Network endpoint profiling using a topical model and semantic analysis

1. A method comprising:

extracting, by a device in a network, words from traffic data for a particular endpoint node in the network;

determining, by the device, one or more topical categories associated with the particular endpoint node by applying a machine learning-based topical model to the extracted words;

identifying, by the device, one or more similar endpoint nodes in the network based on a similarity between the determined one or more topical categories associated with the particular endpoint node and one or more topical categories associated with the one or more similar endpoint nodes;

identifying, by the device, a device type of the one or more similar endpoint nodes;

determining, by the device, a device type for the particular endpoint node that corresponds to the device type of the one or more similar endpoint nodes; and

controlling, by the device, access to content or resources in the network for the particular endpoint node based on the determined device type for the particular endpoint node.

2. The method as in claim 1 , wherein extracting the words from the traffic data for the particular endpoint node comprises:

causing, by the device, one or more networking devices in the network to extract the words; and

receiving, at the device, the extracted words from the one or more networking devices.

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

capturing, by the device, the traffic data for the particular endpoint node.

4. The method as in claim 1 , wherein determining the one or more topical categories associated with the particular endpoint node by applying a machine learning-based topical model to the extracted words comprises:

applying, by the device, a histogram to the extracted words using a nave Bayesian approach, to determine the one or more topical categories.

5. The method as in claim 4 , wherein determining the one or more topical categories associated with the particular endpoint node further comprises:

clustering, by the device, the extracted words using semantic information from a dictionary.

6. The method as in claim 1 , wherein identifying the one or more similar endpoint nodes comprises:

computing, by the device, a similarity score between the one or more topical categories associated with the particular endpoint node and the one or more topical categories associated with the one or more similar endpoint nodes.

7. The method as in claim 6 , wherein the similarity score is based on a Fisher kernel.

8. The method as in claim 1 , wherein the traffic data for the particular endpoint node comprises data regarding an encrypted traffic flow associated with the particular endpoint node.

9. An apparatus, comprising:

one or more network interfaces to communicate with a network;

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

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

extract words from traffic data for a particular endpoint node in the network;

determine one or more topical categories associated with the particular endpoint node by applying a machine learning-based topical model to the extracted words;

identify one or more similar endpoint nodes in the network based on a similarity between the determined one or more topical categories associated with the particular endpoint node and one or more topical categories associated with the one or more similar endpoint nodes;

identify a device type of the one or more similar endpoint nodes;

determine a device type for the particular endpoint node that corresponds to the device type of the one or more similar endpoint nodes; and

control access to content or resources in the network for the particular endpoint node based on the determined device type for the particular endpoint node.

10. The apparatus as in claim 9 , wherein the apparatus extracts the words from the traffic data for the particular endpoint node by:

causing one or more networking devices in the network to extract the words; and

receiving the extracted words from the one or more networking devices.

11. The apparatus as in claim 9 , wherein the process when executed is further operable to:

capture the traffic data for the particular endpoint node.

12. The apparatus as in claim 9 , wherein the apparatus determines the one or more topical categories associated with the particular endpoint node by:

applying a histogram to the extracted words using a nave Bayesian approach, to determine the one or more topical categories.

13. The apparatus as in claim 12 , wherein the apparatus further determines the one or more topical categories associated with the particular endpoint node by:

clustering the extracted words using semantic information from a dictionary.

14. The apparatus as in claim 9 , wherein the apparatus identifies the one or more similar endpoint nodes by:

computing a similarity score between the one or more topical categories associated with the particular endpoint node and the one or more topical categories associated with the one or more similar endpoint nodes.

15. The apparatus as in claim 14 , wherein the similarity score is based on a Fisher kernel.

16. The apparatus as in claim 9 , wherein the traffic data for the particular endpoint node comprises data regarding an encrypted traffic flow associated with the particular endpoint node.

17. A tangible, non-transitory, computer-readable medium storing program instructions that, when executed by a device in a network, cause the device to perform a process comprising:

extracting, by the device, words from traffic data for a particular endpoint node in the network;

determining, by the device, one or more topical categories associated with the particular endpoint node by applying a machine learning-based topical model to the extracted words;

identifying, by the device, one or more similar endpoint nodes in the network based on a similarity between the determined one or more topical categories associated with the particular endpoint node and one or more topical categories associated with the one or more similar endpoint nodes;

identifying, by the device, a device type of the one or more similar endpoint nodes;

determining, by the device, a device type for the particular endpoint node that corresponds to the device type of the one or more similar endpoint nodes; and

controlling, by the device, access to content or resources in the network for the particular endpoint node based on the determined device type for the particular endpoint node.

18. The computer-readable medium as in claim 17 , wherein the device determines the one or more topical categories associated with the particular endpoint node by:

applying, by the device, a histogram to the extracted words using a naïve Bayesian approach, to determine the one or more topical categories.