Identifying and protecting against a computer security threat while preserving privacy of individual client devices using differential privacy machine learning for streaming data

Inventors: Jasjeet Dhaliwal (Morgan Hill, CA); Melanie Beck (Mountain View, CA); Aleatha Parker-Wood (Palo Alto, CA); Geoffrey So (Foster City, CA)

Assignee: NORTONLIFELOCK INC.

G06F21/60 G06N20/00 H04L63/14

View Patent ↗

Loading inventors, assignments & file history…

Monitor This Case

Get email alerts when status or documents change.

Order Certified Copies

Most orders are placed with the USPTO same day — all within 24 business hours.

Order via The Patent Place →

Pre-filled with this patent's details

Quick Facts

Patent No.: US 10,984,113
App. No.: 16/224,394
Granted: Apr 20, 2021
Kind: B1

External Links

USPTO Patent Center↗ Full Patent PDF↗

Abstract

Identifying and protecting against a computer security threat while preserving privacy of individual client devices using differential privacy machine learning for streaming data. In some embodiments, a method may include receiving first actual data values streamed from one or more first local client devices, generating first perturbed data values by adding noise to the first actual data values using a differential privacy mechanism, storing the first perturbed data values, training a machine learning classifier using the first perturbed data values, receiving a second actual data value streamed from a second local client device, generating a second perturbed data value by adding noise to the second actual data value, storing the second perturbed data value, identifying a computer security threat to the second local client device using the second actual data value as input to the trained machine learning classifier, and protecting against the computer security threat.

Claims (51)

1. A computer-implemented method for identifying and protecting against a computer security threat while preserving privacy of individual client devices using differential privacy machine learning for streaming data, at least a portion of the method being performed by a remote server device comprising one or more processors, the method comprising:

receiving, at the remote server device, first actual data values streamed from one or more first local client devices;

generating, at the remote server device, first perturbed data values by adding noise to the first actual data values using a differential privacy mechanism;

storing, at the remote server device, the first perturbed data values;

training, at the remote server device, a machine learning classifier using the first perturbed data values stored at the remote server device;

receiving, at the remote server device, a second actual data value streamed from a second local client device;

generating, at the remote server device, a second perturbed data value by adding noise to the second actual data value using the differential privacy mechanism;

storing, at the remote server device, the second perturbed data value;

identifying, at the remote server device, a computer security threat to the second local client device using the second actual data value as input to the trained machine learning classifier; and

in response to identifying the computer security threat, protecting against the computer security threat by directing performance, at the second local client device or the remote server device, of a remedial action to protect the second local client device from the computer security threat.

2. The method of claim 1 , wherein the performing, at the second local client device, of the remedial action comprises one or more of blocking the second local client device from accessing a network, rolling back one or more changes at the second local client device that were made in response to the computer security threat, or temporarily freezing activity at the second local client device, or some combination thereof.

3. The method of claim 1 , wherein the second actual data value is used as input to the trained machine learning classifier without storing the second actual data value to prevent the second actual data value from being accessed at the remote server device.

4. The method of claim 1 , wherein using the second actual data value instead of the second perturbed data value as input to the trained machine learning classifier increases accuracy of the identifying of the computer security threat.

5. The method of claim 1 , wherein the training, at the remote server device, of the machine learning classifier using the first perturbed data values stored at the remote server device prevents the first actual data values from being exposed due to an explorative attack against the machine learning classifier.

6. The method of claim 1 , further comprising sharing, at the remote server device, the first perturbed data values stored at the remote server device with other devices without revealing the first actual data values.

7. The method of claim 1 , wherein the differential privacy mechanism is not a local differential privacy mechanism.

8. One or more non-transitory computer-readable media comprising one or more computer-readable instructions that, when executed by one or more processors of a remote server device, cause the remote server device to perform a method for identifying and protecting against a computer security threat while preserving privacy of individual client devices using differential privacy machine learning for streaming data, the method comprising:

receiving, at the remote server device, first actual data values streamed from one or more first local client devices;

generating, at the remote server device, first perturbed data values by adding noise to the first actual data values using a differential privacy mechanism;

storing, at the remote server device, the first perturbed data values;

training, at the remote server device, a machine learning classifier using the first perturbed data values stored at the remote server device;

receiving, at the remote server device, a second actual data value streamed from a second local client device;

generating, at the remote server device, a second perturbed data value by adding noise to the second actual data value using the differential privacy mechanism;

storing, at the remote server device, the second perturbed data value at the remote server device;

identifying, at the remote server device, a computer security threat to the second local client device using the second actual data value as input to the trained machine learning classifier; and

9. The one or more non-transitory computer-readable media of claim 8 , wherein the performing, at the second local client device, of the remedial action comprises one or more of blocking the second local client device from accessing a network, rolling back one or more changes at the second local client device that were made in response to the computer security threat, or temporarily freezing activity at the second local client device, or some combination thereof.

10. The one or more non-transitory computer-readable media of claim 8 , wherein the second actual data value is used as input to the trained machine learning classifier without storing the second actual data value to prevent the second actual data value from being accessed at the remote server device.

11. The one or more non-transitory computer-readable media of claim 8 , wherein using the second actual data value instead of the second perturbed data value as input to the trained machine learning classifier increases accuracy of the identifying of the computer security threat.

12. The one or more non-transitory computer-readable media of claim 8 , wherein the training, at the remote server device, of the machine learning classifier using the first perturbed data values stored at the remote server device prevents the first actual data values from being exposed due to an explorative attack against the machine learning classifier.

13. The one or more non-transitory computer-readable media of claim 8 , wherein the method further comprises sharing, at the remote server device, the first perturbed data values stored at the remote server device with other devices without revealing the first actual data values.

14. The one or more non-transitory computer-readable media of claim 8 , wherein the differential privacy mechanism is not a local differential privacy mechanism.

15. A remote server device comprising:

one or more processors; and

one or more non-transitory computer-readable media comprising one or more computer-readable instructions that, when executed by the one or more processors, cause the remote server device to perform a method for identifying and protecting against a computer security threat while preserving privacy of individual client devices using differential privacy machine learning for streaming data, the method comprising: