System and method for gender based authentication of a caller

1. A method of identifying an attempted fraud, the method comprising:

receiving an incoming call from a caller;

determining a gender of the caller;

wherein determining the gender of the caller is done by a unit including a model, the model created by:

training a neural network (NN) using an initial set of labeled data and recording results and associated confidence levels produced by the NN;

selecting, as labeled data, a set of results such that:

a ratio of the number of results related to females to the number of results related to males is within a predefined range, and

a majority of confidence levels respectfully associated with a majority of the results in the set are higher than a threshold; and

retraining the NN using the selected set;

selecting, based on the determined gender, to search for the caller in one of: a first watchlist of untrustworthy female callers and a second watchlist of untrustworthy male callers;

calculating a confidence level for the determination of the gender of the caller; and

if the confidence level is lower than a threshold then performing at least one of:

searching for the caller in the first and second lists; and

updating a model.

2. The method of claim 1 , comprising: if the caller is found in one of the first and second watchlists then performing at least one action related to fraud detection.

3. The method of claim 1 , comprising:

associating a gender of a caller with each of a plurality of recorded interactions;

determining a gender of an untrustworthy caller; and

selecting, based on the determined gender of the untrustworthy caller, to search for the untrustworthy caller in recorded interactions in which the gender of the caller is same as the gender of the untrustworthy caller.

4. The method of claim 3 , comprising: upon identifying the untrustworthy caller in one of the recorded interactions, performing at least one action related to fraud detection.

5. The method of claim 1 , comprising:

associating a gender of a caller with each of a plurality of recorded interactions; and

clustering at least some of the recorded interactions by examining recorded interactions associated with a specific gender.

6. The method of claim 5 , wherein the plurality of recorded interactions includes interactions suspected to be related to fraud.

7. The method of claim 1 , comprising verifying the caller based on matching the determined gender to the gender in a retrieved caller data.

8. The method of claim 1 , comprising iteratively selecting a set of results as labeled data and retraining the NN until the number of results in the selected set meets at least one criterion.

9. The method of claim 1 , wherein the model is generated or updated based on recorded interactions of a specific site.

10. The method of claim 1 , comprising determining how to handle the incoming call based on the determined gender.

11. The method of claim 1 , wherein searching for the caller in one of the first and second watchlists includes matching a voiceprint of the caller with voiceprints of the callers in the watchlist.

12. A method of authenticating a caller, the method comprising:

identifying the gender of the caller;

wherein identifying the gender of the caller is done by a unit including a model, the model created by:

training a neural network (NN) using an initial set of labeled data and recording results and associated confidence levels produced by the NN;

selecting, as labeled data, a set of results such that:

a ratio of the number of results related to females to the number of results related to males is within a predefined range, and

a majority of confidence levels respectfully associated with a majority of the results in the set are higher than a threshold; and

retraining the NN using the selected set:

if the caller is a male, then searching for the caller in a first list of untrustworthy male callers;

if the caller is a female then searching for the caller in a second list of untrustworthy female callers;

calculating a confidence level for the determination of the gender of the caller;

if the confidence level is lower than a threshold then performing at least one of:

searching for the caller in the first and second lists; and

updating a model; and

if the caller is found in one of the lists, then performing at least one action related to security.

13. A system comprising:

a memory; and

a controller configured to:

receive an incoming call from a caller;

determine a gender of the caller;

wherein determining the gender of the caller is done by a unit including a model, the model created by:

training a neural network (NN) using an initial set of labeled data and recording results and associated confidence levels produced by the NN;

selecting, as labeled data, a set of results such that:

a ratio of the number of results related to females to the number of results related to males is within a predefined range, and

a majority of confidence levels respectfully associated with a majority of the results in the set are higher than a threshold; and

retraining the NN using the selected set;

select, based on the determined gender, to search for the caller in one of: a first watchlist of untrustworthy female callers and a second watchlist of untrustworthy male callers;

calculate a confidence level for the determination of the gender of the caller; and

if the confidence level is lower than a threshold then perform at least one of:

search for the caller in the first and second lists; and

update a model.

14. The system of claim 13 , wherein the controller is further configured to: if the caller is found in one of the first and second watchlists then perform at least one action related to fraud detection.

15. The system of claim 13 , wherein the controller is further configured to:

associate a gender of a caller with each of a plurality of recorded interactions;

determine a gender of an untrustworthy caller; and

select, based on the determined gender of the untrustworthy caller, to search for the untrustworthy caller in recorded interactions in which the gender of the caller is same as the gender of the untrustworthy caller.

16. The system of claim 13 , wherein the controller is further configured to:

associate a gender of a caller with each of a plurality of recorded interactions; and

cluster at least some of the recorded interactions by examining recorded interactions associated with a specific gender.

17. The system of claim 13 , wherein the controller is configured to determine the gender of the caller using the model, the model created by:

training a neural network (NN) using an initial set of labeled data and recording results and associated confidence levels produced by the NN;

selecting, as labeled data, a set of results such that:

a ratio of the number of results related to females to the number of results related to males is within a predefined range, and

a majority of confidence levels respectfully associated with a majority of the results in the set are higher than a threshold; and

retraining the NN using the selected set.

18. The system of claim 13 , wherein the controller is configured to determine how to handle the incoming call based on the determined gender.

19. A method of identifying an attempted fraud, the method comprising:

receiving an incoming call from a caller;

determining a gender of the caller; wherein

the gender of the caller is determined by a unit including a model created by:

training a neural network (NN) using an initial set of labeled data and recording results and associated confidence levels produced by the NN;

selecting, as labeled data, a set of results such that:

a ratio of the number of results related to females to the number of results related to males is within a predefined range, and

a majority of confidence levels respectfully associated with a majority of the results in the set are higher than a threshold; and

retraining the NN using the selected set; and

selecting, based on the determined gender, to search for the caller in one of: a first watchlist of untrustworthy female callers and a second watchlist of untrustworthy male callers.