Digital RMB anti-fraud monitoring method, apparatus and device, and medium

Abstract

The invention discloses a digital RMB anti-fraud monitoring method, device and equipment and a medium, and relates to the technical field of digital RMB, and the method comprises the steps: collecting transaction information generated in a transaction process of using the digital RMB in real time; performing transaction feature analysis processing on a current user portrait formed by the transaction information through a preset digital RMB anti-fraud monitoring model to predict a corresponding current fraud behavior index; the preset digital RMB anti-fraud monitoring model is a model obtained after the initial digital RMB anti-fraud monitoring model learns associated information between portrait feature information of historical user portraits and a corresponding relationship between the portrait feature information and a corresponding fraud behavior tag; and judging whether the fraudulent behavior index meets an abnormal transaction mode setting condition or not, and if so, executing a corresponding alarm process and a user verification process. The adverse effects that the wallet of the user is locked and funds are frozen due to misjudgment are reduced, and the user experience is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of digital RMB, and particularly relates to a method, device, equipment and medium for anti-fraud monitoring of digital RMB. Background Art

[0002] When the system determines that a customer's digital RMB transaction is abnormal, the wallet is directly locked, resulting in the customer's wallet being unavailable and in a locked state, and the customer's funds being frozen, making it impossible to perform any operations. Currently, it is necessary to fill out a due diligence questionnaire and apply to the head office for verification and unlocking by email. Among them, the current system's conditions for determining abnormal digital RMB transactions are very strict, and the abnormal transaction conditions are as follows:

[0003] 1. When the user uses digital RMB for transactions during the early morning period, such as from 1 am to 5 am;

[0004] 2. When the geographical location of the user's digital RMB transaction is an infrequently used address, for example, the user opens a digital RMB wallet in City A, but uses it in City B (a non-digital RMB wallet opening area);

[0005] 3. There are frequent small-amount transfer operations in the user's digital RMB wallet;

[0006] 4. The transaction log captured by the system for the Yonghui digital RMB wallet is an abnormal log.

[0007] Once the transaction shows the above situations, the system will directly lock the wallet and freeze it, resulting in the customer being unable to use it normally.

[0008] In summary, how to accurately and real-time monitor anti-fraud for digital RMB transactions is a technical problem to be solved in this field. Summary of the Invention

[0009] In view of this, the purpose of the present invention is to provide a method, device, equipment and medium for anti-fraud monitoring of digital RMB to accurately and real-time monitor anti-fraud for digital RMB transactions. The specific solutions are as follows:

[0010] In the first aspect, the present application discloses a method for anti-fraud monitoring of digital RMB, including:

[0011] Real-time collect transaction information generated during the process of using digital RMB for transactions;

[0012] Performing transaction feature analysis on the current user profile composed of the transaction information through a preset digital RMB anti-fraud monitoring model to predict the corresponding current fraud behavior index; the preset digital RMB anti-fraud monitoring model is obtained after the initial digital RMB anti-fraud monitoring model learns the association information between the portrait feature information of the historical user profiles and the corresponding relationship between the portrait feature information and the fraud behavior labels.

[0013] Judging whether the fraud behavior index meets the abnormal transaction mode setting conditions, and if so, executing the corresponding alarm process and user verification process.

[0014] Optionally, the digital RMB anti-fraud monitoring method further includes:

[0015] Obtaining transaction data, user credit information, and user basic information generated by the user's digital RMB transactions to obtain training data, and using the training data to construct the corresponding historical user profiles.

[0016] Optionally, the obtaining of the transaction data, user credit information, and user basic information generated by the user's digital RMB transactions includes:

[0017] Obtaining the transaction time, transaction amount, transaction frequency, and user consumption behavior generated by the user's digital RMB transactions to obtain the transaction data generated by the user's digital RMB transactions;

[0018] Obtaining the number of credit cards, current arrears, and credit preservation query times of the user to obtain the user credit information;

[0019] Obtaining the age information, educational background information, occupation information, and geographical information of the user's permanent residence to obtain the user basic information.

[0020] Optionally, the constructing of the corresponding historical user profiles using the training data includes:

[0021] Using the mapping relationship between the user basic information in the training data, the user consumption behavior in the transaction data, and the corresponding user consumption patterns to construct a first mapping relationship;

[0022] Using the mapping relationship between the user basic information in the training data, the user credit information, and the corresponding user credit levels to construct a second mapping relationship;

[0023] Constructing the historical user profiles generated by the user's digital RMB according to the first mapping relationship and the second mapping relationship.

[0024] Optionally, the digital RMB anti-fraud monitoring method further includes:

[0025] Construct an initial digital RMB anti-fraud monitoring model including a multi-head attention mechanism layer, a feed-forward layer, a normalization layer, and a skip layer;

[0026] Input the historical user portraits and the corresponding fraud behavior labels into the initial digital RMB anti-fraud monitoring model, so that the multi-head attention mechanism layer of the initial digital RMB anti-fraud monitoring model extracts the importance weights of different feature dimensions in the historical user portraits regarding the anti-fraud task, analyzes the correlation between the portrait feature information to obtain the corresponding correlation information, then inputs the correlation information into the normalization layer for normalization processing, transfers the normalization result to the feed-forward layer, and transfers the importance weights to the feed-forward layer through the skip layer, so that the feed-forward layer learns the correlation between the portrait feature information and the corresponding relationship between the portrait feature information and the fraud behavior labels to obtain a preset digital RMB anti-fraud monitoring model for predicting the corresponding fraud behavior index.

