Attack mitigation using learning machines

1. A method, comprising:

receiving network traffic data for a computer network;

predicting a probability that one or more nodes are under attack based on the network traffic data;

deciding to mitigate a predicted attack by instructing nodes to reroute network traffic around the one or more nodes under attack on an alternative route without altering an existing routing topology of the computer network, wherein the nodes determine the alternate route from examined alternate disjoint paths based on a policy for class of service of the rerouted traffic, an increase in cost of the alternative route and a probability; and, in response,

communicating an attack notification message to the one or more nodes under attack.

2. The method as in claim 1 , wherein predicting further comprises:

using a machine learning model on the network traffic data to predict the probability that the one or more nodes are under attack.

3. The method as in claim 1 , wherein the attack notification message is communicated using a layer-2 broadcast slot.

4. The method as in claim 1 , wherein the attack notification message is communicated using a known multicast layer-3 address encapsulated in a broadcast layer-2 frame.

5. The method as in claim 4 , wherein the encapsulated attack notification message includes a list of one or more nodes predicted to be under attack associated with probabilities of attack for each of the one or more nodes predicted to be under attack.

6. The method as in claim 1 , wherein the one or more nodes predicted to be under attack, in response to receiving the attack notification message, communicate an attack mitigation reroute request to one or more child nodes.

7. The method as in claim 6 , wherein the attack mitigation reroute request provides instructions to the one or more child nodes to search for alternate routing paths without altering existing network topology.

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

receiving real-time network traffic data from the one or more nodes under attack;

determining, based on the real-time network traffic data from the one or more nodes under attack, that the attack is over; and

communicating a notification message to the one or more nodes to stop communicating attack mitigation reroute requests.

9. A method, comprising:

receiving, at a parent node in a computer network, an attack-mitigation reroute request on a layer-2 broadcast slot; and, in response,

communicating an attack mitigation reroute request to one or more child nodes of the parent node to reroute network traffic around the one or more child node under attack on one or more alternative routes, wherein the parent node determines the alternative route from examined alternate disjoint paths based on a policy for class of service of the rerouted traffic, an increase in cost of the alternative route and a probability that one or more child nodes are under attack.

10. The method as in claim 9 , wherein the one or more alternate routes do not alter existing network topology.

11. The method as in claim 9 , wherein the attack mitigation reroute request notifies the one or more child nodes of the probability that the parent node is under attack.

12. The method as in claim 11 , wherein the attack mitigation reroute request instructs, based on the probability that the parent node is under attack, the one or more child nodes to communicate a routing advertisement update.

13. An apparatus, comprising:

one or more network interfaces to communicate within a computer network;

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

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

receive network traffic data for a computer network;

predict a probability that one or more nodes are under attack based on the network traffic data;

decide to mitigate a predicted attack by instructing nodes to reroute network traffic around the one or more nodes under attack on an alternative route without altering an existing routing topology of the computer network, wherein the nodes determine the alternate route from examined alternate disjoint paths based on a policy for class of service of the rerouted traffic, an increase in cost of the alternate route and the probability; and, in response,

communicate an attack notification message to the one or more nodes under attack.

14. The apparatus as in claim 13 , wherein the process when executed is further operable to:

use a machine learning model on the network traffic data to predict the probability that the one or more nodes are under attack.

15. The apparatus as in claim 14 , wherein the machine learning model is hosted on a node selected from the group consisting of a capable node, a network management server (NMS), and a field area router (FAR).

16. The apparatus as in claim 13 , wherein the attack notification message is communicated using a layer-2 broadcast slot.

17. The apparatus as in claim 13 , wherein the attack notification message is communicated using a known multicast layer-3 address encapsulated in a broadcast layer-2 frame.

18. The apparatus as in claim 13 , wherein the encapsulated attack notification message includes a list of one or more nodes predicted to be under attack associated with probabilities of attack for each of the one or more nodes predicted to be under attack.

19. The apparatus as in claim 13 , wherein the one or more nodes predicted to be under attack, in response to receiving the attack notification message, communicate an attack mitigation reroute request to one or more child nodes.

20. The apparatus as in claim 13 , wherein the attack mitigation reroute request provides instructions to the one or more child nodes to search for alternate routing paths without altering existing network topology.