal behavior risk coefficient YCF by summarizing and calculating the obtained user login anomaly index DLY and user order anomaly index DDY. The comprehensive abnormal behavior risk coefficient YCF is calculated using the following formula; ; In the formula, w1 and w2 represent the weighting coefficients of the user login anomaly index DLY and the user order anomaly index DDY, respectively, and w1+w2=1. The specific values are set by the user according to the situation.
6. The information risk analysis method for cross-border e-commerce according to claim 5, characterized in that: S42. Based on the average value of historical user risk data stored in the cloud server, a preset abnormal behavior risk threshold A is set, and a preliminary comparison and evaluation is conducted with the obtained comprehensive abnormal behavior risk coefficient YCF. Corresponding measures are then implemented based on the evaluation results. The specific evaluation scheme is as follows. When the comprehensive abnormal behavior risk coefficient YCF is greater than or equal to the preset abnormal behavior risk threshold A, it indicates that the user has abnormal behavior. At this time, the cloud server will verify the user's identity through the payment platform. If the identity verification fails, the user's account will be frozen, the user will be marked as a risk user, and the risk data of the marked risk users will be classified and stored in one category. When the comprehensive abnormal behavior risk coefficient YCF is less than the preset abnormal behavior risk threshold A, it means that the user does not have abnormal behavior. At this time, monitoring will continue and the second assessment mechanism will be triggered.
7. The information risk analysis method for cross-border e-commerce according to claim 6, characterized in that: S5 specifically includes the following steps: S51: Based on the risk data stored in the cloud server, the PCA dimensionality reduction technology is used to project the data into a low-dimensional space while retaining the characteristics of the data. S52. After data dimensionality reduction, a scatter plot is created using the visualization tool Excel. Scatter points are drawn on the plot based on the dimensionality reduction coordinates of each user. Each point represents a high-risk user. The location of the cluster center JL is obtained by calculating the coordinate concentration of all high-risk users and is marked with specific markers. Based on the psychological behavior of high-risk users who usually stay away from the normal behavior cluster center JL, normal users in a preset area with a radius of R centered on the cluster center JL are distinguished and monitored. The number of times a user enters this area is recorded to obtain the frequency pl of entering and leaving the cluster center.
8. The information risk analysis method for cross-border e-commerce according to claim 7, characterized in that: S6 specifically includes the following steps: S61. The cloud server calculates the distance between a user's data point and its cluster center based on the obtained distance to the cluster center JL and the frequency of the cluster center pl, and obtains the distance YHJ between the user and the cluster center. The distance YHJ between a user and a cluster center is calculated using the following formula; ; In the formula, YHJ(u, JL) j ) represents the distance between user u and the j-th cluster center JL, u i JL represents the value of user u on the i-th feature. j,i表示第j个聚类中心JL在第i个特征上的值,n表示特征的数量。 9. The information risk analysis method for cross-border e-commerce according to claim 8, characterized in that: S62. When the user does not exhibit any abnormal behavior in the initial assessment, a secondary assessment mechanism is triggered. The secondary assessment mechanism compares the distance YHJ between the user and the cluster center with a preset area centered on the cluster center JL and with a radius of R. The specific assessment scheme is as follows. When a user's distance YHJ from the cluster center is less than the preset radius R, and the user frequently enters and exits the cluster center JL, a first key attention signal is generated. This first key attention signal is then transmitted to the cloud server via the Internet for in-depth analysis of the user's historical behavior data. When abnormal data and behaviors are detected, an alert signal is generated by the cloud server to notify relevant personnel to pay attention to the user. When a user's distance YHJ from the cluster center is less than the preset radius R, and the user enters or leaves the cluster center JL for the first time, the user is marked and their behavioral characteristic data in this area is recorded for continued monitoring. When a user's distance YHJ from the cluster center is greater than the preset radius R, and the user frequently enters and exits the cluster center JL, a second key attention signal is generated. Monitoring is maintained, and the second key attention signal is transmitted to the cloud server via the Internet. The cloud server then generates an alert signal to notify relevant personnel to pay attention to this user. When the distance YHJ between a user and the cluster center is greater than the preset radius R, and the user does not enter the cluster center JL, no monitoring is required.
10. A cloud server, characterized in that, The information risk analysis method for cross-border e-commerce as described in any one of claims 1-9 includes a real-time user monitoring module, a storage processing module, a cloud computing analysis module, and a communication module. The real-time user monitoring module monitors user behavior data and user transaction data in real time through the cross-border e-commerce client and cloud server, and transmits the collected user behavior data and user transaction data to the storage and processing module in real time via the Internet. The storage processing module is used to receive user behavior data and user transaction data in real time, store user behavior data and transaction data using sharding and partitioning, and preprocess them using SQL scripts. The cloud computing analysis module uses anomaly index calculation and cluster analysis to assess the risk of abnormal user behavior, and analyzes data through PCA dimensionality reduction and visualization tools to determine the user's risk. The communication module is used for communication between the cloud server and the cross-border e-commerce client. Through the HTTPS protocol, the cloud server can securely obtain user behavior data and user transaction data collected by the user client and transmit them to the storage and processing module in real time via the Internet. It is also responsible for sending assessment results and risk alerts to the monitoring and analysis modules so that appropriate measures can be taken in a timely manner.
