Systems and methods for fraud monitoring

Patent metadata

Field	Value
Publication number	US-12198140-B2
Application number	US-202318386367-A
Country	US
Kind code	B2
Filing date	Nov 2, 2023
Priority date	Aug 23, 2021
Publication date	Jan 14, 2025
Grant date	Jan 14, 2025

Abstract

Systems and methods for fraud monitoring are disclosed, including: receiving a transaction request associated with a first instrument of a user; extracting, characteristics of the transaction request; identifying, by the first processor, user data based on the transaction request; determining a fraud severity value and notification value based on inputting the characteristics and user data into a fraud machine learning model; performing a first fraud action based on the fraud severity value; wherein the first fraud action is at least one selected from the group of locking the first instrument for a period of time, deactivating the first instrument, and electronically transmitting a first query message to a user device associated with the first instrument; and transmitting a fraud notification based on the notification value, wherein the fraud notification includes severity information associated with the fraud severity value.

First claim

What is claimed is:

1. A computer-implemented fraud monitoring method, comprising:
receiving, by a first processor, a transaction request associated with a first instrument of a user;
extracting, by the first processor, one or more transaction data values from the transaction request;
retrieving, from the first instrument of the user identified from the transaction request, one or more application data comprising current location data associated with the first instrument of the user;
computing a deviation in a proximity between a current location of the first instrument indicated by the application data, and a geographical merchant location extracted from the transaction request;
determining, by applying a machine learning model to historical transaction data, one or more correlation values between the one or more transaction data values and one or more fraud data records retrieved from a digital ledger, wherein the digital ledger contains a plurality of fraud data records comprising one or more fraud notifications and user fraud reports;
determining, by the first processor, a fraud severity value representing a probability that the transaction request is fraudulent as a function of the deviation in the proximity and the one or more correlation values;
performing a first fraud action, by the first processor, based on the fraud severity value, wherein the first fraud action is at least one selected from the group of locking the first instrument for a period of time, and deactivating the first instrument; and
determining, based on the fraud severity value, and one or more fraud preference values provided by the user, a fraud notification value representing the necessity of warning one or more account issuers about the fraud,
wherein the first fraud action further comprises electronically transmitting, over a network, a fraud notification message to the first instrument of the user when the fraud notification value exceeds a first notification threshold,
wherein the machine learning model is trained with a training set comprising transaction data from each of a plurality of instruments associated with the user.

2. The computer-implemented fraud monitoring method 1 , wherein the one or more transaction data values include a payment instrument type, a user identifier associated with the payment instrument, the geographical merchant location and one or more merchant identifiers.

3. The computer-implemented fraud monitoring method 1 , wherein the one or more application data further comprises historical location data associated with the first instrument of the user.

4. The computer-implemented fraud monitoring method 1 , further comprising electronically transmitting, by the first processor, a fraud notification message to a second instrument of the user when the fraud notification value exceeds a second notification threshold.

5. The computer-implemented fraud monitoring method 1 , wherein the training set further comprises at least one of a fraud status and fraud type associated with each of the plurality of instruments.

6. The computer-implemented fraud monitoring method 1 , wherein the transaction data is retrieved by the first processor from the distributed ledger of a decentralized network.

7. The computer-implemented fraud monitoring method 1 , wherein the transaction data includes information received from one or more user instruments over the network.

8. The computer-implemented fraud monitoring method 1 , wherein the network is a financial consortium network.

9. The computer-implemented fraud monitoring method 1 , wherein the period of time for locking the first instrument is based on the fraud severity value.

10. A computer-implemented fraud monitoring system, comprising a first processor communicatively coupled to a first server through a network, wherein the first processor is configured to:
receive a transaction request associated with a first instrument of a user;
extract one or more transaction data values from the transaction request;
retrieve, from the first instrument of the user identified from the transaction request, one or more application data comprising current location data associated with the first instrument of the user;
compute a deviation in a proximity between a current location of the first instrument indicated by the application data, and a geographical merchant location extracted from the transaction request;
determine, by applying a machine learning model to historical transaction data, one or more correlation values between the one or more transaction data values and one or more fraud data records retrieved from a digital ledger, wherein the digital ledger contains a plurality of fraud data records comprising one or more fraud notifications and user fraud reports;
determine a fraud severity value representing a probability that the transaction request is fraudulent as a function of the deviation in the proximity and the one or more correlation values;
perform a first fraud action, based on the fraud severity value, wherein the first fraud action is at least one selected from the group of locking the first instrument for a period of time, and deactivating the first instrument; and
determine, based on the fraud severity value, and one or more fraud preference values provided by the user, a fraud notification value representing the necessity of warning one or more account issuers about the fraud,
wherein the first fraud action further comprises electronically transmitting, over a network, a fraud notification message to the first instrument of the user when the fraud notification value exceeds a first notification threshold,
wherein the machine learning model is trained with a training set comprising transaction data from each of a plurality of instruments associated with the user.

11. The computer-implemented fraud monitoring system 10 , wherein the first processor is configured to adjust one of the fraud severity value and the fraud notification value based on the application data.

12. The computer-implemented fraud monitoring system 10 , wherein the first processor is configured to adjust one of the fraud severity value and the fraud notification value based on the one or more fraud preference values.

13. The computer-implemented fraud monitoring system 10 , wherein the first processor is further configured to transmit a fraud notification message to a second instrument of the user when the fraud notification value exceeds a second notification threshold.

14. The computer-implemented fraud monitoring system 10 , wherein the training set further comprises at least one of a fraud status and fraud type associated with each of the plurality of instruments.

15. A non-transitory computer-accessible medium having stored thereon computer-executable instructions, when executed on a computer arrangement, cause the computer arrangement to perform procedures comprising:
receiving a transaction request associated with a first instrument of a user;
extracting one or more transaction data values from the transaction request;
retrieving, from the first instrument of the user identified from the transaction request, one or more application data comprising current location data associated with the first instrument of the user;
computing a deviation in a proximity between a current location of the first instrument indicated by the application data, and a geographical merchant location extracted from the transaction request;
determining, by applying a machine learning model to historical transaction data, one or more correlation values between the one or more transaction data values and on

Assignees

• Capital One Services Llc

Inventors

• Benkreira Abdelkader
• Maiman Tyler
• Way Brendan

Classifications

• G06F18/214

  Generating training patterns; Bootstrap methods, e.g. bagging or boosting · CPC title

• G06N20/00

  Machine learning · CPC title

• G06Q20/389

  Keeping log of transactions for guaranteeing non-repudiation of a transaction · CPC title

• G06Q20/405

  Establishing or using transaction specific rules · CPC title

• G06Q20/407

  Cancellation of a transaction · CPC title