Granted Patent B2

US 12,346,426 · App. 17/947,667 · Granted Jul 1, 2025

Inventors: Abdelkader M'hamed Benkreira (Washington, DC); Joshua Edwards (Philadelphia, PA); Michael Mossoba (Great Falls, VA)

Assignee: Capital One Services, LLC

G06F21/34G06N20/00

Abstract

Claims (86)

1. A computing device comprising:

one or more processors; and

memory storing instructions that, when executed by the one or more processors, cause the computing device to:

train, using training data comprising account records from a plurality of different users, a first machine learning model to output, for a particular device, an indication of device reliability data associated with the particular device, wherein the account records are associated with a plurality of devices used by the plurality of different users to access one or more accounts in the account records;

receive, from a user device, a request for access to an account associated with a user;

receive, from one or more databases, account data corresponding to the account,

wherein the account data indicates one or more logins originated from the user;

determine, based on the account data, device history comprising a set of devices used by the user to login to the account within a predetermined period of time;

provide, as input to the trained first machine learning model, the account data;

receive, from the trained first machine learning model, data indicating device reliability for the set of devices;

determining, based on the device history, one or more false devices that the user has not used to access the account for the predetermined period of time;

generate, based on the data indicating device reliability for the set of devices, a set of modified device choices by excluding one or more devices having corresponding reliability levels below a threshold value, from the set of devices, wherein the set of modified device choices comprise the one or more false devices;

generate an authentication question comprising at least one device choice from the modified set of device choices;

generate, based on the account data and the modified set of device choices, a correct answer to the authentication question;

provide the authentication question to the user device;

receive, from the user device, a response to the authentication question;

compare the response to the authentication question to the correct answer; and

grant the user device access to the account based on the response to the authentication question matching the correct answer.

2. The computing device of claim 1 , wherein the training data comprises device information for the plurality of devices used by the plurality of different users comprising:

a frequency of use for each device of the plurality of devices,

a duration of use for each device of the plurality of devices, and

a time lapsed since a last use for each device of the plurality of devices.

3. The computing device of claim 1 wherein the training data comprises web browser information corresponding to a web browser executed by the plurality of devices used by the plurality of different users.

4. The computing device of claim 1 , wherein the training data comprises account information comprising:

one or more questions previously presented to the plurality of different users, and

responses from the plurality of different users.

5. The computing device of claim 1 , wherein the training data comprises transaction information indicating whether transactions conducted by the plurality of devices were fraudulent.

6. The computing device of claim 1 , wherein the instructions, when executed by the one or more processors, cause the computing device to:

train, based on second training data comprising a history of authentication records, a second machine learning model to determine recommended reliability thresholds, wherein the history of authentication records comprise authentication questions and responses associated with different type of devices used by the plurality of different users and the corresponding scoring schemes;

provide, as input to the trained second machine learning model, input data comprising the authentication question and the response to the authentication question from the user; and

receive, as output from the trained second machine learning model, output data indicating a recommended threshold value associated with the user.

7. The computing device of claim 6 , wherein the instructions, when executed by the one or more processors, cause the computing device to:

receive user feedback information indicating whether the set of devices associated with the account data were valid candidates; and

based on the user feedback information, re-train the second machine learning model to modify the recommended threshold value associated with the set of devices.

8. The computing device of claim 1 , wherein the instructions, when executed by the one or more processors, cause the computing device to:

generate the authentication question comprising a first device from the modified set of device choices and a second device that is not included in the set of devices, wherein the first device and the second device are associated with a same device manufacturer.

9. The computing device of claim 1 , wherein the instructions, when executed by the one or more processors, cause the computing device to:

generate the authentication question comprising a first device from the modified set of device choices and a second device that is not included in the set of devices, wherein the first device and the second device are associated with a similar price point.

10. The computing device of claim 1 , wherein the instructions, when executed by the one or more processors, cause the computing device to:

generate the authentication question comprising a first device from the modified set of device choices and a second device that is not included in the set of devices, wherein the first device and the second device are available at a same period of time.

11. The computing device of claim 1 , wherein the set of modified device choices comprise one or more real devices that the user has used to access the account for the predetermined period of time.

