Fraud detection in voice biometric systems through voice print clustering

A method for fraud prevention upscaling with a fraudster voice print watchlist for a voice biometric system, which method comprises:

• determining that a first number of a plurality of voice prints in the fraudster voice print watchlist is below or at a first scale of voice prints;
• performing, when the first number of voice prints is below or at the first scale of voice prints, first voice authentications for the voice biometric system by matching first received voice prints to the plurality of voice prints in the fraudster voice print watchlist using similarity scores between the first received voice prints and the plurality of voice prints;
• receiving a first voice print of a user during a voice authentication request;
• accessing the fraudster voice print watchlist comprising voice print representatives for a plurality of voice print clusters each having one or more of the plurality of voice prints identified as fraudulent for the voice biometric system;
• determining that one or more of the voice print representatives in the fraudster voice print watchlist meets or exceeds a first biometric threshold for risk detection of the first voice print during fraud prevention operations;
• and
• determining whether the first voice print matches a first one of the plurality of voice print clusters based on a first similarity score between the first voice print and a representative of the first one of the plurality of voice print clusters.

A non-transitory computer-readable medium having stored thereon computer-readable instructions executable to cause fraud prevention upscaling with a fraudster voice print watchlist for a voice biometric system, which fraud prevention operations comprise:

• determining that a first number of voice prints of a plurality of voice prints in the fraudster voice print watchlist meets or exceeds a first scale of voice prints but is below or at a second scale of voice prints;
• generating voice print representatives for a plurality of voice print clusters each having one or more of the plurality of voice prints using a clustering function, wherein the voice print representatives are generated at least in part based on similarity scores between the one or more of the plurality of voice prints in each of the plurality of voice print clusters;
• receiving a first voice print of a user during a voice authentication request;
• accessing the fraudster voice print watchlist comprising voice print representatives for the plurality of voice print clusters each having one or more of the plurality of voice prints identified as fraudulent for the voice biometric system;
• determining that one or more of the voice print representatives in the fraudster voice print watchlist meets or exceeds a first biometric threshold for risk detection of the first voice print during the fraud prevention operations;
• and
• determining whether the first voice print matches a first one of the plurality of voice print clusters based on a first similarity score between the first voice print and a representative of the first one of the plurality of voice print clusters from the voice print representatives.