[0027] Optionally, the inputting the historical user portraits and the corresponding fraud behavior labels into the initial digital RMB anti-fraud monitoring model includes:

[0028] Perform word embedding processing, position embedding processing, and type embedding processing on each portrait text information of the historical user portraits to obtain the text vectors after conversion of each portrait text information;

[0029] Perform low-dimensional mapping on each of the text vectors to obtain a target vector after low-dimensional numericalization, so as to obtain an input feature vector for inputting into the initial digital RMB anti-fraud monitoring model.

[0030] Optionally, the judging whether the fraud behavior index meets the abnormal transaction mode setting condition, if so, performing the corresponding alarm process and user verification process, includes:

[0031] Judge whether the fraud behavior index is greater than a preset fraud index threshold, if so, trigger the step of sending a warning to the regulatory department and the corresponding bank, freeze the current digital RMB account, and then notify the user to perform the corresponding user verification process.

[0032] In a second aspect, the present application discloses a digital RMB anti-fraud monitoring device, including:

[0033] An information collection module, configured to collect transaction information generated during the process of using digital RMB for transactions in real time;

[0034] An exponential prediction module, configured to perform transaction feature analysis and processing on the current user profile composed of the transaction information through the preset digital RMB anti-fraud monitoring model, so as to predict the corresponding current fraud behavior index; the preset digital RMB anti-fraud monitoring model is obtained after the initial digital RMB anti-fraud monitoring model learns the association information between the portrait feature information of historical user profiles and the corresponding relationship between the portrait feature information and the corresponding fraud behavior labels;

[0035] A process execution module, configured to determine whether the fraud behavior index meets the abnormal transaction mode setting condition, and if so, execute the corresponding alarm process and user verification process.

[0036] In a third aspect, the present application discloses an electronic device, including:

[0037] A memory, configured to store a computer program;

[0038] A processor, configured to execute the computer program to implement the steps of the digital RMB anti-fraud monitoring method disclosed above.

[0039] In a fourth aspect, the present application discloses a computer-readable storage medium, configured to store a computer program; wherein, when the computer program is executed by a processor, the steps of the digital RMB anti-fraud monitoring method disclosed above are implemented.

[0040] It can be seen that the present application discloses a method for anti-fraud monitoring of digital RMB, including: collecting in real time the transaction information generated during the process of using digital RMB for transactions; performing transaction feature analysis and processing on the current user profile composed of the transaction information through a preset digital RMB anti-fraud monitoring model to predict the corresponding current fraud behavior index; the preset digital RMB anti-fraud monitoring model is obtained after the initial digital RMB anti-fraud monitoring model learns the correlation information between the portrait feature information of historical user profiles and the corresponding relationship between the portrait feature information and the corresponding fraud behavior labels; determining whether the fraud behavior index meets the set conditions of the abnormal transaction mode, and if so, executing the corresponding alarm process and user verification process. Thus, by extracting and correlating the importance weights of different feature dimensions of the user profile through the digital RMB anti-fraud monitoring model, potential fraud patterns can be captured more comprehensively and accurately, improving the accuracy of anti-fraud. The model can learn and optimize based on a large amount of historical data, better adapting to different types of fraud behaviors and transaction scenarios, and having strong flexibility. It can collect transaction information in real time, analyze and process it, and discover potential fraud behaviors in a timely manner, which is more efficient than discovering them afterwards, and the accuracy of judgment can be improved through model analysis. By analyzing and processing the current user profile composed of the transaction information collected in real time through the preset digital RMB anti-fraud monitoring model, directly predicting the corresponding current fraud behavior index, instead of directly taking extreme measures such as locking the wallet, but first predicting the fraud behavior index and performing the corresponding process, giving a certain buffer and verification space, reducing the excessive interference with the normal use of users. In the prior art, the strict abnormal transaction determination conditions may lead to more misjudgments, while this solution can reduce the adverse effects such as the user's wallet being locked and funds being frozen caused by misjudgments, improving the user experience. The set conditions of the abnormal transaction mode and the subsequent process can be flexibly adjusted according to the actual situation, better adapting to different scenarios and requirements, rather than rigidly following a single strict rule. Through hierarchical management based on the fraud behavior index, different intensities of measures can be taken more targeted, such as adopting a more gentle verification process for lower-risk situations, rather than locking uniformly. It reduces the inconvenience and dissatisfaction brought to users by improper locking, is conducive to maintaining the good relationship between the bank and users, and enhancing the user's trust in the digital RMB system. Description of the Drawings

[0041] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only the embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on the provided drawings.