12. A method comprising:

training, using training data comprising account records from a plurality of different users, a first machine learning model to output, for a particular device, an indication of device reliability data associated with the particular device, wherein the account records are associated with a plurality of devices used by the plurality of different users to access one or more accounts in the account records;

receiving, from a user device, a request for access to an account associated with a user;

receiving, from one or more databases, account data corresponding to the account,

wherein the account data indicates one or more logins originated from the user;

determining, based on the account data, device history comprising a set of devices used by the user to login to the account within a predetermined period of time;

providing, as input to the trained first machine learning model, the account data;

receiving, from the trained first machine learning model, data indicating device reliability of the set of devices;

determining, based on the device history, one or more false devices that the user has not used to access the account for the predetermined period of time;

generating, based on the data indicating device reliability of the set of devices, a set of modified device choices by excluding one or more devices having corresponding confidence levels below a threshold value, from the set of devices, wherein the set of modified device choices comprise the one or more false devices;

generating an authentication question comprising at least one device choice from the modified set of device choices;

generating, based on the account data and the modified set of device choices, a correct answer to the authentication question;

providing the authentication question to the user device;

receiving, from the user device, a response to the authentication question;

comparing the response to the authentication question to the correct answer; and

granting the user device access to the account based on the response to the authentication question matching the correct answer.

13. The method of claim 12 , wherein the training data comprises device information for the plurality of devices used by the plurality of different users comprising:

a frequency of use for each device of the plurality of devices,

a duration of use for each device of the plurality of devices, and

a time lapsed since a last use for each device of the plurality of devices.

14. The method of claim 12 , wherein the training data comprises web browser information corresponding to a web browser executed by the plurality of devices used by the plurality of different users.

15. The method of claim 12 , wherein the training data comprises account information comprising:

one or more questions previously presented to the plurality of different users, and responses from the plurality of different users.

16. The method of claim 12 , wherein the training data comprises transaction information indicating whether transactions conducted by the plurality of devices were fraudulent.

17. One or more non-transitory computer-readable media storing instructions that, when executed by one or more processors, cause a computing device to:

train, using training data comprising account records from a plurality of different users, a first machine learning model to output, for a particular device, an indication of device reliability data associated with the particular device, wherein the account records are associated with a plurality of devices used by the plurality of different users to access one or more accounts in the account records;

receive, from a user device, a request for access to an account associated with a user;

receive, from one or more databases, account data corresponding to the account, wherein the account data indicates one or more logins originated from the user;

determine, based on the account data, device history comprising a set of devices used by the user to login to the account within a predetermined period of time;

provide, as input to the trained first machine learning model, the account data;

receive, from the trained first machine learning model, data indicating device reliability of the set of devices;

determining, based on the device history, one or more false devices that the user has not used to access the account for the predetermined period of time;

generate, based on the data indicating device reliability of the set of devices, a set of modified device choices by excluding one or more devices having corresponding confidence levels below a threshold value, from the set of devices, wherein the set of modified device choices comprise the one or more false devices;

generate an authentication question comprising at least one device choice from the modified set of device choices;

generate, based on the account data and the modified set of device choices, a correct answer to the authentication question;

provide the authentication question to the user device;

receive, from the user device, a response to the authentication question;

compare the response to the authentication question to the correct answer; and

grant the user device access to the account based on the response to the authentication question matching the correct answer.

18. The computer-readable media of claim 17 , wherein the instructions, when executed by the one or more processors, cause the computing device to:

generate the authentication question comprising a first device from the modified set of device choices and a second device that is not included in the set of devices, wherein the first device and the second device are associated with a same device manufacturer.

19. The computer-readable media of claim 17 , wherein the instructions, when executed by the one or more processors, cause the computing device to:

generate the authentication question comprising a first device from the modified set of device choices and a second device that is not included in the set of devices, wherein the first device and the second device are associated with a similar price point.

20. The computer-readable media of claim 17 , wherein the instructions, when executed by the one or more processors, cause the computing device to:

generate the authentication question comprising a first device from the modified set of device choices and a second device that is not included in the set of devices, wherein the first device and the second device are available at a same period of time.

Assignments (1)

Continuity (1)

References Cited (11)