Method and System for Enrolling a Voiceprint in a Fraudster Database

Abstract

Disclosed is a method for enrolling a voiceprint in a fraudster database, the method comprising: a) defining a fraud model comprising at least one hypothesis indicative of a fraudulent transaction; b) processing audio data based on the fraud model to identify at least one suspect voiceprint in the audio data suspected of belonging to a fraudster; and c) enrolling the at least one suspect voiceprint in the fraudster database.

1 - 20 . (canceled) 
     
     
         21 . A method for enrolling a voiceprint into a fraudster database, the method comprising:
 storing a fraud model in a memory device;   receiving at a fraud detection engine a batch of transactions before detection of a fraud for transactions in the batch, each transaction including audio data and a caller identity;   processing the batch of transactions to extract a subset of the batch of transactions based on the fraud model, the processing performed using a selection engine;   processing the audio data of each of the extracted transactions using a voice processing engine to identify a group of transactions in the extracted transactions, the identified group of transactions including audio data characteristic of a unique individual who has called multiple times;   processing the caller identities of the identified group of transactions using the voice processing engine to determine that the unique individual has used multiple caller identities;   generating a voiceprint from the audio data of the identified group of transactions using the voice processing engine; and   enrolling the generated voiceprint into the fraudster database using an enrollment engine.   
     
     
         22 . The method according to  claim 21 , wherein the fraud model comprises transactions related to a phone number pattern belonging to a specific geographical location. 
     
     
         23 . The method according to  claim 21 , wherein the fraud model comprises transactions related to a specific action requested by the caller. 
     
     
         24 . The method according to  claim 23 , wherein the specific action is change of address. 
     
     
         25 . The method according to  claim 21 , wherein processing the audio data of each of the extracted transactions comprises generating a score for each transaction and comparing scores to determine a likelihood that two callers are the same individual. 
     
     
         26 . The method according to  claim 21 , wherein fraud model comprises at least one of a predetermined time period, random sampling, calls made to a same phone number, calls received from a same phone number, calls made to a specific phone number pattern, all calls for a same account, calls received from a specific phone number pattern, call forwarded calls, calls from a specific service provider, similar transactions, and specific out-of-pattern transactions. 
     
     
         27 . The method according to  claim 21 , wherein the fraud model comprises transactions related to a specific geographical location. 
     
     
         28 . The method according to  claim 27 , wherein the geographic location comprises at least one of a billing address, a place of shipment, a zip code, and an address on an account. 
     
     
         29 . A system for early detection of fraudsters, the system comprising:
 a device memory for storing a fraud model;   an early detection engine configured to receive a batch of transactions before the system receives a fraud report for transactions in the batch, each transaction including caller audio data and caller identity data;   a selection engine configured to extract a subset of transactions from the batch of transactions based on the fraud model;   a voice processing engine configured to process the caller audio data and caller identity data of the subset of transactions extracted from the batch to:
 identify a group of transactions including audio data characteristic of a unique individual who has called multiple times using more than one caller identities, and 
 generate a voiceprint from the audio data of the identified group of transactions; and 
   an enrollment engine configured to enroll the generated voiceprint into the fraudster database.   
     
     
         30 . The system according to  claim 29 , wherein the caller audio data includes a voiceprint. 
     
     
         31 . The system according to  claim 29 , wherein the batch of caller audio data is recorded by a third party on behalf of an enterprise. 
     
     
         32 . The system according to  claim 29 , wherein the fraud model includes transactions that exceed a predetermined value. 
     
     
         33 . The system according to  claim 29 , wherein the fraud model includes comparing an expected channel associated with a phone number and a detected channel. 
     
     
         34 . The system according to  claim 33 , wherein the voice processing engine is further configured to analyze audio signal characteristics to determine the detected channel. 
     
     
         35 . The system according to  claim 29 , wherein the fraud model includes calls from frequently used Automatic Number Identifications. 
     
     
         36 . The system according to  claim 29 , wherein the fraud model includes determining if a call is received from a VOIP, landline, or mobile service provider. 
     
     
         37 . The system according to  claim 29 , wherein the fraud model includes calls received from the same phone number. 
     
     
         38 . A non-transitory computer readable storage media having a program embodied thereon, the program being executable by a processor to perform a method for enrolling voiceprints into a database of suspected fraudsters, the method comprising:
 analyzing caller audio data in a batch of transactions related to calls to identify a plurality of groups of multiple calls, each group characteristic of a unique individual;   determining from caller identity data for each of the identified groups of multiple calls if the transactions related to the group of multiple calls include multiple caller identities; and   enrolling a voiceprint of the unique individual into the database of suspected fraudsters.   
     
     
         39 . The system according to  claim 38 , wherein the analyzing caller audio data is performed before receiving a fraud report for the batch of transactions. 
     
     
         40 . The system according to  claim 38 , further comprising extracting a subset of transactions from the batch of transactions by analyzing audio signal characteristics of the call audio data for channel detection.