A real-time risk monitoring and anti-fraud control method for game accessory transactions

Abstract

The application discloses a kind of real-time risk monitoring and anti-fraud control method of game ornament transaction, belong to the technical field of game data supervision;Including the following steps: step one: user data acquisition, step two: user data is integrated calculation, obtains transaction risk and transaction risk threshold;Step three: training identification model;Step four: according to transaction risk threshold and identification model, judge whether transaction can be allowed;Step five: according to the result of identification model, select to block transaction or continue transaction.The application uses the above method, through user data acquisition as foundation, carries out user risk calculation to the two sides of transaction, clearly high-risk transaction, combined with early warning mechanism, risk warning is carried out to the two sides of high-risk transaction, and secondary identification is carried out through identification model, when judging as fraud transaction, transaction behavior is prevented, realize real-time risk monitoring and anti-fraud control to transaction, reduce the probability of successful fraud.

Description

Technical Field

[0001] The present invention relates to the technical field of game item trading, and in particular to a method for real-time risk monitoring and anti-fraud control. Background Technology

[0002] Against the backdrop of the booming development of the current gaming industry, the trading of game items has developed into a large-scale and highly liquid secondary market. Players can freely trade items in the market. However, free trading can lead to fraud. Existing regulatory measures often rely on post-event recovery to address fraud, without corresponding risk detection and anti-fraud solutions. As a result, the post-event recovery period is often long and the resolution efficiency is low. Summary of the Invention

[0003] The purpose of this invention is to provide a real-time risk monitoring and anti-fraud control method for game item transactions. It establishes a two-level management system. The first level of management uses mathematical statistics and weight calculation methods to detect both parties in the transaction. The detection cost is low and it can provide real-time judgment basis for both parties. When the transaction risk is high, the second level of management intervenes through an identification model to determine whether the transaction is fraudulent. If the identification model determines that the transaction is fraudulent, the transaction is blocked.

[0004] To achieve the above objectives, this invention provides a method for real-time risk monitoring and anti-fraud control in game item trading, comprising the following steps: Step 1: Collect user data, including account transaction data, account activity, and account login information; Step 2: Perform integrated calculations on user data to calculate the transaction risk for each user and define the transaction risk threshold; Step 3: Collect historical fraud cases and normal transaction cases as training datasets to train the recognition model; Step 4: Based on the transaction risk threshold, classify transactions into low-risk, medium-risk, and high-risk transactions. In low-risk transactions, both parties are allowed to continue the transaction. In medium-risk transactions, the party with the lower transaction risk is alerted that there is a risk of deception, but both parties are allowed to continue the transaction. In high-risk transactions, the transaction is suspended, and Step 5 is executed to use the identification model for further identification. Step 5: Based on the results of the identification model, choose to block the transaction or continue the transaction.

[0005] Preferably, in step one, the specific content of the user data includes the following: Account transaction data: Records the difference in amount and item type for all transactions made by the account; Account activity: Records the login IP and game duration for each account login; Account Login: Records user registration duration, real-name authentication status, recharge amount, and account value distribution; Preferably, in step two, the process of calculating transaction risk is as follows: Transaction risks The formula is as follows: ; In the above formula, , and Indicates the weighting coefficient. This indicates a risk associated with account transaction data. Indicates dynamic account risk. This indicates account login risk; the buyer's transaction risk is calculated separately. Transaction risks with the seller Processed The trading risk thresholds are defined as follows: Low trading risk thresholds are set separately. , medium transaction risk threshold and high transaction risk threshold ,pass The magnitude of the transaction risk threshold is used to determine the risk of the transaction.

[0006] Preferably, the risk calculation for the account transaction data is as follows: The calculation of account transaction data risk includes: Amount Difference: the difference between the recorded transaction amount and the current average market price; Item Type: based on the price of the traded item, a significance coefficient is defined. The formula for account transaction data risk is as follows: ; In the above formula, For the transaction price, Indicates the average market price. The importance coefficient is indicated by the following formula: ; In the above formula, These are warning parameters that users can set themselves.

