Systems, methods, and media for determining fraud patterns and creating fraud behavioral models

• US 9,503,571 B2
• Filed: 10/29/2015
• Issued: 11/22/2016
• Est. Priority Date: 04/21/2005
• Status: Active Grant

• Current Assignee: Verint Americas Incorporated (Verint Systems Incorporated)
• Original Assignee: Verint Americas Incorporated (Verint Systems Incorporated)
• Inventors: Guerra, Lisa, Gutierrez, Richard, Hartig, David, Rajakumar, Anthony, Vyas, Vipul
• Primary Examiner(s): WANG, LIANG CHE A
• Application Number:
  US14/926,998
• Publication Number:
  US 20160142534A1
• Time in Patent Office: 390 Days
• Field of Search: 709/202, 709/219, 709/224, 709/225, 709/227, 709/229, 713/186
• US Class Current: 1/1
• CPC Class Codes: G06Q 50/01 Social networking
• CPC Class Codes: G06Q 99/00 Subject matter not provided for in other groups of this subclass
• CPC Class Codes: G10L 17/00 Speaker identification or verification techniques
• CPC Class Codes: G10L 17/04 Training, enrolment or model building
• CPC Class Codes: G10L 17/14 Use of phonemic categorisation or speech recognition prior to speaker recognition or verification
• CPC Class Codes: H04L 67/52 specially adapted for the location of the user terminal
• CPC Class Codes: H04M 2242/14 with services dependent on location
• CPC Class Codes: H04M 3/2281 Call monitoring, e.g. for law enforcement purposes; Call tracing; Detection or prevention of malicious calls
• CPC Class Codes: H04M 3/4936 Speech interaction details speech recognition per se G10L15/00
• CPC Class Codes: H04M 3/51 Centralised call answering arrangements requiring operator intervention , e.g. call or contact centers for telemarketing

Abstract

Systems, methods, and media for analyzing fraud patterns and creating fraud behavioral models are provided herein. In some embodiments, methods for analyzing call data associated with fraudsters may include executing instructions stored in memory to compare the call data to a corpus of fraud data to determine one or more unique fraudsters associated with the call data, associate the call data with one or more unique fraudsters based upon the comparison, generate one or more voiceprints for each of the one or more identified unique fraudsters from the call data, and store the one or more voiceprints in a database.

Claims

1. A method for analyzing call data associated with fraudsters, the method comprising:

executing instructions stored in memory, the instructions to be executed by a processor, the instructions being configured to;

responsive to receiving first call data associated with fraudsters;

(a) generate a voiceprint for each fraudster in the first call data such that a first plurality of voiceprints are generated;

(b) retrieve from a blacklist database a second plurality of voiceprints;

(c) compare each voiceprint of the first plurality of voiceprints with each voiceprint of the second plurality of voiceprints to generate a plurality of relative comparison scores, wherein the comparing is performed using a matrix arranged such that the first plurality of voiceprints are disposed on a first axis of the matrix, the second plurality of voiceprints are disposed on a second axis of the matrix, and such that cells of the matrix include a relative comparison score for each voiceprint pair, wherein relative comparison scores are numerical values that indicate a degree of similarity between two voiceprints;

(d) group together voiceprint pairs having relative comparison scores above or below a predetermined threshold values;

(e) generate a master voiceprint for the group; and

(f) store the master voiceprint in the blacklist database, wherein the master voiceprint is subsequently used for screening new call data to detect fraud in a new call.

2. The method of claim 1, wherein the instructions are further configured to:

generate a fraud behavioral model for one or more voiceprints in the group, the fraud behavioral model including data that identifies behaviors of a unique fraudster; and

validate the association of two or more voiceprints by comparing call data of the two or more voiceprints with the fraud behavioral model.

3. The method of claim 2, wherein the fraud behavior model is generated at least by executing further instructions stored in memory configured to:

extract keywords from the call data that are indicative of fraud;

identify fraud patterns from the keywords; and

store the fraud patterns in a database.

4. The method of claim 3, wherein the instructions are further configured to:

generate a fraud analytics report based on the fraud patterns, the fraud analytics report comprising a set of visual graphs illustrating the fraud patterns, wherein a visual graph of the set of visual graphs comprises a first axis associated with a uniquely identified fraudster and a second axis associated with a number of calls made.

5. The method of claim 3, wherein a visual graph of the set of visual graphs comprises a first axis associated with a uniquely identified fraudster and a second axis associated with a number of accounts.

6. The method of claim 3, wherein keywords include any of a credit card number, a social security number, a bank account number, a telephone number, a name, a location, a username, a password, personally identifiable information, and combinations thereof.

