Granted Patent B2

US 12,725,158 · App. 17/358,575 · Granted Sep 1, 2026

Inventors: Shi Cao (Austin, TX); Chiranjeet Chetia (Round Rock, TX); Liang Wang (San Jose, CA); Junpeng Wang (San Jose, CA); Morvarid Jamalian (Saratoga, CA)

Assignee: Visa International Service Association

G06Q20/4016G06N20/00G06Q30/0205

Abstract

Claims (107)

1 . A computer-implemented method for detecting collusive transaction fraud, comprising:

identifying, with at least one processor of a fraud detection system, a merchant category code for a first merchant system operating in an electronic payment processing network to process electronic payment transactions, the merchant category code provided by the first merchant system;

storing, with the at least one processor of the fraud detection system, the merchant category code in association with the first merchant system in a database of the electronic payment processing network;

extracting, with at least one processor of a data extractor of the fraud detection system and from the database of the electronic payment processing network storing data associated with electronic payment transactions processed by the electronic payment processing network, transaction data baseline data for the merchant category code and time series baseline data for the merchant category code, the transaction data baseline data comprising at least one of the following: distribution of dollar amount per transaction, dollar amount per payment device used, dollar amount per cross-border transaction, fraud rate, decline rate for the merchant category code, or any combination thereof, and the time series baseline data comprising at least one of the following: an average transaction count, a median transaction count, an average transaction spend, a median transaction spend, or any combination thereof, over a time period for the merchant category code;

generating, with at least one processor of a baseline generator of the fraud detection system, a merchant baseline for the merchant category code, the merchant baseline comprising: (i) a transaction data baseline generated from the transaction data baseline data for the merchant category code; and (ii) a time series baseline generated from the time series baseline data for the merchant category code;

extracting, with the at least one processor of the data extractor of the fraud detection system and from the database of the electronic payment processing network, time series data for the first merchant system, the time series data for the first merchant system comprising at least one of the following: an average transaction count, a median transaction count, an average transaction spend, a median transaction spend, or any combination thereof over a time period for the first merchant system;

extracting, with the at least one processor of the data extractor of the fraud detection system and from the database of the electronic payment processing network, transaction data for the first merchant system, the transaction data for the first merchant system comprising at least one of the following: a distribution of dollar amount per transaction, dollar amount per payment device used, dollar amount per cross-border transaction, fraud rate, decline rate for the first merchant system, or any combination thereof;

inputting, with at least one processor of the fraud detection system, the transaction data for the first merchant system and the transaction data baseline to a deep learning model of the fraud detection system;

generating, with the deep learning model, a first score based on comparing the transaction data for the first merchant system to the transaction data baseline;

inputting, with at least one processor of the fraud detection system, the time series data for the first merchant system and the time series baseline to the deep learning model of the fraud detection system;

generating, with the deep learning model, a second score based on comparing the time series data for the first merchant system to the time series baseline, wherein generating the first score and/or the second score comprises the deep learning model:

compressing the transaction data and the transaction data baseline and/or compressing the time series data and the time series baseline by generating at least one feature vector for each;

comparing the at least one feature vector for the transaction data and the at least one feature vector for the transaction data baseline and/or comparing the at least one feature vector for the time series data and the at least one feature vector for the time series baseline; and

generating the first score and/or the second score based on the comparison;

generating, with the at least one processor of the fraud detection system, a first merchant risk score associated with the first merchant system based on the first score and the second score;

based on the first merchant risk score, determining, with the at least one processor of the fraud detection system, that the merchant category code for the first merchant is miscoded;

determining, with the at least one processor of the fraud detection system, a plurality of related entities related to the first merchant system based on the data associated with electronic payment transactions processed by the electronic payment processing network stored in the database of the electronic payment processing network, the plurality of related entities comprising at least one of the following: an acquirer system associated with the first merchant system, a second merchant system associated with the acquirer system associated with the first merchant system, a payment device which initiated a payment transaction with the first merchant system, or any combination thereof;

classifying, with the at least one processor of the fraud detection system, the first merchant system and at least one related entity of the plurality of related entities in a first group risk class based on at least one risk score generated by the fraud detection system for the at least one related entity, the first group risk class based on a determination that the first merchant system and the at least one related entity are engaged in collusive transaction fraud;

determining, with the at least one processor of the fraud detection system, a corrected merchant category code for the first merchant system by comparing the transaction data for the first merchant system to transaction data baselines of a plurality of merchant category codes and/or comparing the time series data for the first merchant system to time series baselines of a plurality of merchant category codes;

automatically modifying, with the at least one processor of the fraud detection system, the merchant category code for the first merchant system by modifying the merchant category code stored in the database of the electronic payment processing network to the corrected merchant category code;

after modifying the merchant category code for the first merchant system, receiving, with a transaction processing system in the electronic payment processing network, a payment transaction request associated with the first merchant system and the at least one related entity; and

processing, with the transaction processing system, the payment transaction request using the corrected merchant category code based on the corrected category code stored in the database of the electronic payment processing network.

