US 12,437,301 B2
Real-time updating of a security model
Shriyans Lenkala, Carpentersville, IL (US); Jonathan Barbalas, Chicago, IL (US); and Joel Mampilly, Flushing, NY (US)
Assigned to Capital One Services, LLC, McLean, VA (US)
Filed by Capital One Services, LLC, McLean, VA (US)
Filed on Jan. 4, 2021, as Appl. No. 17/140,680.
Prior Publication US 2022/0215393 A1, Jul. 7, 2022
Int. Cl. G06Q 20/40 (2012.01); G06N 20/00 (2019.01); G06Q 20/32 (2012.01)
CPC G06Q 20/4016 (2013.01) [G06N 20/00 (2019.01); G06Q 20/3224 (2013.01); G06Q 20/325 (2013.01); G06Q 20/4015 (2020.05)] 20 Claims
OG exemplary drawing
 
1. A fraud notification platform for updating a security model using real-time feedback from a user device, the platform comprising:
one or more non-transitory memories configured to store instructions and data; and
one or more processors, coupled to the one or more memories, configured to execute the instructions to:
receive, from the user device, user data that indicates at least one of a credential of a card associated with a user of the user device or an identifier associated with the user device;
receive an indication of an exchange initiated using the credential of the card associated with the user data, wherein the indication identifies exchange data associated with the exchange;
determine, using the security model, a score associated with the exchange based on the exchange data, wherein the score indicates a likelihood that the exchange is valid, wherein the security model comprises a machine learning model trained using real-time feedback and location data to enhance fraud detection accuracy;
transmit, to the user device, a notification based on the score satisfying a first threshold,
wherein the notification requests an indication of whether the exchange is valid regardless of whether the score satisfies a second threshold that is higher than the first threshold, wherein the exchange is enabled to be completed in instances where the score satisfies the first threshold and does not satisfy the second threshold;
receive, from the user device, a response to the notification indicating whether the exchange is valid, wherein the response indicates user device location information associated with the user device;
compare a user device location indicated by the user device location information with an exchange location indicated by the exchange data, wherein the exchange location comprises geographic coordinates associated with the exchange;
determine whether the response is valid based on comparing the user device location and the exchange location; and
update the security model in real-time using an initial determination of the score and an indication of whether the exchange is valid that is based on the response to the notification and whether the response is valid, wherein the updating includes dynamically adjusting model parameters to improve fraud detection accuracy and system security by deploying and training the security model.