[0007] Preferably, the dynamic risk calculation for the account is as follows: Account dynamic risks include the login IP address for each login, obtaining the login location each time through the login IP address, and defining login parameters. Game duration is the ratio of recorded game operation time to total game time. Account dynamic risk is calculated as follows: ; ; In the above formula, This indicates the number of times the login IP address appears. This indicates the total game time. This indicates the time spent playing the game.

[0008] Preferably, the account login risk is calculated as follows: Account login risks include user registration duration and defined duration parameters. The formula for calculating account login risk is as follows: Real-name parameter b is obtained based on real-name status, recharge parameter c is obtained based on recharge amount, and the value ratio of currency and goods is calculated based on the corresponding account value distribution. ; ; ; In the above formula, This indicates the total amount of currency currently in the account. This indicates the total amount of currency corresponding to the current items in the account.

[0009] Preferably, in step three, the specific process of training the recognition model is as follows: The recognition model uses the XGBoost model. For the training dataset, there are a total of There are 3 samples, each with a corresponding label. For the i-th fraud case, a label is set. For the first A typical case: Setting tags A feature vector is set for each transaction sample. Including transaction behavior characteristics Account characteristics Login behavior characteristics Each feature vector For training and recognition of the XGBoost model, the model training parameters are set as follows: learning rate, number of decision trees, maximum depth, regularization parameter, and subsample ratio. The loss function is defined as follows: ; In the above formula, This represents the probability that the identification model identifies the i-th sample as fraudulent; the model is optimized using gradient descent until the performance of the identification model meets the requirements.

[0010] Preferably, the feature vector The content is as follows: Transaction behavior feature vector Account feature vector Login behavior feature vector .

[0011] Therefore, the present invention employs the above-mentioned method for real-time risk monitoring and anti-fraud control in game item transactions, which has the following advantages: In this invention, a two-tiered management system is established through mathematical statistics and identification models. Based on user data collection, user risk is calculated for both parties in a transaction to identify high-risk transactions. Combined with an early warning mechanism, risk alerts are given to both parties in high-risk transactions. Furthermore, a secondary identification is performed through the identification model. When a transaction is determined to be fraudulent, the transaction is blocked, thereby achieving real-time risk monitoring and anti-fraud control of transactions and reducing the probability of successful fraud.

[0012] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0013] Figure 1 This is a flowchart illustrating a real-time risk monitoring and anti-fraud control method for game item transactions according to the present invention. Detailed Implementation

[0014] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Specific model specifications need to be selected and determined according to the actual specifications of the device, etc. The specific selection calculation method adopts existing technology in the art, and therefore will not be described in detail.

[0015] Example like Figure 1 As shown, this invention provides a method for real-time risk monitoring and anti-fraud control in game item trading, comprising the following steps: Step 1: Collect user data, including account transaction data, account activity, and account login information; The specific content of user data includes the following: Account transaction data: Records the difference in amount and item type for all transactions made by the account; Account activity: Records the login IP and game duration for each account login; Account Login: Records user registration duration, real-name authentication status, recharge amount, and account value distribution; The data from the aforementioned users will be used for subsequent integrated calculations and identification of the recognition model; Step 2: Perform integrated calculations on user data to calculate the transaction risk for each user and define the transaction risk threshold; In step two, the process of calculating transaction risk is as follows: Transaction risks The formula is as follows: ; In the above formula, , and Indicates the weighting coefficient. This indicates a risk associated with account transaction data. Indicates dynamic account risk. This indicates account login risk; the buyer's transaction risk is calculated separately. Transaction risks with the seller Processed The processing method here is set according to the user's needs, and includes, but is not limited to, summation and subtraction. In this embodiment, The formula is as follows: ; Define the trading risk thresholds as follows: Set low trading risk thresholds respectively. , medium transaction risk threshold and high transaction risk threshold The specific value of the transaction risk threshold can be set by the user based on existing data. The magnitude of the transaction risk threshold is used to determine the risk of the transaction.

