Shared secret voice authentication

A method for voice authentication for a user of a device, comprising:

• acquiring information including at least one of information about an authorized user, information about an environment in which the device is located and information about prior interactions with the device by the authorized user;
• based on the acquired information, generating an authentication utterance and a prompt therefor;
• providing the prompt to the user, the user being a person attempting to use the device;
• capturing an acoustic sound including a speech of the user;
• detecting the authentication utterance in the speech;
• and
• based on the detection, determining a confidence score which indicates a probability that the user is the authorized user,
• wherein the information is used to generate the authentication utterance that, when spoken by the user, highlights a distinctiveness of the authorized user's voice with respect to the voices of other users,
• wherein generating the authentication utterance includes determining a secret that the authorized user would uniquely be able to provide, wherein the authentication utterance includes the secret.

An apparatus for performing voice authentication for a user of a device, the apparatus having a processor adapted to:

• acquire information including at least one of information about an authorized user, information about an environment in which the device is located and information about prior interactions with the device by the authorized user;
• based on the acquired information, generate an authentication utterance and a prompt therefor;
• cause the prompt to be provided to the user, the user being a person attempting to use the device;
• receive an acoustic sound including a speech of the user;
• detect the authentication utterance in the speech;
• and
• based on the detection, determine a confidence score which indicates a probability that the user is the authorized user,
• wherein the information is used to generate the authentication utterance that, when spoken by the user, highlights a distinctiveness of the authorized user's voice with respect to the voices of other users.