Heuristic identity authentication engine
AssigneeState Farm Mutual Automobile Insurance Company
InventorsElizabeth Flowers, Puneit Dua, Eric Balota, Shanna L. Phillips
Life of this patent
Estimating the term…
Abstract
A heuristic engine includes capabilities to collect an unstructured data set and authenticate a user's identity. By providing a heuristic algorithm with unstructured data, such as question and answer sets related to a user's activities in the past, a correlation factor can be calculated with a current context. Authenticating users using heuristic algorithms ensures precision and predictability of authentication decisions. Such heuristic algorithms may learn from each interaction and improve the accuracy of authentication decisions in the future.
Claims28
A computer-implemented method, executed with a computer processor, to authenticate an identity in a transaction, comprising:
- retrieving, with the computer processor, an un-structured transaction data set associated with a customer;
- receiving, with the computer processor and via a human-machine interface, an encoded context indicating a natural language input received from the customer, wherein the natural language input is associated with past usage of at least one account;
- accessing, with the computer processor, a heuristic algorithm;
- generating, by executing the heuristic algorithm with the computer processor, and using the un-structured transaction data set and the natural language input as inputs to the heuristic algorithm:
- a suggested set of identity authentication questions, and
- a set of expected identity authentication answers corresponding to the suggested set of identity authentication questions;
- receiving, with the computer processor, an indication of a customer response to at least one question of the suggested set of identity authentication questions;
- and
- training, with the computer processor, the heuristic algorithm using machine learning, wherein the machine learning updates the heuristic algorithm based at least partly on a correlation factor between the set of expected identity authentication answers and the customer response to the at least one question.
(canceled)
(canceled)
(canceled)
A computer system configured to authenticate an identity in a transaction, the computer system comprising at least one of a processor or a transceiver configured to:
- retrieve an un-structured transaction data set associated with a customer;
- receive an encoded context via a human-machine interface, the encoded context indicating a natural language input received from the customer, wherein the natural language input is associated with past usage of at least one account;
- access a heuristic algorithm;
- generate, by executing the heuristic algorithm using the un-structured transaction data set and the natural language input as inputs to the heuristic algorithm:
- a suggested set of identity authentication questions, and
- a set of expected identity authentication answers corresponding to the suggested set of identity authentication questions;
- receiving an indication of a customer response to at least one question of the suggested set of identity authentication questions;
- and
- train the heuristic algorithm using machine learning, wherein the machine learning updates the heuristic algorithm based at least partly on a correlation factor between the set of expected identity authentication answers and the customer response to the at least one question.
(canceled)
(canceled)
A non-transitory computer readable medium, comprising computer readable instructions that, when executed by a computer processor, cause the computer processor to:
- retrieve an un-structured transaction data set associated with a customer;
- receive an encoded context via a human-machine interface, the encoded context indicating a natural language input received from the customer, wherein the natural language input is associated with past usage of at least one account;
- access a heuristic algorithm;
- generate, by executing the heuristic algorithm using the un-structured transaction data set and the natural language input as inputs to the heuristic algorithm:
- a suggested set of identity authentication questions, and
- a set of expected identity authentication answers corresponding to the suggested set of identity authentication questions;
- receive an indication of a customer response to at least one question of the suggested set of identity authentication questions;
- and
- train the heuristic algorithm using machine learning, wherein the machine learning updates the heuristic algorithm based at least partly on a correlation factor between the set of expected identity authentication answers and the customer response to the at least question.
(canceled)
(canceled)
(canceled)
Dependencies are inferred from the claim text.