Automatic retraining of machine learning models to detect DDoS attacks

1. A method comprising:

monitoring, at an attack detector, network traffic to detect a Distributed Denial of Service (DDoS) attack by applying a machine learning-based attack detection model against one or more attributes of the network traffic;

responsive to detection by the attack detector of a potential DDoS attack, transmitting an attack mitigation message indicating the potential DDoS attack to an attack mitigation server;

causing, by the attack mitigation server, an assessment of the network traffic in response to the attack mitigation message;

initiating, based on the assessment of the network traffic, an attack mitigation response;

determining that the attack detector associated with the attack mitigation message incorrectly assessed the network traffic based on the assessment of the network traffic;

receiving, at the attack detector, an update message relating to the attack detection model in response to the determination that the attack detector incorrectly assessed the network traffic; and

using, by the attack detector, data in the received update message to update the attack detection model.

2. The method of claim 1 , wherein the attack mitigation server uses a detection model different from the attack detection model of the attack detector.

3. The method of claim 1 , wherein the attack mitigation server receives data from other hosts in the network related to the network traffic.

4. The method of claim 1 , wherein the attack mitigation response comprises dropping one or more network packets associated with a DDoS attack.

5. The method of claim 1 , wherein the data in the update message comprises an indication that the attack detector incorrectly assessed the network traffic.

6. The method of claim 1 , wherein the data in the update message comprises one or more labels associated with the network traffic; and wherein the attack detector uses the data to retrain the attack detection model.

7. The method of claim 1 , wherein the data in the update message comprises one or more updated parameters for the attack detection model; and wherein the attack detector uses the data to update one or more parameters of the attack detection model.

8. The method of claim 1 , wherein the attack detector is a DDoS Open Threat Signaling (DOTS) client.

9. The method of claim 1 , wherein the attack mitigation server is a DOTS server.

10. The method of claim 1 , further comprising performing one or more operations at the attack detector to locally mitigate the potential DDoS attack.

11. A system comprising:

an attack detector, wherein the attack detector comprises:

one or more network interfaces to communicate with a network;

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

a memory configured to store a process that is executable by the processor, the process when executed causing the attack detector to:

monitor network traffic to detect a Distributed Denial of Service (DDoS) attack by applying a machine learning-based attack detection model against one or more attributes of the network traffic, and

responsive to detection by the attack detector of a potential DDoS attack, transmit an attack mitigation message indicating the potential DDoS attack to an attack mitigation server; and

the attack mitigation server, wherein the attack mitigation server comprises:

one or more network interfaces to communicate with a network;

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

a memory configured to store a process that is executable by the processor, the process when executed causing the attack mitigation server to:

cause an assessment of the network traffic in response to the attack mitigation message,

initiate, based on the assessment of the network traffic, an attack mitigation response, and

determine that the attack detector associated with the attack mitigation message incorrectly assessed the network traffic based on the assessment of the network traffic;

wherein the attack detector is further operable to:

receive an update message relating to the attack detection model in response to the determination that the attack detector incorrectly assessed the network traffic, and

use data in the received update message to update the attack detection model.

12. The system of claim 11 , wherein the attack mitigation server uses a detection model different from the attack detection model of the attack detector.

13. The system of claim 11 , wherein the attack mitigation server is further operable to receive data from other hosts in the network related to the network traffic.

14. The system of claim 11 , wherein the attack mitigation response comprises dropping one or more network packets associated with a DDoS attack.

15. The system of claim 11 , wherein the data in the update message comprises an indication that the attack detector incorrectly assessed the network traffic.

16. The system of claim 11 , wherein the data in the update message comprises one or more labels associated with the network traffic; and wherein the attack detector uses the data to retrain the attack detection model.

17. The system of claim 11 , wherein the data in the update message comprises one or more updated parameters for the attack detection model; and wherein the attack detector is further operable to use the data to update one or more parameters of the attack detection model.

18. The system of claim 11 , wherein the attack detector is a DDoS Open Threat Signaling (DOTS) client.

19. The system of claim 11 , wherein the attack mitigation server is a DOTS server.

20. The system of claim 11 , wherein the attack detector is further operable to perform one or more operations to locally mitigate the potential DDoS attack.