Adversarially robust voice biometrics, secure recognition, and identification

A computer-implemented method, comprising:

• receiving, by a system having a processor and a memory, voice data corresponding to a user account;
• analyzing, by the system, the voice data to determine one or more characteristics of the voice data;
• comparing, by the system, the one or more characteristics of the voice data to one or more characteristics of a voice fingerprint corresponding to the user account;
• determining, by the system, that a first similarity score corresponding to the one or more characteristics of the voice data and the one or more characteristics of the voice fingerprint is above a first threshold similarity score;
• in response to determining that the first similarity score corresponding to the one or more characteristics of the voice data and the one or more characteristics of the voice fingerprint is above the first threshold similarity score, comparing, by the system, the one or more characteristics of the voice data to a set of previously stored voice fingerprints corresponding to the user account, wherein the set of previously stored voice fingerprints is different from the voice fingerprint, and wherein the set of previously stored voice fingerprints correspond to one or more interactions between the system and a user corresponding to the user account;
• determining, by the system, whether a second similarity score corresponding to the one or more characteristics of the voice data and the set of previously stored voice fingerprints is above a second threshold similarity score;
• and
• in response to determining that the voice information is potentially invalid or fraudulent, presenting an authentication challenge comprising a request in a second language different from a first language to which the voice information is related.

A non-transitory machine-readable medium, comprising executable instructions that, when executed by a processor, facilitate performance of operations, comprising:

• receiving voice information associated with a user account;
• determining one or more characteristics of the voice information based on the analyzing of the voice information;
• determining a first voice likeness score based on comparing the one or more characteristics of the voice information to one or more characteristics of a voice print associated with the user account;
• determining that the first voice likeness score is greater than a first threshold voice likeness score based on comparing the first voice likeness score to the first threshold voice likeness score;
• in response to determining that the first voice likeness score is above the first threshold voice likeness score, comparing the one or more characteristics of the voice information to one or more characteristics of a set of previously stored voice prints associated with the user account, wherein the set of previously stored voice prints is different from the voice print, and wherein the set of previously stored voice prints relate to one or more interactions between the system and a user associated with the user account;
• determining a second voice likeness score based on the comparing of the one or more characteristics of the voice information to the one or more characteristics of the set of previously stored voice prints;
• and
• determining whether the second voice likeness score is greater than a second threshold voice likeness score;
• and
• in response to determining that the voice information is potentially invalid or fraudulent, presenting an authentication challenge comprising a request in a second language different from a first language to which the voice information is related.