7. A non-transitory computer readable storage media having a program embodied thereon, the program being executable by a processor to perform a method for analyzing call data associated with fraudsters, the method comprising:

responsive to receiving first call data associated with fraudsters;

(a) generating a voiceprint for each fraudster in the first call data such that a first plurality of voiceprints are generated;

(b) retrieving from a blacklist database a second plurality of voiceprints;

(c) comparing each voiceprint of the first plurality of voiceprints with each voiceprint of the second plurality of voiceprints to generate a plurality of relative comparison scores, wherein the comparing is performed using a matrix arranged such that the first plurality of voiceprints are disposed on a first axis of the matrix, the second plurality of voiceprints are disposed on a second axis of the matrix, and such that cells of the matrix include a relative comparison score for each voiceprint pair, wherein relative comparison scores are numerical values that indicate a degree of similarity between two voiceprints;

(d) grouping together voiceprint pairs having relative comparison scores above or below a predetermined threshold values;

(e) generating a master voiceprint for the group; and

(f) storing the master voiceprint in the blacklist database, wherein the master voiceprint is subsequently used for screening new call data to detect fraud in a new call.

8. The non-transitory computer readable storage media of claim 7, wherein the method further includes:

generating a fraud behavioral model for one or more voiceprints in the group, the fraud behavioral model including data that identifies behaviors of a unique fraudster; and

validating the association of two or more voiceprints by comparing call data of the two or more voiceprints with the fraud behavioral model.

9. The non-transitory computer readable storage media of claim 8, wherein the fraud behavior model is generated at least by:

extracting keywords from the call data that are indicative of fraud;

identifying fraud patterns from the keywords; and

storing the fraud patterns in a database.

10. The non-transitory computer readable storage media of claim 9, wherein the method further includes:

generating a fraud analytics report based on the fraud patterns, the fraud analytics report comprising a set of visual graphs illustrating the fraud patterns, wherein a visual graph of the set of visual graphs comprises a first axis associated with a uniquely identified fraudster and a second axis associated with a number of calls made.

11. The non-transitory computer readable storage media of claim 9, wherein a visual graph of the set of visual graphs comprises a first axis associated with a uniquely identified fraudster and a second axis associated with a number of accounts.

12. The non-transitory computer readable storage media of claim 9, wherein keywords include any of a credit card number, a social security number, a bank account number, a telephone number, a name, a location, a username, a password, personally identifiable information, and combinations thereof.

13. A system for analyzing call data associated with fraudsters, the system comprising:

a memory for storing executable instructions;

a processor for executing the instructions, the executable instructions including;

a voice printing module that, responsive to receiving first call data associated with fraudsters, generates a voiceprint for each fraudster in the first call data such that a first plurality of voiceprints are generated;

a comparator module that retrieves from a blacklist database a second plurality of voiceprints, compares each voiceprint of the first plurality of voiceprints with each voiceprint of the second plurality of voiceprints to generate a plurality of relative comparison scores, wherein the comparing is performed using a matrix arranged such that the first plurality of voiceprints are disposed on a first axis of the matrix, the second plurality of voiceprints are disposed on a second axis of the matrix, and such that cells of the matrix include a relative comparison score for each voiceprint pair, wherein relative comparison scores are numerical values that indicate a degree of similarity between two voiceprints; and

a grouping module that groups together voiceprint pairs having relative comparison scores above or below a predetermined threshold values;

a master voice printing module that generates a master voiceprint for the group and stores the master voiceprint in the blacklist database, wherein the master voiceprint is subsequently used for screening new call data to detect fraud in a new call.

14. The system of claim 13, further comprising a fraud pattern identifying module that:

generates a fraud behavioral model for one or more voiceprints in the group, the fraud behavioral model including data that identifies behaviors of a unique fraudster; and

validates the association of two or more voiceprints by comparing call data of the two or more voiceprints with the fraud behavioral model.

15. The system of claim 14 further comprising:

an analytics module that extract keywords from the call data that are indicative of fraud; and

wherein the fraud pattern identifying module generates the fraud behavior model by at least identifying fraud patterns from the keywords and storing the fraud patterns in a database.

16. The system of claim 15, wherein the method further includes:

a reports generator that generates a fraud analytics report based on the fraud patterns, the fraud analytics report comprising a set of visual graphs illustrating the fraud patterns, wherein a visual graph of the set of visual graphs comprises a first axis associated with a uniquely identified fraudster and a second axis associated with a number of calls made.

17. The system of claim 15, wherein a visual graph of the set of visual graphs comprises a first axis associated with a uniquely identified fraudster and a second axis associated with a number of accounts.

18. The system of claim 15, wherein keywords include any of a credit card number, a social security number, a bank account number, a telephone number, a name, a location, a username, a password, personally identifiable information, and combinations thereof.