[0042] Figure 1Flowchart of a digital RMB anti-fraud monitoring method disclosed in this application;

[0043] Figure 2 Flowchart of obtaining the first mapping relationship disclosed in this application;

[0044] Figure 3 Flowchart of obtaining the second mapping relationship disclosed in this application;

[0045] Figure 4 Flowchart of a method for using a preset digital RMB anti-fraud monitoring model disclosed in this application;

[0046] Figure 5 Schematic diagram of the structure of a digital RMB anti-fraud monitoring device disclosed in this application;

[0047] Figure 6 Structural diagram of an electronic device disclosed in this application. Detailed implementation manners

[0048] Next, the technical solutions in the embodiments of this application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of this application. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0049] When the system determines that a customer's digital RMB transaction is abnormal, the wallet is directly locked, resulting in the customer's wallet being unavailable and in a locked state, and the customer's funds are frozen and no operations can be performed. Currently, it is necessary to fill out a due diligence questionnaire and apply to the head office for verification and unlocking by email. Among them, the conditions for the current system to determine abnormal digital RMB transactions are very strict, and the abnormal transaction conditions are as follows:

[0050] 1. When the user uses digital RMB for transactions during the early morning period, such as from 1 o'clock to 5 o'clock;

[0051] 2. When the geographical location of the user's digital RMB transaction is an infrequently used address, for example, the user opens a digital RMB wallet in City A, but the user uses it in City B (a non-digital RMB wallet opening area);

[0052] 3. Frequent small-amount transfer operations occur in the user's digital RMB wallet;

[0053] 4. The transaction log captured by the system for the Yonghui digital RMB wallet is an abnormal log.

[0054] Once the transaction shows the above situations, the system will directly lock the wallet and freeze it, resulting in the customer being unable to use it normally.

[0055] To this end, the present invention provides a digital RMB anti-fraud monitoring solution, which can accurately and real-time monitor digital RMB transactions for anti-fraud.

[0056] Refer to Figure 1 As shown, an embodiment of the present invention discloses a digital RMB anti-fraud monitoring method, including:

[0057] Step S11: Real-time collect transaction information generated during the process of using digital RMB transactions.

[0058] In this embodiment, high-speed data processing technology (such as Apache Kafka or Apache Flink, which provides a unified, high-throughput, and low-latency platform for processing real-time data) is used to collect and process transaction data streams in real-time to ensure data real-time. Among them, the transaction information specifically includes, but is not limited to: obtaining any one or several of the transaction time, transaction amount, transaction frequency, user consumption behavior generated by the user using digital RMB transactions, the number of the user's credit cards, the current outstanding amount, the number of credit preservation inquiries, the user's age information, educational background information, occupation information, and geographical information of the user's permanent residence. Specifically, the steps of using Apache Kafka to collect and process transaction data streams in real-time are as follows:

[0059] The transaction system sends the generated transaction data as messages to the Kafka topic, and the Kafka cluster is responsible for storing and distributing the above messages. The processing application reads the messages from the Kafka topic as a consumer for further processing.

[0060] The steps of using the combination of Apache Flink and Kafka to collect and process transaction data streams in real-time are as follows:

[0061] Create a Flink task: Write a Flink program to define the data processing logic.

[0062] Connect to Kafka: Establish a connection with the Kafka topic through the Kafka connector of Flink.

[0063] Data processing flow: Implement various operations on the transaction data read from Kafka in the Flink task, such as filtering, transformation, aggregation, etc.

[0064] Result output: Output the processed results to a suitable destination, such as a database or other storage systems.

[0065] For example, in the processing of transaction data, the total transaction amount per minute can be calculated in real-time through Flink, or real-time screening and early warning can be performed on specific types of transactions. Assuming that the transaction data includes fields such as transaction amount and transaction type, the Flink program can implement the following logic:

[0066] DataStream<Tuple2<String, Double>> transactions = env

[0067] .addSource(new FlinkKafkaConsumer<>("transactionTopic"));

[0068] DataStream<Tuple2<String, Double>> totalAmountPerMinute =transactions

[0069] .keyBy(tuple -> tuple.f0)

[0070] .timeWindow(Time.minutes(1))

[0071] .sum(1);

[0072] In this way, the total amount of each transaction type per minute can be calculated in real time. This is just a simple example. In actual applications, more complex processing logics will be designed according to specific requirements, and no specific limitations are made here.

[0073] Step S22: Perform transaction feature analysis and processing on the current user portrait composed of the transaction information through a preset digital RMB anti-fraud monitoring model to predict the corresponding current fraud behavior index; the preset digital RMB anti-fraud monitoring model is a model obtained after the initial digital RMB anti-fraud monitoring model learns the association information between the portrait feature information of historical user portraits and the corresponding relationship between the portrait feature information and the corresponding fraud behavior labels.

