Enrollment and authentication over a phone call in call centers
Embodiments described herein provide for a voice biometrics system execute machine-learning architectures capable of passive, active, continuous, or static operations, or a combination thereof. Systems passively and/or continuously, in some cases in addition to actively and/or statically, enrolling speakers. The system may dynamically generate and update profiles corresponding to end-users who contact a call center. The system may determine a level of enrollment for the enrollee profiles that limits the types of functions that the user may access. The system may update the profiles as new contact events are received or based on certain temporal triggering conditions.
1. A computer-implemented method comprising:
receiving, by a computer, from one or more devices, contact data associated with a plurality of contact events;
generating, by the computer, a profile for each contact event of the plurality of contact events using a set of features extracted using the contact data, wherein at least one voiceprint based on the set of features is associated with a plurality of profiles;
generating, by the computer, a profile similarity score using a first profile for a first contact event and a second profile for a second contact event; and
responsive to the computer determining that the profile similarity score fails to satisfy a profile similarity threshold, storing, by the computer, the second profile as a new profile using a profile status indicator and a second set of features having a second voiceprint generated from the second contact event and used to generate the profile similarity score.
2. The method according to claim 1 , further comprising:
generating, by the computer, a second profile similarity score using the first profile for the first contact event and a third profile for a third contact event; and
responsive to the computer determining that the third profile similarity score fails the profile similarity threshold, updating, by a computer, the first profile using a third set of features.
3. The method according to claim 1 , further comprising:
receiving, by the computer, an identity claim in the contact data of the second contact event indicating one or more accounts associated with the identity claim; and
querying, by the computer, a database for the first profile associated with the account associated with the identity claim.
4. The method according to claim 1 , further comprising:
generating, by the computer, a second profile similarity score for the first profile using first contact data and third contact data in the first profile; and
in response to determining that the second profile similarity score fails to satisfy the profile similarity threshold, splitting, by the computer, the first profile by generating a third profile as another new profile, the first profile containing the first contact data and the third profile containing the third contact data.
5. The method according to claim 1 , further comprising:
receiving, by the computer, inbound contact data for an inbound end-user;
generating, by the computer, an authentication score for the inbound end-user using the inbound contact data and first contact data of the first profile; and
authenticating, by the computer, the inbound end-user in response to determining that the authentication score satisfies an authentication threshold.
6. The method according to claim 1 , further comprising storing, by the computer, the second profile into a quarantine memory location.
7. The method according to claim 6 , further comprising updating, by the computer, a status of the second profile from a quarantine status to an active status in response to authenticating a third profile matched to the second profile.
8. The method according to claim 1 , further comprising:
generating, by the computer, a risk score for the second profile using the second profile data based upon at least one of a local risk factor and a global risk factor; and
updating, by the computer, the second profile to a fraud status in response to determining that the risk score fails to satisfy a risk threshold score.
9. The method according to claim 1 , further comprising linking, by the computer, a subset of one or more profiles associated with one or more end-users of the plurality of profiles with an account associated with the one or more end-users.
10. The method according to claim 9 , further comprising unlinking, by the computer, at least one profile of the subset of profiles from the account associated with the one or more end-users.
11. A system comprising:
a database configured to store a plurality of profiles; and
a computer comprising a processor configured to:
receive from one or more devices, contact data associated with a plurality of contact events;
generate a profile for each contact event of the plurality of contact events using a set of features extracted using the contact data, wherein at least one voiceprint based on the set of features is associated with a plurality of profiles;
generate a profile similarity score using a first profile for a first contact event and a second profile for a second contact event; and
responsive to determining that the profile similarity score fails to satisfy a profile similarity threshold, store the second profile as a new profile using a profile status indicator and a second set of features having a second voiceprint generated from the second contact event and used to generate the profile similarity score.
12. The system according to claim 11 , wherein the computer is further configured to:
generate a second profile similarity score using the first profile for the first contact event and a third profile for a third contact event;
responsive to determining that the third profile similarity score fails the profile similarity threshold:
update the first profile using a third set of features.
13. The system according to claim 11 , wherein the computer is further configured to:
receive an identity claim in second contact data indicating one or more accounts associated with the identity claim; and
query a database for the first profile associated with the account associated with the identity claim.
14. The system according to claim 11 , wherein the computer is further configured to:
generate a second profile similarity score for the first profile using first contact data and third contact data in the first profile; and
in response to determining that the second profile similarity score fails to satisfy the profile similarity threshold:
split the first profile by generating a third profile as another new profile, the first profile containing the first contact data and the third profile containing the third contact data.
15. The system according to claim 11 , wherein the computer is further configured to:
receive inbound contact data for an inbound end-user;
generate an authentication score for the inbound end-user using the inbound contact data and first contact data of the first profile; and
authenticate the inbound end-user in response to determining that the authentication score satisfies an authentication threshold.
16. The system according to claim 11 , wherein the computer is further configured to store the second profile into a quarantine memory location.
17. The system according to claim 11 , wherein the computer is further configured to update a status of the second profile from a quarantine status to an active status in response to authenticating a third profile matched to the second profile.
18. The system according to claim 11 , wherein the computer is further configured to:
generate a risk score for the second profile using the second profile data based upon at least one of a local risk factor and a global risk factor; and
update the second profile to a fraud status in response to determining that the risk score fails to satisfy a risk threshold score.
19. The system according to claim 11 , wherein the computer is further configured to link a subset of one or more profiles associated with one or more end-users of the plurality of profiles with an account associated with the one or more end-users.
20. The system according to claim 19 , wherein the computer is further configured to unlink at least one profile of the subset of profiles from the account associated with the one or more end-users.