US 12,287,856 B2
Computer-based systems applying textual embedding space software engines to authentication question and answers for identifying a user and methods of use thereof
Kevin Osborn, Newton Highlands, MA (US); and Brian McClanahan, Hyattsville, MD (US)
Assigned to Capital One Services, LLC, McLean, VA (US)
Filed by Capital One Services, LLC, McLean, VA (US)
Filed on Sep. 12, 2022, as Appl. No. 17/942,401.
Prior Publication US 2024/0086501 A1, Mar. 14, 2024
Int. Cl. G06F 21/31 (2013.01)
CPC G06F 21/31 (2013.01) 20 Claims
OG exemplary drawing
 
1. A method, comprising:
receiving, by a processor, at least one authentication answer to each of a plurality of identity-authentication questions provided in an initial authentication question-answer registration by at least one verified user;
wherein the at least one authentication answer comprises at least one word, at least one phrase of words, or any combination thereof;
generating, by the processor, at least one authentication answer embedding vector in a textual embedding space by inputting the at least one authentication answer into at least one embedding engine;
wherein the at least one authentication answer embedding vector in the textual embedding space represents a location in the textual embedding space;
wherein the location in the textual embedding space represents:
(i) the at least one word in the at least one authentication answer, and
(ii) a discourse of the at least one phrase of words in the at least one authentication answer;
wherein the discourse is based on the at least one word in the at least one authentication answer;
storing, by the processor, the at least one authentication answer embedding vector for each of the plurality of identity-authentication questions of the at least one verified user in a non-transient memory;
receiving, by the processor, at least one unverified-user authentication answer for at least one specific identity-authentication question from the plurality of identity-authentication questions of the at least one verified user, in response to posing to at least one unverified user in an identity-authentication session, the at least one specific identity-authentication question;
generating, by the processor, at least one unverified-user authentication answer embedding vector representing at least one unverified-user authentication answer location in the textual embedding space by inputting the at least one unverified-user authentication answer into the at least one embedding engine for each of the at least one specific identity-authentication question;
computing, by the processor, an embedding space distance between the at least one unverified-user authentication answer location and the location in the textual embedding space of the at least one authentication answer embedding vector for each of the at least one specific identity-authentication question of the at least one verified user posed to the at least one unverified user;
wherein, when the at least one unverified user is the at least one verified user, the embedding space distance accounts for semantic variations, discourse variations, or both in the at least one unverified-user authentication answer relative to the at least one authentication answer previously provided in the initial authentication question-answer registration;
computing, by the processor, an accumulated similarity score based the embedding space distance computed for each of the at least one specific identity-authentication question posed to the at least one unverified user; and
identifying, by the processor, the at least one unverified user as the at least one verified user when the accumulated similarity score is higher than a predefined verification threshold score.