[0074] In this embodiment, before performing transaction feature analysis and processing on the current user profile composed of the transaction information through a preset digital RMB anti-fraud monitoring model, the following steps are also included: obtaining transaction data, user credit information, and user basic information generated by the user's digital RMB transactions to obtain training data, and using the training data to construct a corresponding historical user profile. Specifically, obtaining the transaction time, transaction amount, transaction frequency, and user consumption behavior generated by the user's digital RMB transactions to obtain the transaction data generated by the user's digital RMB transactions; obtaining the number of credit cards of the user, the current outstanding amount, and the number of credit preservation inquiries to obtain the user credit information; obtaining the age information, educational background information, occupation information, and geographical information of the user's permanent residence to obtain the user basic information. It can be understood that obtaining the transaction data, user credit information, and user basic information generated by the user's digital RMB transactions from the group asset library, data lake, data warehouse, and transaction table to obtain the training data. It should be noted that the transaction data specifically includes, but is not limited to: the transaction time, transaction amount, transaction frequency, and user consumption behavior generated by the user's digital RMB transactions; the user credit information specifically includes, but is not limited to: the number of credit cards of the user, the current outstanding amount, and the number of credit preservation inquiries; the user basic information includes, but is not limited to: the age information, educational background information, occupation information, and geographical information of the user's permanent residence.

[0075] In this embodiment, a first mapping relationship is constructed by using the mapping relationship between the user basic information in the training data and the user consumption behavior in the transaction data and the corresponding user consumption pattern; it can be understood that, as Figure 2 shown, predicting the user consumption pattern based on the current user basic information and user consumption behavior, that is, constructing a first mapping relationship between the user basic information, user consumption behavior, and user consumption pattern. Specifically, the collected corresponding training data is respectively subjected to token embedding, position embedding, and type embedding processing, that is, vector representation of data such as user basic information, usage habits, and consumption behavior. Specifically:

[0076] token embedding: Convert data such as the user's basic information, usage habits, and consumption behavior into vector form for easy processing and analysis by the model.

[0077] position embedding: Consider the position information of the data, such as transaction time, transaction order, etc., and encode the position information into a vector so that the model can capture the sequential and time features in the data.

[0078] type embedding: Distinguish and encode different types of data. For example, distinguish credit information such as the number of credit cards and the amount of debt from basic information such as age and education level, so that the model can better understand and process different types of data.

[0079] Through the above-mentioned processing, the original training data can be converted into a vector representation suitable for model input, and information such as semantics, position, and type in the data can be captured. Then, through the embedding layer, the discrete vector representation obtained after the above-mentioned token embedding, position embedding, and type embedding processing is further mapped to a relatively low-dimensional continuous vector space. Through this mapping, features that may originally have a large number of different categories or states are represented as vectors of a fixed dimension, achieving a compression and dimensionality reduction process for the original discrete vector representation, obtaining the target vector representation. The target vector representation that retains semantic information is input into the initial user consumption pattern prediction model, so as to train the initial user consumption pattern prediction model using the target vector and the corresponding user consumption pattern label, and obtain the target user consumption pattern prediction model for predicting the corresponding user consumption pattern. Among them, the consumption pattern label includes: normal consumption pattern and abnormal consumption pattern; the initial user consumption pattern prediction model includes: multi-head attention mechanism layer, skip layer, normalization layer, and feed-forward layer. Specifically, the multi-head attention mechanism layer: The multi-head attention mechanism layer of the model processes the input target vector, extracts the importance weights of different feature dimensions, and analyzes the correlation between feature information to obtain the corresponding correlation information. The normalization layer: Input the correlation information into the normalization layer for normalization processing. The feed-forward layer: The normalization result and the importance weight transmitted by the skip layer are input into the feed-forward layer together. The feed-forward layer learns the correlation between feature information and the correspondence between feature information and the consumption pattern label. Use the training data to train the entire model, and continuously adjust the parameters of the model through the backpropagation algorithm to improve the prediction accuracy of the model. In this way, a first mapping relationship between user basic information, user consumption behavior, and user consumption pattern can be constructed.

[0080] In this embodiment, use the mapping relationship between the user basic information, the user credit information in the training data and the corresponding user credit rating to construct a second mapping relationship; it can be understood that, such as Figure 3As shown above, in the process of constructing the first mapping relationship, the user basic information and user credit information in the training data are represented in text vector form, and then they are processed twice by the embedding layer to obtain the target vector representation after dimensionality reduction. The target vector representation after dimensionality reduction and the corresponding user credit rating label are input into the initial user credit rating prediction model to train the initial user credit rating prediction model, so as to obtain the target user credit rating prediction model for predicting the corresponding user credit rating. Among them, each layer of the initial user credit rating prediction model is exactly the same as each layer of the initial user consumption pattern prediction model, and the training process thereof will not be elaborated herein. In this way, a second mapping relationship is constructed by using the mapping relationship between the user basic information, user credit information and the corresponding user credit rating. Among them, the user credit rating can be divided into ten levels, and the higher the credit rating, the better the user's credit.

[0081] In this embodiment, a historical user portrait generated by the user's use of digital RMB is constructed according to the first mapping relationship and the second mapping relationship. It can be understood that in the transaction process of the user using digital RMB constructed according to the above first mapping relationship and second mapping relationship, a historical user portrait of the user is comprehensively constructed by using information such as user basic information, usage habits, consumption behaviors, psychological states, credit ratings, geographical distributions, etc. and the corresponding user consumption patterns and user credit ratings. So as to train the initial digital RMB anti-fraud monitoring model by using the historical user portrait.

[0082] In this embodiment, word embedding processing, position embedding processing and type embedding processing are respectively performed on each portrait text information of the historical user portrait to obtain the text vectors after conversion of each portrait text information; the text vectors are subjected to low-dimensional mapping to obtain the target vectors after low-dimensional numericalization, so as to obtain the input feature vectors for inputting into the initial digital RMB anti-fraud monitoring model. It can be understood that performing token embedding, position embedding, and type embedding processing on the portrait text information respectively, that is, the vector representation of each user information. Specifically: Token embedding: Convert data such as the user's basic information, usage habits, and consumption behaviors into vector form for easy processing and analysis by the model. Position embedding: Considering the position information of the data, such as transaction time, transaction order, etc., the position information is also encoded as a vector so that the model can capture the sequential and time features in the data. Type embedding: Distinguish and encode different types of data, such as separating credit information such as the number of credit cards and the amount of arrears from basic information such as age and education level, so that the model can better understand and process different types of data.

[0083] In this embodiment, as Figure 4As shown, the initial digital RMB anti-fraud monitoring model can be: Convolutional Neural Network (CNN), Recurrent Neural Network (RNN), or Long Short-Term Memory network (LSTM), which is used to detect patterns and anomalies of fraud behaviors. The present invention proposes a model for detecting fraud behaviors based on Transformer. Specifically, an initial digital RMB anti-fraud monitoring model including a multi-head attention mechanism layer, a feed-forward layer, a normalization layer, and a skip layer is constructed; the historical user portraits and the corresponding fraud behavior labels are input into the initial digital RMB anti-fraud monitoring model, so that the multi-head attention mechanism layer of the initial digital RMB anti-fraud monitoring model extracts the importance weights of different feature dimensions in the historical user portraits regarding the anti-fraud task, and analyzes the relevance between the portrait feature information to obtain the corresponding correlation information, and then inputs the correlation information into the normalization layer for normalization processing, transfers the normalization result to the feed-forward layer, and transfers the importance weights to the feed-forward layer through the skip layer, so that the feed-forward layer learns the correlation between the portrait feature information and the corresponding relationship between the portrait feature information and the fraud behavior labels to obtain a preset digital RMB anti-fraud monitoring model for predicting the corresponding fraud behavior index. It can be understood that the multi-head attention mechanism layer of the model processes the input vectors, extracts the importance weights of different feature dimensions, and analyzes the relevance between the portrait feature information to obtain the corresponding correlation information. Normalization layer: Input the correlation information into the normalization layer for normalization processing. Feed-forward layer: The normalization result and the importance weights transferred by the skip layer are input into the feed-forward layer together, and the feed-forward layer learns the correlation between the portrait feature information and the corresponding relationship between the portrait feature information and the fraud behavior labels. The skip layer refers to a structure in the neural network that spans certain layers or connections and is used to achieve the skip transmission of information. The role of the skip layer is to introduce a more flexible information flow mode in the neural network, enabling information to skip some intermediate layers and be directly transmitted to the subsequent layers. This skip connection can help alleviate the problem of gradient disappearance, promote the propagation and sharing of information, and thus improve the performance and expressive ability of the network. Common skip layer structures include Residual Connection and Dense Connection, etc. The residual connection directly adds the input to the output of the subsequent layer, enabling information to be directly transmitted and avoiding the problem of gradient disappearance. The dense connection connects the outputs of all previous layers to the subsequent layers, achieving a wider range of information sharing. Model training: Use the training data to train the entire model, and continuously adjust the parameters of the model through the backpropagation algorithm to improve the prediction accuracy of the model.Model evaluation: Use the test data to evaluate the trained model. The evaluation metrics can include accuracy, recall rate, F1 value, etc. Model adjustment: According to the results of model evaluation, adjust and optimize the model, such as adjusting the model structure, hyperparameters, etc. Model application: Apply the trained model to the actual digital RMB anti-fraud monitoring, collect user transaction information in real time, and conduct prediction and analysis. It should be noted that the target digital RMB anti-fraud monitoring model automatically updates and optimizes the model by continuously learning new transaction data and fraud patterns.

[0084] Step S13: Determine whether the fraud behavior index meets the abnormal transaction mode setting conditions. If so, execute the corresponding alarm process and user verification process.

[0085] In this embodiment, determine whether the fraud behavior index is greater than the preset fraud index threshold. If so, trigger the step of sending a warning to the regulatory department and the corresponding bank, freeze the current digital RMB account, and then notify the user to execute the corresponding user verification process. It can be understood that, according to the comparison and judgment of the size relationship between the fraud behavior index output by the digital RMB anti-fraud monitoring model and the preset fraud index threshold, this fraud behavior index is predicted by comprehensively using all user data of the user regarding the digital RMB wallet for data characteristics. Therefore, when it is compared that the fraud behavior index is greater than the preset fraud index threshold, it means that the user's digital RMB transaction mode is identified as an abnormal transaction mode. For example: unusual transaction frequency, abnormal amount pattern, uncommon geographical location jump, etc. Then trigger the step of sending a warning to the regulatory department and the corresponding bank, send an alarm to the regulatory department and the relevant bank in real time, freeze the current digital RMB account, and then notify the user to execute the corresponding user verification process. Specifically, provide a detailed abnormal transaction report, including the account information involved, transaction time, transaction amount, and the reason for being marked as a suspicious account. If it involves individual users, the system should be able to send real-time warnings to users via text messages, emails, or in-app notifications. Users can respond to the warning by verifying the transaction or reporting unauthorized activities.

