Granted Patent B2

US 12,657,587 · App. 18/121,364 · Granted Jun 16, 2026

Inventors: Jennifer Kwok (Brooklyn, NY); Vyjayanthi Vadrevu (Pflugerville, TX); Michael Saia (Weehawken, NJ); Viraj Chaudhary (Katy, TX); Phoebe Atkins (Midlothian, VA); Tyler Maiman (Melville, NY); Leeyat Bracha Tessler (Arlington, VA); Ray Cheng (Long Island City, NY)

Assignee: Capital One Services, LLC

G06Q20/40145G06N20/00G06Q20/4016

Abstract

Claims (56)

1 . A method for fraud prevention using behavioral biometrics, the method comprising the steps of:

recording, by a processor, a plurality of behavioral biometrics of a user interacting with a user device;

generating, by the processor and based on the recorded plurality of behavioral biometrics, a behavioral biometric use-print for the user;

storing, by the processor in a database, the behavioral biometric use-print;

monitoring, by the processor, one or more behavioral biometrics of the user while the user accesses a user account on the user device;

applying, by the processor, a machine learning algorithm to the one or more behavioral biometrics of the user and the user's behavioral biometric use-print to determine that there is a material deviation in at least one of the one or more behavioral biometrics from the user's behavioral biometric use-print;

determine, by the processor and based on the material deviation, that the user is under stress;

providing, by the processor, an intervention to the user device based on the determination that the user is under stress, wherein the intervention comprises at least one of:

(i) a pop-up dialogue box window inquiring if the user's actions are currently being prompted by any third-party, or

(ii) a change to privileges in the user account comprising at least one selected from the group of requiring step-up authentication, reducing credit limits, lowering transfer amount maximums, and lowering cash advance maximums; and

receiving, by the processor, user feedback on the intervention, whereby the user feedback is used as an input to the machine learning algorithm to train and refine the machine learning algorithm;

improving, by the processor, the machine learning algorithm by:

optimizing weights of established relationships between and among metrics;

optimizing relationships based on the user feedback; and

testing, by the processor using the machine learning algorithm, detected relationships and analyses based on the relationships through feedback on predictions over time.

2 . The method of claim 1 , wherein the user device is a smart phone with a mobile application associated with the user account.

3 . The method of claim 1 , wherein the user device is a computer with a web browser capable of accessing a website associated with the user account.

4 . The method of claim 1 , wherein the behavioral biometrics comprise at least one selected from a group of: device holding preferences comprising, device orientation, screen focus, interaction gestures, screen interaction pressures, typing speed, typing error rates, typing habits including misspellings, browsing and interaction flow and order, scrolling speed, frequency, and cadence, click rates, frequency, and cadence, swipe speed, frequency, and cadence, jerkiness, stillness, and device elevation.

5 . The method of claim 1 , wherein the behavioral biometric use-print for the user contains enough behavioral biometrics to uniquely identify the user.

6 . The method of claim 1 , wherein the machine learning algorithm's determination of a material deviation is based on a degree of divergence from the user's behavioral biometric use-print and the monitored behavioral biometrics of the user while the user accesses a user account on the user device as well as a type of access attempted by the user in the user account.

7 . The method of claim 6 , wherein the type of access attempted comprises one or more of attempting a password change, attempting an address change, attempting a funds transfer transaction over a threshold amount.

8 . A system for using behavioral biometrics to prevent digital scams, the system comprising:

a memory storing a behavioral biometric use-print for a user; and

a processor configured to:

record a plurality of behavioral biometrics of a user interacting with a user device;

generate, based on the recorded plurality of behavioral biometrics, the behavioral biometric use-print for the user;

monitor one or more behavioral biometrics of the user while the user accesses a user account on the user device;

apply a machine learning algorithm to the one or more behavioral biometrics of the user and the user's behavioral biometric use-print to determine that there is a material deviation in at least one of the one or more behavioral biometrics from the user's behavioral biometric use-print;

conclude, based on the material deviation, that the user is under stress;

provide an intervention to the user device based on the conclusion that the user is under stress, wherein the intervention comprises at least one of:

(i) a pop-up dialogue box window inquiring if the user's actions are currently being prompted by any third-party, or

(ii) a change to privileges in the user account comprising at least one selected from the group of requiring step-up authentication, reducing credit limits, lowering transfer amount maximums, and lowering cash advance maximums; and

receive user feedback on the intervention, whereby the user feedback is used as an input to the machine learning algorithm to train and refine the machine learning algorithm;

improving the machine learning algorithm by:

optimizing weights of established relationships between and among metrics;

optimizing relationships based on the user feedback; and

testing, using the machine learning algorithm, detected relationships and analyses based on the relationships through feedback on predictions over time.

9 . The system of claim 8 , wherein the user device is a smart phone with a mobile application associated with the user account.

10 . The system of claim 8 , wherein the user device is a computer with a web browser capable of accessing a website associated with the user account.

11 . The system of claim 8 , wherein the behavioral biometrics comprise one or more of: device holding preferences, device orientation, screen focus, interaction gestures, screen interaction pressures, typing speed, typing error rates, typing habits including misspellings, browsing and interaction flow and order, scrolling speed, frequency, and cadence, click rates, frequency, and cadence, swipe speed, frequency, and cadence, jerkiness, stillness, and device elevation.

12 . The system of claim 8 , wherein the machine learning algorithm's determination of a material deviation is based on a degree of divergence from the user's behavioral biometric use-print and the monitored behavioral biometrics of the user while the user accesses a user account on the user device as well as a type of access attempted by the user in the user account.

13 . The system of claim 12 , wherein the type of access attempted comprises one or more of attempting a password change, attempting an address change, attempting a funds transfer transaction over a threshold amount.

14 . A computer-readable non-transitory medium comprising computer executable instructions that, when executed by at least one processor, perform procedures comprising the steps of:

recording a plurality of behavioral biometrics of a user interacting with a user device;

generating, based on the recorded plurality of behavioral biometrics, a behavioral biometric use-print for the user;

monitoring one or more behavioral biometrics of the user while the user accesses a user account on the user device;

applying a machine learning algorithm to the one or more behavioral biometrics of the user and the user's behavioral biometric use-print to determine that there is a material deviation in at least one of the one or more behavioral biometrics from the user's behavioral biometric use-print;

concluding, based on the material deviation, that the user is under stress;

providing an intervention to the user device based on the conclusion that the user is under stress, wherein the intervention comprises at least one of:

(i) a pop-up dialogue box window inquiring if the user's actions are currently being prompted by any third-party, or

(ii) a change to privileges in the user account comprising at least one selected from the group of requiring step-up authentication, reducing credit limits, lowering transfer amount maximums, and lowering cash advance maximums; and

receiving user feedback on the intervention, whereby the user feedback is used as an input to the machine learning algorithm to train and refine the machine learning algorithm;

improving the machine learning algorithm by:

optimizing weights of established relationships between and among metrics;

optimizing relationships based on the user feedback; and

testing, using the machine learning algorithm, detected relationships and analyses based on the relationships through feedback on predictions over time.

Assignments (1)

Continuity (1)

References Cited (14)