ANOMALY DETECTION IN A NETWORK COUPLING STATE INFORMATION WITH MACHINE LEARNING OUTPUTS

• US 20170104774A1
• Filed: 10/08/2015
• Published: 04/13/2017
• Est. Priority Date: 10/08/2015
• Status: Active Grant

• Current Assignee: Cisco Technology, Inc. (Cisco Systems, Inc.)
• Original Assignee: Cisco Technology, Inc. (Cisco Systems, Inc.)
• Inventors: Vasseur, Jean-Philippe, Flacher, Fabien, Mermoud, Grgory
• Granted Patent: US 10,063,575 B2
• Time in Patent Office: Days
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: H04L 63/1425 Traffic logging, e.g. anomaly detection

Abstract

In one embodiment, a device in a network receives an output of an anomaly detection model. The device receives state information surrounding the output of the anomaly detection model. The device determines whether the state information supports the output of the anomaly detection model. The device causes the anomaly detection model to be adjusted based on a determination that the state information does not support the output of the anomaly detection model.

Claims

1. A method, comprising:

receiving, at a device in a network, an output of an anomaly detection model;

receiving, at the device, state information surrounding the output of the anomaly detection model;

determining, by the device, whether the state information supports the output of the anomaly detection model; and

causing, by the device, the anomaly detection model to be adjusted based on a determination that the state information does not support the output of the anomaly detection model.

2. The method as in claim 1, wherein the state information comprises at least one of:

information indicative of congestion in the network or a measured quality of service (QoS) metric.

3. The method as in claim 1, wherein the state information comprises at least one of:

system metrics of a particular node in the network or link metrics of a particular set of one or more links in the network.

4. The method as in claim 1, wherein receiving the state information surrounding the output of the anomaly detection model comprises:

sending, by the device, a request to a node in the network for the state information, wherein the request is generated based on a type of the anomaly detection model.

5. The method as in claim 1, wherein receiving the state information surrounding the output of the anomaly detection model comprises:

retrieving the state information from a local memory of the device.

6. The method as in claim 1, further comprising:

determining, by the device, a severity associated with the output of the anomaly detection model based on the state information.

7. The method as in claim 6, wherein determining the severity associated with the output of the anomaly detection model comprises:

comparing, by the device, the state information to one or more threshold values.

8. The method as in claim 6, wherein determining the severity associated with the output of the anomaly detection model comprises:

providing, by the device, the output of the anomaly detection model and the state information to a user interface; and

receiving, via the user interface, data indicative of the severity.

9. The method as in claim 1, wherein causing the anomaly detection model to be adjusted comprises:

updating, locally at the device, the anomaly detection model using incremental learning.

10. The method as in claim 1, wherein the state information comprises information that was not used as input to the anomaly detection model.

11. An apparatus, comprising:

one or more network interfaces to communicate with a network;

a processor coupled to the network interfaces and adapted to execute one or more processes; and

a memory configured to store a process executable by the processor, the process when executed configured to;

receive an output of an anomaly detection model;

receive state information surrounding the output of the anomaly detection model;

determine whether the state information supports the output of the anomaly detection model; and

cause the anomaly detection model to be adjusted based on a determination that the state information does not support the output of the anomaly detection model.

12. The apparatus as in claim 11, wherein the state information comprises at least one of:

information indicative of congestion in the network or a measured quality of service (QoS) metric.

13. The apparatus as in claim 11, wherein the state information comprises at least one of:

system metrics of a particular node in the network or link metrics of a particular set of one or more links in the network.

14. The apparatus as in claim 11, wherein the apparatus receives the state information surrounding the output of the anomaly detection model by sending, a request to a node in the network for the state information, wherein the request is generated based on a type of the anomaly detection model.

15. The apparatus as in claim 11, wherein the apparatus receives the state information surrounding the output of the anomaly detection model by retrieving the state information from a local memory of the device.

16. The apparatus as in claim 11, wherein the process when executed is further configured to:

determine a severity associated with the output of the anomaly detection model based on the state information.

17. The apparatus as in claim 16, wherein the apparatus determines the severity associated with the output of the anomaly detection model by comparing the state information to one or more threshold values.

18. The apparatus as in claim 16, wherein the apparatus determines the severity associated with the output of the anomaly detection model by:

providing the output of the anomaly detection model and the state information to a user interface; and

receiving, via the user interface, data indicative of the severity.

19. The apparatus as in claim 11, wherein the apparatus causes the anomaly detection model to be adjusted by retraining the anomaly detection model locally using incremental learning.

20. A tangible, non-transitory, computer-readable media having software encoded thereon, the software when executed by a processor of a device configured to:

receive an output of an anomaly detection model;

receive state information surrounding the output of the anomaly detection model;

determine whether the state information supports the output of the anomaly detection model; and

cause the anomaly detection model to be adjusted based on a determination that the state information does not support the output of the anomaly detection model.