[0016] Account transaction data risk is calculated as follows: The calculation includes: Amount Difference: the difference between the recorded transaction amount and the current average market price; Item Type: based on the price of the traded item, a significance coefficient is defined. The formula for calculating account transaction data risk is as follows: ; In the above formula, For the transaction price, Indicates the average market price. The importance coefficient is indicated by the following formula: ; In the above formula, These are warning parameters that users can set themselves.

[0017] Account dynamic risk is calculated as follows: Account dynamic risk includes the login IP address for each login, which is used to obtain the login location for each login, and login parameters are defined. Game duration is the ratio of recorded game operation time to total game time. Account dynamic risk is calculated as follows: ; ; In the above formula, This indicates the number of times the login IP address appears. This indicates the total game time. This indicates the time spent playing the game.

[0018] Account login risk is calculated as follows: Account login risk includes the user's registration duration, defined by the duration parameter. The formula for calculating account login risk is as follows: Real-name parameter b is obtained based on real-name status, recharge parameter c is obtained based on recharge amount, and the value ratio of currency and goods is calculated based on the corresponding account value distribution. ; ; ; In the above formula, This indicates the total amount of currency currently in the account. This indicates the total amount of currency corresponding to the current items in the account.

[0019] Step 3: Collect historical fraud cases and normal transaction cases as training datasets to train the identification model; the specific process of training the identification model is as follows: The recognition model uses the XGBoost model. For the training dataset, there are a total of There are 3 samples, each with a corresponding label. For the i-th fraud case, a label is set. For the first A typical case: Setting tags A feature vector is set for each transaction sample. Including transaction behavior characteristics Account characteristics Login behavior characteristics Each feature vector For training and recognition of the XGBoost model, the model training parameters are set as follows: learning rate, number of decision trees, maximum depth, regularization parameter, and subsample ratio. The loss function is defined as follows: ; In the above formula, This represents the probability that the identification model identifies the i-th sample as fraudulent; the model is optimized using gradient descent until the performance of the identification model meets the requirements.

[0020] Feature vector The content is as follows: Transaction behavior feature vector Account feature vector Login behavior feature vector .

[0021] Step 4: Based on the transaction risk threshold, the transactions are divided into low-risk, medium-risk, and high-risk transactions. In low-risk transactions, both parties are allowed to continue the transaction; in medium-risk transactions, the party with the lower transaction risk is warned of the risk of fraud, and both parties are allowed to continue the transaction; in high-risk transactions, the transaction is suspended, and an identification model is introduced for identification. Step 5: Based on the results of the identification model, choose to block the transaction or continue the transaction.

[0022] Therefore, this invention employs a real-time risk monitoring and anti-fraud control method for game item transactions. It establishes a two-tiered management system using mathematical statistics and an identification model. The first tier uses low-cost mathematical statistics as the primary management level. When the first tier determines a high risk, the identification model intervenes to effectively identify fraudulent transactions. Based on user data collection, user risk calculations are performed on both parties in the transaction to identify high-risk transactions. Combined with an early warning mechanism, risk alerts are issued to both parties in high-risk transactions. Furthermore, a second identification process is performed using the identification model. When a transaction is determined to be fraudulent, the transaction is blocked, achieving real-time risk monitoring and anti-fraud control, and reducing the probability of successful fraud.

[0023] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for real-time risk monitoring and anti-fraud control in game item trading, characterized in that: Includes the following steps: Step 1: Collect user data, including account transaction data, account activity, and account login information; ‌ Step 2: Perform integrated calculations on user data to calculate the transaction risk for each user and define the transaction risk threshold; Step 3: Collect historical fraud cases and normal transaction cases as training datasets to train the recognition model; Step 4: Based on the transaction risk threshold, classify transactions into low-risk, medium-risk, and high-risk transactions. In low-risk transactions, both parties are allowed to continue the transaction. In medium-risk transactions, the party with the lower transaction risk is alerted that there is a risk of deception, but both parties are allowed to continue the transaction. In high-risk transactions, the transaction is suspended, and Step 5 is executed to use the identification model for further identification. Step 5: Based on the results of the identification model, choose to block the transaction or continue the transaction.

2. The method for real-time risk monitoring and anti-fraud control of game item transactions according to claim 1, characterized in that: In step one, the specific content of the user data includes the following: Account transaction data: Records the difference in amount and item type for all transactions made by the account; Account activity: Records the login IP and game duration for each account login; Account Login: Records user registration duration, real-name authentication status, recharge amount, and account value distribution.

3. The method for real-time risk monitoring and anti-fraud control of game item transactions according to claim 2, characterized in that: In step two, the process of calculating transaction risk is as follows: Transaction risks The formula is as follows: ； In the above formula, , and Indicates the weighting coefficient. This indicates a risk associated with account transaction data. Indicates dynamic account risk. This indicates account login risk; the buyer's transaction risk is calculated separately. Transaction risks with the seller Processed The trading risk thresholds are defined as follows: Low trading risk thresholds are set separately. , medium transaction risk threshold and high transaction risk threshold ,pass The magnitude of the transaction risk threshold is used to determine the risk of the transaction.

4. The method for real-time risk monitoring and anti-fraud control of game item transactions according to claim 3, characterized in that: The risk calculation for the account transaction data is as follows: The calculation of account transaction data risk includes: Amount Difference: the difference between the recorded transaction amount and the current average market price; Item Type: based on the price of the traded item, a significance coefficient is defined. The formula for account transaction data risk is as follows: ； In the above formula, For the transaction price, Indicates the average market price. The importance coefficient is indicated by the following formula: ； In the above formula, These are warning parameters that users can set themselves.

5. The method for real-time risk monitoring and anti-fraud control of game item transactions according to claim 3, characterized in that: The dynamic risk calculation for the account is as follows: Account dynamic risks include the login IP address for each login, obtaining the login location each time through the login IP address, and defining login parameters. ; Game duration is the ratio of recorded game action time to total game time. Account dynamic risk is calculated as follows: ； ； In the above formula, This indicates the number of times the login IP address appears. This indicates the total game time. This indicates the time spent playing the game.

6. The method for real-time risk monitoring and anti-fraud control of game item transactions according to claim 3, characterized in that: The account login risk is calculated as follows: Account login risks include user registration duration and the definition of duration parameters. The formula for calculating account login risk is as follows: Real-name parameter b is obtained based on real-name status, recharge parameter c is obtained based on recharge amount, and the value ratio of currency and goods is calculated based on the corresponding account value distribution. ； ； ； In the above formula, This indicates the total amount of currency currently in the account. This indicates the total amount of currency corresponding to the current items in the account.

7. The method for real-time risk monitoring and anti-fraud control of game item transactions according to claim 1, characterized in that: In step three, the specific process of training the recognition model is as follows: The recognition model uses the XGBoost model. For the training dataset, there are a total of There are 3 samples, each with a corresponding label. For the i-th fraud case, a label is set. For the first A typical case: Setting tags A feature vector is set for each transaction sample. Including transaction behavior characteristics Account characteristics Login behavior characteristics Each feature vector For training and recognition of the XGBoost model, the model training parameters are set as follows: learning rate, number of decision trees, maximum depth, regularization parameter, and subsample ratio. The loss function is defined as follows: ； In the above formula, This represents the probability that the identification model identifies the i-th sample as fraudulent; the model is optimized using gradient descent until the performance of the identification model meets the requirements.

8. The method for real-time risk monitoring and anti-fraud control of game item transactions according to claim 7, characterized in that: The feature vector The content is as follows: Transaction behavior feature vector ; Account feature vector Login behavior feature vector .

Images