US-20260127615-A1 - FRAUD DETECTION SYSTEMS AND METHODS

US20260127615A1US 20260127615 A1US20260127615 A1US 20260127615A1US-20260127615-A1

Abstract

A payment and authentication network may include a communications interface, one or more processors, and a memory. The memory may have instructions stored thereon that, when executed by the one or more processors cause the one or more processors to receive, using the communications interface, transaction information associated with a transaction from a merchant system and generate a fraud risk score based on the transaction information. The instructions may cause the one or more processors to determine that the fraud risk score is indicative that the transaction is likely fraudulent and transmit an alert to a user device that informs a user of the user device that the transaction is likely fraudulent prior to generating an approval decision for the transaction.

Inventors

Chris Burke
Nagaraj Kollar
Jacob Bellman
Ronald Scott Alcorn
Jie He
Laura E. Weinflash
Jack Kohoutek
Clayton Meyer

Assignees

EARLY WARNING SERVICES, LLC

Dates

Publication Date: 20260507
Application Date: 20251222

Claims (20)

1 . A payment and authentication network, comprising: a communications interface; one or more processors; and a memory having instructions stored thereon that, when executed by the one or more processors cause the one or more processors to: receive, using the communications interface, transaction information associated with a transaction of a user and validation information of the user from a requesting device, wherein the transaction information comprises a payment amount of the transaction, a time and date of the transaction, and user identification information of the user; validate the user based on the validation information of the user; communicate an indication to the requesting device that the user has been successfully validated; provide the transaction information to a machine learning model that has been trained to generate fraud risk scores based on a probability that a given transaction is fraudulent; generate, using the machine learning model, a fraud risk score for the transaction that indicates a likelihood that the transaction is fraudulent, wherein the fraud risk score is generated based on the machine learning model identifying one or more risk factors that match with one or more characteristics of the transaction information; determine that the fraud risk score is indicative that the transaction is likely fraudulent; and transmit an alert to a user device that informs a user of the user device that the transaction is likely fraudulent prior to generation of an approval decision for the transaction.
2 . The payment and authentication network of claim 1 , wherein: the validation information is extracted from the transaction information.
3 . The payment and authentication network of claim 1 , wherein: validating comprises comparing the validation information of the user with similar information from one or more sources.
4 . The payment and authentication network of claim 3 , wherein: the one or more sources comprise profile data from an account of the user with the payment and authentication network, a merchant system associated with the transaction, a mobile phone network operator.
5 . The payment and authentication network of claim 3 , wherein: the similar information comprises a device identifier associated with a device of the user.
6 . The payment and authentication network of claim 1 , wherein: validating the user comprises performing a default authentication procedure that applies to all fraud risk determinations of the payment and authentication network.
7 . The payment and authentication network of claim 1 , wherein: validating the user comprises performing an authentication procedure selected by a merchant system associated with the transaction.
8 . A method of performing fraud detection, comprising: receiving, by a payment and authentication network, transaction information associated with a transaction of a user and validation information of the user from a requesting device, wherein the transaction information comprises a payment amount of the transaction, a time and date of the transaction, and user identification information of the user; validating, by the payment and authentication network, the user based on the validation information of the user; communicating, by the payment and authentication network, an indication to the requesting device that the user has been successfully validated; providing, by the payment and authentication network, the transaction information to a machine learning model that has been trained to generate fraud risk scores based on a probability that a given transaction is fraudulent; generating, using the machine learning model, a fraud risk score for the transaction that indicates a likelihood that the transaction is fraudulent, wherein the fraud risk score is generated based on the machine learning model identifying one or more risk factors that match with one or more characteristics of the transaction information; determining, by the payment and authentication network, that the fraud risk score is indicative that the transaction is likely fraudulent; and transmitting, by the payment and authentication network, an alert to a user device that informs a user of the user device that the transaction is likely fraudulent prior to generation of an approval decision for the transaction.
9 . The method of performing fraud detection of claim 8 , wherein: validating the user comprises using one or both of biometric authentication and multi-factor authentication.
10 . The method of performing fraud detection of claim 8 , further comprising: training the machine learning model by providing data from a plurality of prior fraudulent transactions and a plurality of prior valid transactions to the machine learning model as input variables, wherein: each of the plurality of prior fraudulent transactions and each of the plurality of prior valid transactions comprises an indication of whether a particular transaction was fraudulent; and each of the plurality of prior fraudulent transactions and each of the plurality of prior valid transactions comprises a payment amount of the particular transaction, a time and date of the particular transaction, and an identifier of at least one party to the particular transaction.
11 . The method of performing fraud detection of claim 8 , further comprising: performing one or more fraud risk actions based on a comparison between the fraud risk score and one or more threshold scores.
12 . The method of performing fraud detection of claim 11 , wherein: the one or more fraud risk actions comprise at least one of approving the transaction, canceling the transaction, or triggering a manual review of the transaction.
13 . The method of performing fraud detection of claim 11 , wherein: the one or more fraud risk actions comprise: alerting the user that the transaction is likely fraudulent; and prompting the user for a confirmation of whether to cancel or continue with the transaction.
14 . The method of performing fraud detection of claim 8 , wherein: the alert comprises an indication of why the transaction has been flagged as being likely fraudulent.
15 . A non-transitory computer-readable medium having instructions stored thereon that, when executed by one or more processors of a payment and authentication network, cause the one or more processors to: receive transaction information associated with a transaction of a user and validation information of the user from a requesting device, wherein the transaction information comprises a payment amount of the transaction, a time and date of the transaction, and user identification information of the user; validate the user based on the validation information of the user; communicate an indication to the requesting device that the user has been successfully validated; provide the transaction information to a machine learning model that has been trained to generate fraud risk scores based on a probability that a given transaction is fraudulent; generate, using the machine learning model, a fraud risk score for the transaction that indicates a likelihood that the transaction is fraudulent, wherein the fraud risk score is generated based on the machine learning model identifying one or more risk factors that match with one or more characteristics of the transaction information; determine that the fraud risk score is indicative that the transaction is likely fraudulent; and transmit an alert to a user device that informs a user of the user device that the transaction is likely fraudulent prior to generation of an approval decision for the transaction.
16 . The non-transitory computer-readable medium of claim 15 , wherein: the alert comprises one or more warning signs regarding one or more known scam types.
17 . The non-transitory computer-readable medium of claim 16 , wherein: the one of more known scam types are related to scams that the payment and authentication network has deemed as being potentially similar to the transaction.
18 . The non-transitory computer-readable medium of claim 15 , wherein: validating the user comprises comparing at least one of a device identifier, an IP address, or a browser cookie with data from an account of the user associated with the payment and authentication network.
19 . The non-transitory computer-readable medium of claim 18 , wherein: validating the user is performed passively without requiring the user to access an authentication page.
20 . The non-transitory computer-readable medium of claim 15 , wherein: the requesting device is the user device.