2 . The computer-implemented method of claim 1 , further comprising:

in response to classifying the first merchant system and the at least one related entity of the plurality of related entities in the first group risk class, automatically initiating, with the at least one processor, an investigation protocol.

3 . The computer-implemented method of claim 1 , further comprising:

determining, with the at least one processor, a region code associated with the first merchant system; and

detecting illicit transaction activity of the first merchant system based on the corrected merchant category code and the region code.

4 . The computer-implemented method of claim 1 , further comprising:

identifying, with the at least one processor, at least one payment transaction conducted between the first merchant system and the at least one related entity of the plurality of related entities in the first group risk class;

determining, with the at least one processor, that the at least one payment transaction is a cross-border payment transaction associated with a first region and a second region;

generating, with at least one processor, a transaction risk score associated with the at least one payment transaction based on at least one of the following: a status of the at least one payment transaction as a cross-border payment transaction, the first region, the second region, and the merchant category code;

determining, with the at least one processor, that the transaction risk score satisfies a threshold; and

classifying, with the at least one processor, the at least one payment transaction in a first transaction risk class in response to determining that the transaction risk score satisfies the threshold.

5 . A system for detecting collusive transaction fraud, comprising at least one processor programmed or configured to:

identify a merchant category code for a first merchant system operating in an electronic payment processing network to process electronic payment transactions, the merchant category code provided by the first merchant system;

store the merchant category code in association with the first merchant system in a database of the electronic payment processing network;

extract, with a data extractor of a fraud detection system and from the database of the electronic payment processing network storing data associated with electronic payment transactions processed by the electronic payment processing network, transaction data baseline data for the merchant category code and time series baseline data for the merchant category code, the transaction data baseline data comprising at least one of the following: distribution of dollar amount per transaction, dollar amount per payment device used, dollar amount per cross-border transaction, fraud rate, decline rate for the merchant category code, or any combination thereof, and the time series baseline data comprising at least one of the following: an average transaction count, a median transaction count, an average transaction spend, a median transaction spend, or any combination thereof, over a time period for the merchant category code;

generate, with a baseline generator of the fraud detection system, a merchant baseline for the merchant category code, the merchant baseline comprising: (i) a transaction data baseline generated from the transaction data baseline data for the merchant category code; and (ii) a time series baseline generated from the time series baseline data for the merchant category code;

extract, with the data extractor of the fraud detection system and from the database of the electronic payment processing network, time series data for the first merchant system, the time series data for the first merchant system comprising at least one of the following: an average transaction count, a median transaction count, an average transaction spend, a median transaction spend, or any combination thereof over a time period for the first merchant system;

extract, with the data extractor of the fraud detection system and from the database of the electronic payment processing network, transaction data for the first merchant system, the transaction data for the first merchant system comprising at least one of the following: a distribution of dollar amount per transaction, dollar amount per payment device used, dollar amount per cross-border transaction, fraud rate, decline rate for the first merchant system, or any combination thereof;

input the transaction data for the first merchant system and the transaction data baseline to a deep learning model of the fraud detection system;

generate, with the deep learning model, a first score based on comparing the transaction data for the first merchant system to the transaction data baseline;

input the time series data for the first merchant system and the time series baseline to the deep learning model of the fraud detection system;

generate, with the deep learning model, a second score based on comparing the time series data for the first merchant system to the time series baseline, wherein generating the first score and/or the second score comprises the deep learning model:

compressing the transaction data and the transaction data baseline and/or compressing the time series data and the time series baseline by generating at least one feature vector for each;

comparing the at least one feature vector for the transaction data and the at least one feature vector for the transaction data baseline and/or comparing the at least one feature vector for the time series data and the at least one feature vector for the time series baseline; and

generating the first score and/or the second score based on the comparison;

generate a first merchant risk score associated with the first merchant system based on the first score and the second score;