[0086] It should be noted that after training is completed and during the process of model application, set a dynamic threshold. When certain characteristics of transaction activities (transaction data, user credit information, user basic information) exceed the set dynamic threshold, an alarm is triggered. The dynamic threshold is dynamically adjusted based on historical data and the results of continuous learning.

[0087] It can be seen that the present application discloses a digital RMB anti-fraud monitoring method, including: collecting in real time the transaction information generated during the process of using digital RMB for transactions; analyzing and processing the transaction characteristics of the current user portrait composed of the transaction information through a preset digital RMB anti-fraud monitoring model to predict the corresponding current fraud behavior index; the preset digital RMB anti-fraud monitoring model is obtained after the initial digital RMB anti-fraud monitoring model learns the association information between the portrait feature information of historical user portraits and the corresponding relationship between the portrait feature information and the corresponding fraud behavior labels; determining whether the fraud behavior index meets the set conditions of the abnormal transaction mode, and if so, executing the corresponding alarm process and user verification process. Thus, by extracting and associating the importance weights of different feature dimensions of the user portrait through the digital RMB anti-fraud monitoring model, potential fraud patterns can be captured more comprehensively and accurately, improving the accuracy of anti-fraud. The model can be learned and optimized based on a large amount of historical data, better adapting to different types of fraud behaviors and transaction scenarios, and having strong flexibility. It can collect transaction information in real time and perform analysis and processing to timely detect potential fraud behaviors, which is more efficient than detecting them afterwards, and the accuracy of judgment can be improved through model analysis. By analyzing and processing the current user portrait composed of the transaction information collected in real time through the preset digital RMB anti-fraud monitoring model, directly predicting the corresponding current fraud behavior index, instead of directly taking extreme measures such as locking the wallet, but first predicting the fraud behavior index and performing the corresponding process, giving a certain buffer and verification space, reducing the excessive interference with the normal use of users. In the prior art, the strict abnormal transaction determination conditions may lead to more misjudgments, while this solution can reduce the adverse effects such as the user's wallet being locked and funds being frozen caused by misjudgments, improving the user experience. The set conditions of the abnormal transaction mode and the subsequent process can be flexibly adjusted according to the actual situation, better adapting to different scenarios and requirements, rather than rigidly following a single strict rule. Through hierarchical management based on the fraud behavior index, different intensity measures can be taken more targeted, such as adopting a relatively mild verification process for lower-risk situations, rather than locking uniformly. It reduces the inconvenience and dissatisfaction brought to users by improper locking, is conducive to maintaining the good relationship between the bank and users, and enhancing the user's trust in the digital RMB system.

[0088] Referring to Figure 5 as shown, the present invention also correspondingly discloses a digital RMB anti-fraud monitoring device, including:

[0089] An information collection module 11, configured to collect in real time the transaction information generated during the process of using digital RMB for transactions;

[0090] An exponential prediction module 12 is configured to perform transaction feature analysis and processing on the current user profile composed of the transaction information through the preset digital RMB anti-fraud monitoring model to predict the corresponding current fraud behavior index; the preset digital RMB anti-fraud monitoring model is a model obtained after the initial digital RMB anti-fraud monitoring model learns the association information between the portrait feature information of the historical user profiles and the corresponding relationship between the portrait feature information and the corresponding fraud behavior labels;

[0091] A process execution module 13 is configured to determine whether the fraud behavior index meets the abnormal transaction mode setting conditions, and if so, execute the corresponding alarm process and user verification process.

[0092] It can be seen that the present application discloses real-time collection of transaction information generated during the process of using digital RMB for transactions; performing transaction feature analysis and processing on the current user portrait composed of the transaction information through a preset digital RMB anti-fraud monitoring model to predict the corresponding current fraud behavior index; the preset digital RMB anti-fraud monitoring model is a model obtained after the initial digital RMB anti-fraud monitoring model learns the correlation information between the portrait feature information of historical user portraits and the corresponding relationship between the portrait feature information and the fraud behavior labels; determining whether the fraud behavior index meets the abnormal transaction mode setting conditions, and if so, executing the corresponding alarm process and user verification process. Thus, through the extraction and correlation analysis of the importance weights of different feature dimensions of the user portrait by the digital RMB anti-fraud monitoring model, potential fraud patterns can be captured more comprehensively and accurately, improving the accuracy of anti-fraud. The model can learn and optimize based on a large amount of historical data, better adapting to different types of fraud behaviors and transaction scenarios, and having strong flexibility. It can collect transaction information in real time and perform analysis and processing to promptly discover potential fraud behaviors, which is more efficient than discovering them afterwards, and the accuracy of judgment can be improved through model analysis. By analyzing and processing the current user portrait composed of the real-time collected transaction information through the preset digital RMB anti-fraud monitoring model, directly predicting the corresponding current fraud behavior index, instead of directly taking extreme measures such as locking the wallet, but first predicting the fraud behavior index and performing the corresponding process, giving a certain buffer and verification space, reducing the excessive interference with the normal use of users. In the prior art, the strict abnormal transaction determination conditions may lead to more misjudgments, while this solution can reduce the adverse effects such as the user's wallet being locked and funds being frozen due to misjudgments, improving the user experience. The abnormal transaction mode setting conditions and subsequent processes can be flexibly adjusted according to the actual situation, better adapting to different scenarios and requirements, rather than rigidly following a single strict rule. Through hierarchical management based on the fraud behavior index, different intensities of measures can be taken more pertinently, such as adopting a relatively mild verification process for lower-risk situations instead of locking uniformly. It reduces the inconvenience and dissatisfaction brought to users by improper locking, is conducive to maintaining a good relationship between the bank and users, and enhances the user's trust in the digital RMB system.

