Fraudulent call detection

1. A mobile telephone, comprising:

a hardware platform comprising a processor and a memory;

a telecommunication transceiver; and

instructions encoded within the memory to instruct the processor to:

identify a call made via the telecommunication transceiver;

analyze the call, comprising analyzing audio of the call while the call is in progress to identify indicia of deceit, and based on the analysis assign the call a predicted local reputation according to the analysis, including a legitimacy confidence score;

if the legitimacy confidence score is less than a first threshold, terminate the call;

if the legitimacy confidence score is greater than a second threshold, cease analysis of the call; and

if the legitimacy confidence score is between the first and second thresholds, continue analysis of the call.

2. The mobile telephone of claim 1 , wherein the call is an incoming call.

3. The mobile telephone of claim 1 , wherein the call is an outgoing call.

4. The mobile telephone of claim 1 , wherein analyzing the call comprises querying a local contact list for a phone number associated with the call.

5. The mobile telephone of claim 1 , wherein the instructions are to provide a mode wherein the call is not analyzed.

6. The mobile telephone of claim 5 , wherein the instructions are to enter the mode if the call is to or from a known or trusted number.

7. The mobile telephone of claim 5 , wherein the instructions are to, if the call is to or from a number that is not known or trusted, enter the mode only with user confirmation.

8. The mobile telephone of claim 7 , wherein the instructions are to determine that the number is stored in a local contact list but is not trusted before the analysis.

9. The mobile telephone of claim 8 , wherein determining that the number is not trusted comprises determining that a call to or from the number has not previously been analyzed.

10. The mobile telephone of claim 1 , wherein the instructions are to provide a machine learning model for the analysis.

11. One or more tangible, non-transitory computer-readable storage media having stored thereon executable instructions to:

determine that a voice call has been activated;

determine that the voice call is to be handled in a first mode, the first mode to provide analysis of the voice call;

analyze the voice call, including analyzing audio of the call while the call is in progress to identify indicia of deceit, and based on the analysis, assign the voice call a score, wherein the score includes a probability that the voice call is legitimate;

terminate the voice call if the probability is below a lower threshold;

terminate analysis if the probability is above an upper threshold; and

otherwise, subject the voice call to additional analysis.

12. The one or more tangible, non-transitory computer-readable storage media of claim 11 , wherein analyzing the voice call comprises collecting a voice sample and analyzing the voice sample.

13. The one or more tangible, non-transitory computer-readable storage media of claim 12 , wherein analyzing the voice sample comprises voice recognition.

14. The one or more tangible, non-transitory computer-readable storage media of claim 12 , wherein analyzing the voice sample comprises language detection.

15. The one or more tangible, non-transitory computer-readable storage media of claim 12 , wherein analyzing the voice sample comprises accent analysis.

16. The one or more tangible, non-transitory computer-readable storage media of claim 12 , wherein analyzing the voice sample comprises stress or frequency analysis of the voice sample.

17. The one or more tangible, non-transitory computer-readable storage media of claim 12 , wherein analyzing the voice sample comprises text-to-speech conversion.

18. The one or more tangible, non-transitory computer-readable storage media of claim 17 , wherein analyzing the voice sample comprises searching converted text for key words.

19. A method of detecting fraudulent voice calls on a device, comprising:

detecting a call;

determining that the call is of a class to be subjected to analysis;

analyzing call data for the call, comprising analyzing audio of the call for indicia of deceit while the call is in progress;

if the call data indicate, with a first confidence, that the call is fraudulent, terminating the call;

if the call data indicate, with a second confidence, that the call is legitimate, allowing the call; and

otherwise, subjecting the call to ongoing analysis.

20. The method of claim 19 , wherein the call is an incoming call.