Sampling of telemetry events to control event volume cost and address privacy vulnerability

1. A system for sampling telemetry events, comprising:

a processor; and

a memory, coupled to the processor, configured to store executable instructions, wherein the instructions, when executed by the processor, cause the processor to:

receive a plurality of telemetry events, related to an application, from a plurality of client devices;

generate a sampling model for collecting a telemetry event based on the plurality of telemetry events, wherein the sampling model defines under what conditions the telemetry event is to be reported by a client device;

generate an instruction for determining whether to report an incoming same telemetry event by the client device based on the sampling model; and

provide the instruction to the client device, to allow the client device to determine whether to report the incoming same telemetry event based on the instruction.

2. The system of claim 1 , wherein, to generate the sampling model and the instruction, the instructions further cause the processor to:

determine whether a statistically significant insight is obtained for the telemetry event based on the plurality of telemetry events;

if the statistically significant insight is obtained for the telemetry event, generate a corresponding instruction to instruct the client device to discard new instances of the incoming same telemetry event.

3. The system of claim 1 , wherein, to generate the sampling model and the corresponding instruction, the instructions further cause the processor to:

determine whether a statistically significant insight is obtained for the telemetry event based on one or more insights received from one or more of the plurality of client devices;

if the statistically significant insight is obtained for the telemetry event, generate a corresponding instruction to instruct the client device to discard the incoming same telemetry event.

4. The system of claim 1 , wherein, to generate the sampling model for collecting the telemetry event based on the plurality of telemetry events, the instructions further cause the processor to:

dynamically update the generated sampling model based on additional telemetry events reported by the plurality of client devices.

5. The system of claim 4 , wherein:

to generate the sampling model, the instructions further cause the processor to generate a client-version model for the plurality of client devices; and

to generate and provide the instruction to the client device, the instructions further cause the processor to provide the client-version model to the client device, to allow the client device to determine whether to report the incoming same telemetry event based on the client-version model.

6. The system of claim 1 , wherein the instructions further cause the processor to:

determine that one of the plurality of telemetry events logs sensitive data; and

ping a client device that has reported the one of the plurality of telemetry events not to collect a same telemetry event that logs the sensitive data.

7. The system of claim 1 , wherein the instructions further cause the processor to:

determine that one of the plurality of telemetry events logs sensitive data; and

ping a client device that has reported the one of the plurality of telemetry events to hash a field that logs the sensitive data before reporting a same telemetry event that logs the sensitive data.

8. The system of claim 1 , wherein the instructions further cause the processor to:

determine that one of the plurality of telemetry events logs sensitive data; and

ping a client device that has reported the one of the plurality of telemetry events to report a same telemetry event that logs the sensitive data through a pipeline that has a higher-level security.

9. A method for sampling telemetry events, comprising:

receiving, by a cloud-based server, a plurality of telemetry events, related to an application, from a plurality of client devices;

generating, by the cloud-based server, a sampling model for collecting a telemetry event based on the plurality of telemetry events, wherein the sampling model defines under what conditions the telemetry event is to be reported by a client device;

generating, by the cloud-based server, an instruction for determining whether to report an incoming same telemetry event by the client device based on the sampling model; and

providing, by the cloud-based server, the instruction to the client device, to allow the client device to determine whether to report the incoming same telemetry event based on the instruction.

10. The method of claim 9 , wherein generating the sampling model and the instruction comprises:

determining, by the cloud-based server, whether a statistically significant insight is obtained for the telemetry event based on the plurality of telemetry events;

if the statistically significant insight is obtained for the telemetry event, generating a corresponding instruction to instruct the client device to discard new instances of the incoming same telemetry event without reporting the incoming same telemetry event to the cloud-based server.

11. The method of claim 9 , wherein generating the sampling model and the instruction comprises:

determining, by the cloud-based server, whether a statistically significant insight is obtained for the telemetry event based on one or more insights received from one or more of the plurality of client devices;

if the statistically significant insight is obtained for the telemetry event, generating a corresponding instruction to instruct the client device to discard the incoming same telemetry event without reporting the incoming same telemetry event to the cloud-based server.

12. The method of claim 11 , wherein each of the one or more insights is obtained by a corresponding client device through aggregating a plurality of same telemetry events that occurred on the corresponding client device.

13. The method of claim 9 , wherein generating the sampling model for collecting the telemetry event based on the plurality of telemetry events further comprises dynamically updating the generated sampling model based on additional telemetry events received by the cloud-based server.

14. The method of claim 9 , wherein

generating the sampling model further comprises generating a client-version model for the plurality of client devices; and

generating and providing the instruction to the client device further comprises providing the client-version model to the client device, to allow the client device to determine whether to report the incoming same telemetry event based on the client-version model.

15. The method of claim 14 , wherein the client-version model is generated based on patterns of data associated with the telemetry event.

16. The method of claim 15 , wherein the patterns of data correspond to one or more of a locale, country, device type, build version, and time taken to perform an action associated with the telemetry event.

17. The method of claim 9 , further comprising:

determining, by the cloud-based server, that one of the plurality of telemetry events logs sensitive data; and

pinging a client device that has reported the one of the plurality of telemetry events not to collect a same telemetry event that logs the sensitive data.

18. The method of claim 9 , further comprising:

determining, by the cloud-based server, that one of the plurality of telemetry events logs sensitive data; and

pinging a client device that has reported the one of the plurality of telemetry events to hash a field that logs the sensitive data before reporting a same telemetry event that logs the sensitive data.

19. The method of claim 18 , wherein the field that logs the sensitive data is hashed using a salt that is only available in a client boundary.

20. The method of claim 19 , further comprising:

determining, by the cloud-based server, that one of the plurality of telemetry events logs sensitive data; and

pinging a client device that has reported the one of the plurality of telemetry events to report a same telemetry event that logs the sensitive data through a pipeline that has a higher-level security.