based on the first merchant risk score, determine that the merchant category code for the first merchant is miscoded;

determine a plurality of related entities related to the first merchant system based on the data associated with electronic payment transactions processed by the electronic payment processing network stored in the database of the electronic payment processing network, the plurality of related entities comprise at least one of the following: an acquirer system associated with the first merchant system, a second merchant system associated with the acquirer system associated with the first merchant system, a payment device which initiated a payment transaction with the first merchant system, or any combination thereof;

classify the first merchant system and at least one related entity of the plurality of related entities in a first group risk class based on at least one risk score generated by the fraud detection system for the at least one related entity, the first group risk class based on a determination that the first merchant system and the at least one related entity are engaged in collusive transaction fraud;

determine a corrected merchant category code for the first merchant system by comparing the transaction data for the first merchant system to transaction data baselines of a plurality of merchant category codes and/or comparing the time series data for the first merchant system to time series baselines of a plurality of merchant category codes;

automatically modify the merchant category code for the first merchant system by modifying the merchant category code stored in the database of the electronic payment processing network to the corrected merchant category code;

after modifying the merchant category code for the first merchant system, receive, with a transaction processing system in the electronic payment processing network, a payment transaction request associated with the first merchant system and the at least one related entity; and

process, with the transaction processing system, the payment transaction request using the corrected merchant category code based on the corrected category code stored in the database of the electronic payment processing network.

6 . The system of claim 5 , wherein the at least one processor is further programmed or configured to:

in response to classifying the first merchant system and the at least one related entity of the plurality of related entities in the first group risk class, automatically initiate an investigation protocol.

7 . The system of claim 5 , wherein the at least one processor is further programmed or configured to:

determine a region code associated with the first merchant system; and

detect illicit transaction activity of the first merchant system based on the corrected merchant category code and the region code.

8 . The system of claim 5 , wherein the at least one processor is further programmed or configured to:

identify at least one payment transaction conducted between the first merchant system and the at least one related entity of the plurality of related entities in the first group risk class;

determine that the at least one payment transaction is a cross-border payment transaction associated with a first region and a second region;

generate a transaction risk score associated with the at least one payment transaction based on at least one of the following: a status of the at least one payment transaction as a cross-border payment transaction, the first region, the second region, and the merchant category code;

determine that the transaction risk score satisfies a threshold; and

classify the at least one payment transaction in a first transaction risk class in response to determining that the transaction risk score satisfies the threshold.

9 . A computer program product for detecting collusive transaction fraud, the computer program product comprising at least one non-transitory computer-readable medium including one or more instructions that, when executed by at least one processor, cause the at least one processor to:

identify a merchant category code for a first merchant system operating in an electronic payment processing network to process electronic payment transactions, the merchant category code provided by the first merchant system;

store the merchant category code in association with the first merchant system in a database of the electronic payment processing network;

extract, with a data extractor of a fraud detection system and from the database of the electronic payment processing network storing data associated with electronic payment transactions processed by the electronic payment processing network, transaction data baseline data for the merchant category code and time series baseline data for the merchant category code, the transaction data baseline data comprising at least one of the following: distribution of dollar amount per transaction, dollar amount per payment device used, dollar amount per cross-border transaction, fraud rate, decline rate for the merchant category code, or any combination thereof, and the time series baseline data comprising at least one of the following: an average transaction count, a median transaction count, an average transaction spend, a median transaction spend, or any combination thereof, over a time period for the merchant category code;

generate, with a baseline generator of the fraud detection system, a merchant baseline for the merchant category code, the merchant baseline comprising: (i) a transaction data baseline generated from the transaction data baseline data for the merchant category code; and (ii) a time series baseline generated from the time series baseline data for the merchant category code;

extract, with the data extractor of the fraud detection system and from the database of the electronic payment processing network, time series data for the first merchant system, the time series data associated with the first merchant system comprising at least one of the following: an average transaction count, a median transaction count, an average transaction spend, a median transaction spend, or any combination thereof over a time period for the first merchant system;

extract, with the data extractor of the fraud detection system and from the database of the electronic payment processing network, transaction data for the first merchant system, the transaction data for the first merchant system comprising at least one of the following: a distribution of dollar amount per transaction, dollar amount per payment device used, dollar amount per cross-border transaction, fraud rate, decline rate for the first merchant system, or any combination thereof;

input the transaction data for the first merchant system and the transaction data baseline to a deep learning model of the fraud detection system;

generate, with the deep learning model, a first score based on comparing the transaction data for the first merchant system to the transaction data baseline;

