Method and system for enrolling a voiceprint in a fraudster database
AssigneeVictrio, Inc.
InventorsRichard Gutierrez, Anthony Rajakumar, Lisa Marie Guerra, David Hartig
Life of this patent
Estimating the term…
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.
Claims23
A method for enrolling a voiceprint in a fraudster database, the method comprising:
- defining storing a fraud model in a memory device, the fraud model comprising at least one hypothesis indicative of a fraudulent extraction of transactions related to a specific geographic location, wherein the fraud model is stored in a memory device;
- processing a batch of transactions audio data based on the fraud model to extract a plurality of transactions related to the specific geographical location, each of the extracted transactions including audio data and a caller identity, the processing performed using an early fraud detection engine;
- processing the audio data and caller identities of the extracted transactions to identify a group of similar audio data representing a unique individual, to determine that the unique individual has called multiple times and used multiple caller identities, and to generate a voiceprint from the audio data of a plurality if similar voices, the processing performed using a voice processing engine;
- and
- enrolling the at least one suspect generated voiceprint in a fraudster database using an enrollment engine.
The method of claim, 1 wherein the enrolling is performed prior to receiving a fraud report from a modern enterprise.
A system for enrolling a voiceprint in a fraudster database, the system comprising:
- a memory device capable of storing a fraud model comprising a plurality of
- transactions originating from one or more phone numbers that match a phone number pattern at least one hypothesis indicative of a fraudulent transaction;
- an early fraud detection engine configured to process a batch of transactions based on the fraud model to extract the plurality of transactions originating from the specific phone number pattern, each of the extracted transactions including audio data and caller identities;
- a voice processing engine capable of processing the audio data and caller identities of the extracted transactions to group similar voices, a batch of audio data based on the fraud model to identify at least one suspect to generate a single voiceprint for each group of similar voices, and to identify a generated voice print that matches audio data for multiple caller identities in the batch of audio data suspected of belonging to a fraudster;
- and an enrollment engine capable of enrolling the at least one suspect matched voiceprint in the fraudster database.
A computer readable non-transitory medium containing a computer program product for enrolling a voiceprint in a fraudster database, the computer program product comprising:
- program code for defining a fraud model selection criteria comprising transactions occurring during a predetermined time period at least one hypothesis indicative of a fraudulent transaction;
- program code for processing a batch of audio data transactions based on the fraud model selection criteria to select the plurality of transactions occurring during the predetermined time period, each of the transactions in the batch of transactions including audio data and caller identity information;
- program code for analyzing the audio data and caller identity information of the selected transactions to identify a suspected fraudster who has called multiple times during the predetermined period and used multiple caller identities identify at least one suspect voiceprint in the batch audio data suspected of belonging to a fraudster;
- and
- program code for enrolling a voiceprint of the suspected fraudster in the fraudster database.
Dependencies are inferred from the claim text.