[0093] Further, the embodiment of the present application also discloses an electronic device, Figure 6 It is the structure diagram of the electronic device 20 shown according to an exemplary embodiment, and the content in the figure cannot be considered as any limitation on the scope of use of the present application.

[0094] Figure 6Schematic diagram of the structure of an electronic device 20 provided by an embodiment of the present application. The electronic device 20 may specifically include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26. Among them, the memory 22 is used to store a computer program, and the computer program is loaded and executed by the processor 21 to implement the relevant steps in the digital RMB anti-fraud monitoring method disclosed in any of the foregoing embodiments. In addition, the electronic device 20 in this embodiment may specifically be an electronic computer.

[0095] In this embodiment, the power supply 23 is used to provide working voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and external devices, and the communication protocol it follows is any communication protocol applicable to the technical solution of the present application, and no specific limitation is imposed on it here; the input / output interface 25 is used to obtain external input data or output data to the outside, and its specific interface type can be selected according to specific application needs, and no specific limitation is made here.

[0096] Among them, the processor 21 may include one or more processing cores, such as a 4-core processor, an 8-core processor, etc. The processor 21 may be implemented in at least one of the following hardware forms: DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), and PLA (Programmable Logic Array). The processor 21 may also include a main processor and a coprocessor. The main processor is a processor used to process data in the wake state, also known as the CPU (Central Processing Unit); the coprocessor is a low-power processor used to process data in the standby state. In some embodiments, the processor 21 may be integrated with a GPU (Graphics Processing Unit), and the GPU is responsible for rendering and drawing the content to be displayed on the display screen. In some embodiments, the processor 21 may also include an AI (Artificial Intelligence) processor, which is used to process computing operations related to machine learning.

[0097] In addition, the memory 22, as a carrier for resource storage, may be a read-only memory, a random access memory, a disk, or an optical disc, etc. The resources stored thereon may include an operating system 221, a computer program 222, etc., and the storage method may be short-term storage or permanent storage.

[0098] Among them, the operating system 221 is used to manage and control each hardware device on the electronic device 20 and the computer program 222, so as to implement the operation and processing of the massive data 223 in the memory 22 by the processor 21. It can be Windows Server, Netware, Unix, Linux, etc. In addition to the computer program that can be used to complete the digital RMB anti-fraud monitoring method executed by the electronic device 20 disclosed in any of the foregoing embodiments, the computer program 222 can further include computer programs that can be used to complete other specific tasks. In addition to the data that can include the data transmitted by the external device received by the electronic device, the data 223 can also include the data collected by its own input / output interface 25, etc.

[0099] Furthermore, the present application also discloses a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, it implements the digital RMB anti-fraud monitoring method disclosed above. For the specific steps of this method, reference can be made to the corresponding content disclosed in the foregoing embodiments, and details will not be repeated here.

[0100] In this specification, the various embodiments are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. The same or similar parts among the various embodiments can be referred to each other. For the devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the description of the method part.

[0101] Those skilled in the art may further realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described according to functions in the above description. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered as exceeding the scope of this application. The steps of the methods or algorithms described in combination with the embodiments disclosed herein can be directly implemented by hardware, software modules executed by a processor, or a combination of both. The software modules can be placed in a random access memory (RAM), internal memory, read-only memory (ROM), electrically programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), registers, hard disks, removable disks, CD-ROM (Compact Disc-Read Only Memory), or any other form of storage medium known in the technical field.

[0102] Finally, it should also be noted that in this document, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including", or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article, or device comprising a series of elements not only includes those elements but also includes other elements not expressly listed, or elements inherent to such process, method, article, or device. Without further limitation, an element defined by the statement "comprising an..." does not exclude the existence of additional identical elements in the process, method, article, or device comprising the element.

[0103] The above has introduced the solution provided by the present invention in detail. Specific examples are used herein to elaborate on the principles and implementation manners of the present invention. The description of the above embodiments is only used to help understand the method and its core idea of the present invention; at the same time, for those of ordinary skill in the art, according to the idea of the present invention, there will be changes in the specific implementation manners and application scopes. In summary, the content of this specification should not be construed as a limitation to the present invention.

