Call classification through analysis of DTMF events
Systems, methods, and computer-readable media for call classification and for training a model for call classification, an example method comprising: receiving DTMF information from a plurality of calls; determining, for each of the calls, a feature vector including statistics based on DTMF information such as DTMF residual signal comprising channel noise and additive noise; training a model for classification; comparing a new call feature vector to the model; predicting a device type and geographic location based on the comparison of the new call feature vector to the model; classifying the call as spoofed or genuine; and authenticating a call or altering an IVR call flow.
1. A computer-implemented method comprising:
training, by a computer, a classifier model for a phone number based upon one or more prior feature vectors for one or more past calls associated with the phone number, each prior feature vector for a past call is generated based upon dual-tone multifrequency (DTMF) information received during an interactive voice response (IVR) session of the past call;
receiving, by the computer, the DTMF information during the IVR session of a new call purportedly received from the phone number;
generating, by the computer, a feature vector for the new call based upon the DTMF information of the new call; and
authenticating, by the computer, the new call based upon comparing the feature vector for the new call against the classifier model for the phone number trained on the one or more past calls from the phone number.
2. The method according to claim 1 , further comprising:
generating, by the computer, an ideal DTMF tone corresponding to a DTMF tone received during the IVR session of the new call; and
generating, by the computer, the feature vector for the new call based upon one or more differences between the ideal DTMF tone and the DTMF tone of the new call.
3. The method according to claim 2 , further comprising:
estimating, by the computer, additive noise in the DTMF tone of the new call based upon the difference between the ideal DTMF tone and the DTMF tone of the new call.
4. The method according to claim 2 , further comprising:
estimating, by the computer, channel noise in the DTMF tone of the new call based upon the difference between the ideal DTMF tone and the DTMF tone of the new call.
5. The method according to claim 1 , wherein training the classifier model includes:
obtaining, by the computer, the DTMF information of the one or more past calls associated with the phone number; and
generating, by the computer, the one or more prior feature vectors based upon the DTMF information of each past call.
6. The method according to claim 5 , further comprising, for each past call:
generating, by the computer, an ideal DTMF tone corresponding to a DTMF tone received during the IVR session of the past call; and
generating, by the computer, the prior feature vector for the past call based upon one or more differences between the ideal DTMF tone and the DTMF tone of the past call.
7. The method according to claim 1 , wherein authenticating the new call includes:
determining, by the computer, a probability that a device that generated the DTMF information of the new call is a type of device associated with the one or more prior feature vectors according to the classifier model for the phone number.
8. The method according to claim 1 , wherein authenticating the new call includes:
classifying, by the computer, the new call as fraudulent or non-fraudulent according to the classifier model for the phone number.
9. The method according to claim 1 , wherein authenticating the new call includes:
transmitting, by the computer, to a party of the new call an indicator that the new call is authenticated as genuine.
10. The method according to claim 1 , wherein the DTMF information for the new call comprises as at least one of an analog audio signal and an audio packet.
11. A system comprising:
one or more network interfaces configured to receive dual-tone multifrequency (DTMF) information for a plurality of calls associated with a plurality of phone numbers; and
a processor configured to:
train a classifier model for a phone number based upon one or more prior feature vectors for one or more past calls associated with the phone number, each prior feature vector for a past call is generated based upon the DTMF information received during an interactive voice response (IVR) session of the past call;
receive the DTMF information during the IVR session of a new call purportedly received from the phone number;
generate a feature vector for the new call based upon the DTMF information of the new call; and
authenticate the new call based upon comparing the feature vector for the new call against the classifier model for the phone number trained on the one or more past calls from the phone number.
12. The system according to claim 11 , wherein the processor is further configured to:
generate an ideal DTMF tone corresponding to a DTMF tone received during the IVR session of the new call; and
generate the feature vector for the new call based upon one or more differences between the ideal DTMF tone and the DTMF tone of the new call.
13. The system according to claim 12 , wherein the processor is further configured to:
estimate additive noise in the DTMF tone of the new call based upon the difference between the ideal DTMF tone and the DTMF tone of the new call.
14. The system according to claim 12 , wherein the processor is further configured to:
estimate channel noise in the DTMF tone of the new call based upon the difference between the ideal DTMF tone and the DTMF tone of the new call.
15. The system according to claim 11 , wherein, when training the classifier model, the processor is further configured to:
obtain the DTMF information of the one or more past calls associated with the phone number; and
generate the one or more prior feature vectors based upon the DTMF information of each past call.
16. The system according to claim 15 , wherein the processor is further configured to, for each past call:
generate an ideal DTMF tone corresponding to a DTMF tone received during the IVR session of the past call; and
generate the prior feature vector for the past call based upon one or more differences between the ideal DTMF tone and the DTMF tone of the past call.
17. The system according to claim 11 , wherein, when authenticating the new call, the processor is further configured to:
determine a probability that a device that generated the DTMF information of the new call is a type of device associated with the one or more prior feature vectors according to the classifier model for the phone number.
18. The system according to claim 11 , wherein, authenticating the new call, the processor is further configured to:
classify the new call as fraudulent or non-fraudulent according to the classifier model for the phone number.
19. The system according to claim 11 , wherein, wherein authenticating the new call, the processor is further configured to:
transmit to a party of the new call an indicator that the new call is authenticated as genuine.
20. The system according to claim 11 , wherein the DTMF information for the new call comprises as at least one of an analog audio signal and an audio packet.