input the time series data for the first merchant system and the time series baseline to the deep learning model of the fraud detection system;

generate, with the deep learning model, a second score based on comparing the time series data for the first merchant system to the time series baseline, wherein generating the first score and/or the second score comprises the deep learning model:

compressing the transaction data and the transaction data baseline and/or compressing the time series data and the time series baseline by generating at least one feature vector for each;

comparing the at least one feature vector for the transaction data and the at least one feature vector for the transaction data baseline and/or comparing the at least one feature vector for the time series data and the at least one feature vector for the time series baseline; and

generating the first score and/or the second score based on the comparison;

generate a first merchant risk score associated with the first merchant system based on the first score and the second score;

based on the first merchant risk score, determine that the merchant category code for the first merchant is miscoded;

determine a plurality of related entities related to the first merchant system based on the data associated with electronic payment transactions processed by the electronic payment processing network stored in the database of the electronic payment processing network, the plurality of related entities comprise at least one of the following: an acquirer system associated with the first merchant system, a second merchant system associated with the acquirer system associated with the first merchant system, a payment device which initiated a payment transaction with the first merchant system, or any combination thereof;

classify the first merchant system and at least one related entity of the plurality of related entities in a first group risk class based on at least one risk score generated by the fraud detection system for the at least one related entity, the first group risk class based on a determination that the first merchant system and the at least one related entity are engaged in collusive transaction fraud;

determine a corrected merchant category code for the first merchant system by comparing the transaction data for the first merchant system to transaction data baselines of a plurality of merchant category codes and/or comparing the time series data for the first merchant system to time series baselines of a plurality of merchant category codes;

automatically modify the merchant category code for the first merchant system by modifying the merchant category code stored in the database of the electronic payment processing network to the corrected merchant category code;

after modifying the merchant category code for the first merchant system, receive, with a transaction processing system in the electronic payment processing network, a payment transaction request associated with the first merchant system and the at least one related entity; and

process, with the transaction processing system, the payment transaction request using the corrected merchant category code based on the corrected category code stored in the database of the electronic payment processing network.

10 . The computer program product of claim 9 , wherein the one or more instructions further cause the at least one processor to:

in response to classifying the first merchant system and the at least one related entity of the plurality of related entities in the first group risk class, automatically initiate an investigation protocol.

11 . The computer program product of claim 9 , wherein the one or more instructions further cause the at least one processor to:

determine a region code associated with the first merchant system; and

detect illicit transaction activity of the first merchant system based on the corrected merchant category code and the region code.

12 . The computer program product of claim 9 , wherein the one or more instructions further cause the at least one processor to:

identify at least one payment transaction conducted between the first merchant system and the at least one related entity of the plurality of related entities in the first group risk class;

determine that the at least one payment transaction is a cross-border payment transaction associated with a first region and a second region;

generate a transaction risk score associated with the at least one payment transaction based on at least one of the following: a status of the at least one payment transaction as a cross-border payment transaction, the first region, the second region, and the merchant category code;

determine that the transaction risk score satisfies a threshold; and

classify the at least one payment transaction in a first transaction risk class in response to determining that the transaction risk score satisfies the threshold.

13 . The computer-implemented method of claim 1 , wherein the determining the plurality of related entities related to the first merchant system comprises retrieving from an entity relationship database the plurality of related entities related to the first merchant system based on data stored in the entity relationship database, wherein the computer-implemented method further comprises:

for each entity of the plurality of related entities related to the first merchant system, generating, with the at least one processor, an entity risk score, the entity risk score for the at least one related entity comprises the at least one risk score associated with the at least one related entity;

determining, with the at least one processor, that the at least one risk score associated with the at least one related entity satisfies a second threshold;

based on the first merchant risk score satisfying the threshold and the at least one risk score associated with the at least one related entity satisfying the second threshold, determining, with the at least one processor, a collusive group; and

in response to determining the collusive group, classifying, with the at least one processor, the first merchant system and at least one related entity of the plurality of related entities in the first group risk class.

14 . The computer-implemented method of claim 1 , further comprising:

modifying, with the transaction processing system, the payment transaction request with the corrected merchant category code by modifying the merchant category code in the payment transaction request with the corrected merchant category code based on the corrected merchant category code stored in the database of the electronic payment processing network; and

processing, with the transaction processing system, the payment transaction request using the corrected merchant category code.

Assignments (1)

Continuity (1)

References Cited (25)