Claims

1. A digital RMB anti-fraud monitoring method, characterized in that: include: Collect transaction information generated during the use of digital RMB transactions in real time; The current user portrait composed of the transaction information is analyzed and processed by a preset digital RMB anti-fraud monitoring model for transaction characteristics to predict the corresponding current fraud behavior index; the preset digital RMB anti-fraud monitoring model is a model obtained after the initial digital RMB anti-fraud monitoring model learns the correlation information between the portrait feature information of the historical user portrait and the correspondence between the portrait feature information and the corresponding fraud behavior label; Determine whether the fraud behavior index meets the abnormal transaction mode setting conditions. If so, execute the corresponding alarm process and user verification process.

2. The digital RMB anti-fraud monitoring method according to claim 1 is characterized in that: Also includes: Obtain transaction data, user credit information, and user basic information generated by users using digital RMB transactions to obtain training data, and use the training data to construct corresponding historical user portraits.

3. The digital RMB anti-fraud monitoring method according to claim 2 is characterized in that: The acquisition of transaction data, user credit information, and user basic information generated by users using digital RMB transactions includes: Obtain the transaction time, transaction amount, transaction frequency, and user consumption behavior generated by the user's use of digital RMB transactions to obtain the transaction data generated by the user's use of digital RMB transactions; Obtain the user's credit card number, current outstanding balance, and credit save query times to obtain the user's credit information; Obtain the user's age information, education information, occupation information and geographic information of the user's permanent residence to obtain the user's basic information.

4. The digital RMB anti-fraud monitoring method according to claim 2 is characterized in that: The using the training data to construct the corresponding historical user portrait includes: Constructing a first mapping relationship by using the mapping relationship between the user basic information in the training data, the user consumption behavior in the transaction data and the corresponding user consumption pattern; Constructing a second mapping relationship by using the mapping relationship between the user basic information, the user credit information and the corresponding user credit level in the training data; A historical user profile generated by the user's use of digital RMB is constructed based on the first mapping relationship and the second mapping relationship.

5. The digital RMB anti-fraud monitoring method according to any one of claims 1 to 4 is characterized in that: Also includes: Construct an initial digital RMB anti-fraud monitoring model that includes a multi-head attention mechanism layer, a feedforward layer, a normalization layer, and a skip layer; The historical user portraits and the corresponding fraud behavior labels are input into the initial digital RMB anti-fraud monitoring model, so that the multi-head attention mechanism layer of the initial digital RMB anti-fraud monitoring model extracts the importance weights of different feature dimensions in the historical user portraits regarding the anti-fraud task, and analyzes the correlation between the portrait feature information to obtain the corresponding correlation information, and then inputs the correlation information into the normalization layer for normalization, passes the normalization result to the feedforward layer, and passes the importance weight to the feedforward layer through the jump layer, so that the feedforward layer learns the correlation between the portrait feature information and the correspondence between the portrait feature information and the fraud behavior label, so as to obtain a preset digital RMB anti-fraud monitoring model for predicting the corresponding fraud behavior index.

6. The digital RMB anti-fraud monitoring method according to claim 5 is characterized in that: The inputting of historical user portraits and corresponding fraud behavior labels into the initial digital RMB anti-fraud monitoring model includes: Performing word embedding processing, position embedding processing, and type embedding processing on each portrait text information of the historical user portrait, so as to obtain a text vector after the portrait text information is converted; Each of the text vectors is low-dimensionally mapped to obtain a low-dimensional digitized target vector to obtain an input feature vector for inputting the initial digital RMB anti-fraud monitoring model.

7. The digital RMB anti-fraud monitoring method according to claim 1 is characterized in that: The step of judging whether the fraudulent behavior index meets the abnormal transaction mode setting conditions, and if so, executing the corresponding alarm process and user verification process includes: Determine whether the fraud behavior index is greater than the preset fraud index threshold. If so, it triggers the step of sending a warning to the regulatory authorities and the corresponding bank, freezes the current digital RMB account, and then notifies the user to perform the corresponding user verification process.

8. A digital RMB anti-fraud monitoring device, characterized in that: include: An information collection module, used to collect transaction information generated during the transaction process using digital RMB in real time; An index prediction module is used to perform transaction feature analysis on the current user portrait composed of the transaction information through a preset digital RMB anti-fraud monitoring model to predict the corresponding current fraud behavior index; the preset digital RMB anti-fraud monitoring model is a model obtained after the initial digital RMB anti-fraud monitoring model learns the correlation information between the portrait feature information of the historical user portrait and the correspondence between the portrait feature information and the corresponding fraud behavior label; The process execution module is used to determine whether the fraud behavior index meets the setting conditions of the abnormal transaction mode. If so, the corresponding alarm process and user verification process are executed.

9. An electronic device, characterized in that: include: Memory, used to store computer programs; A processor for executing the computer program to implement the steps of the digital RMB anti-fraud monitoring method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: Used to store computer programs; wherein, when the computer program is executed by the processor, the steps of the digital RMB anti-fraud monitoring method as described in any one of claims 1 to 7 are implemented.

Images