Description

CROSS-REFERENCES TO RELATED APPLICATIONS This application is a continuation of U.S. Non-Provisional patent application Ser. No. 17/824,688, filed May 25, 2022, entitled “FRAUD DETECTION SYSTEMS AND METHODS”, which claims the benefit of priority to U.S. Provisional Patent Application No. 63/192,979, filed May 25, 2021, entitled “FRAUD DETECTION SYSTEMS AND METHODS”, the contents of which are herein incorporated in their entirety. BACKGROUND OF THE INVENTION Conventional payment systems facilitate transactions using one or more payment sources. While many financial systems have fraud detection services that help identify attempts to use stolen cards and other misuse of payment media, such payment systems do not sufficiently protect users from authorizing payments to scammers and/or other fraudsters. Therefore, users are often exposed to risks of various fraudulent transactions even while in sole possession of their payment media. This may result in financial losses for the user and/or the user's financial institution. Therefore, improvements in fraud detection in payment systems are desired. BRIEF SUMMARY OF THE INVENTION Embodiments of the present invention are directed to payment systems that provide real-time fraud risk scoring during a transaction. Embodiments may provide risk scores prior to payment to enable the user or the user's financial institution to cancel the transaction prior to the transfer of any funds. Embodiments may provide alerts to the user that may educate the user as to the potential fraudulent activity, and may provide the user with one or more sources of additional information. This may slow down the transaction to provide the user an opportunity to make an informed decision about whether to proceed with the transaction. Some embodiments of the present technology may encompass payment and authentication networks. The networks may include a communications interface. The networks may include one or more processors. The networks may include a memory having instructions stored thereon. When executed by the one or more processors the instructions may cause the one or more processors to receive, using the communications interface, transaction information associated with a transaction from a requesting system. The instructions may cause the one or more processors to generate a fraud risk score based on the transaction information. The instructions may cause the one or more processors to determine that the fraud risk score is indicative that the transaction is likely fraudulent. The instructions may cause the one or more processors to transmit an alert to a user device that informs a user of the user device that the transaction is likely fraudulent prior to generating an approval decision for the transaction. In some embodiments, the instructions further cause the one or more processors to validate the user prior to generating the fraud risk score. The alert may include information about one or more possible forms of fraudulent activity. The alert may include a link to one or more sources of more detailed information about one or more possible forms of fraudulent activity. The alert may include an override feature that enables the user to accept the transaction. The alert may include a cancellation feature that enables the user to terminate the transaction without completing payment. The transaction information may include one or more variables selected from the group consisting of a payment amount, an identifier of the user, a recipient identifier, location data of the user, and location data of the recipient. Some embodiments of the present technology may encompass methods of performing fraud detection. The methods may include receiving, using a communications interface, transaction information associated with a transaction from a requesting system. The methods may include generating a fraud risk score based on the transaction information. The methods may include determining that the fraud risk score is indicative that the transaction is likely fraudulent. The methods may include transmitting an alert to a user device that informs a user of the user device that the transaction is likely fraudulent prior to generating an approval decision for the transaction. In some embodiments, the fraud risk score may be generated using a machine learning model. The machine learning model may be trained using transaction information from known fraudulent transactions. The machine learning model may include a deterministic machine learning model. The machine learning model may include a non-deterministic machine learning model. The methods may include determining one or more device characteristics associated with the requesting system. A form of the alert may be based on the one or more device characteristics. The one or more device characteristics may include device capabilities of the requesting system. Some embodiments of the present technology may encompass non-transitory computer-readable mediums havi