Creating aggregate network flow time series in network anomaly detection systems

1. A computer implemented method for processing network flow data over a time series associated with a device-circuit pair, comprising:

receiving, by a server, flow data for one or more flows that correspond to the device-circuit pair;

calculating, by the server, a time difference for each flow of the one or more flows that correspond to the device-circuit pair, wherein calculating the time difference for each flow is based on a start time and an end time of each flow at one of the device-circuit pair and a file stamp time of a network flow record indicating each flow received by the server;

based on the calculated time differences and the received flow data, updating a probability distribution model associated with the device-circuit pair;

determining, by the server, whether a time bucket, of the time series, is complete or open based on the updated probability distribution model; and

when the time bucket is determined to be complete, determining, by the server, based on the received flow data corresponding to the time bucket of the time series, that a denial of service attack is occurring based on a number of network flows being too high for a specific time period.

2. The method of claim 1 , further comprising:

when the time bucket is determined to be complete, ignoring further flow data that corresponds to the time bucket; and

when the time bucket is determined to be open, incorporating further flow data that corresponds to the time bucket.

3. The method of claim 1 , wherein the probability distribution model comprises flow data that corresponds to the device-circuit pair and time differences for flows that correspond to the device-circuit pair.

4. The method of claim 3 , wherein the updating the probability distribution model comprises:

incorporating the received flow data and the calculated time differences into the probability distribution model;

calculating a mean value based on the time differences and the flow data included in the probability distribution model; and

calculating a standard deviation value based the time differences and the flow data included in the probability distribution model.

5. The method of claim 4 , wherein the determining whether the time bucket is complete or open comprises:

calculating a time delay value based on the standard deviation value; and

determining whether the time bucket is complete or open based on the time delay value and a file stamp time value of a network flow record containing the received flow data.

6. The method of claim 5 , wherein the calculating the time delay value comprises calculating the time delay value based on the standard deviation value and the mean value.

7. The method of claim 5 , wherein the determining whether the time bucket is complete or open comprises:

creating an expiry time based on an end time of the time bucket and the calculated time delay value;

determining that the time bucket is complete if the file stamp time is beyond the created expiry time; and

determining that the time bucket is open if the file stamp time is not beyond the created expiry time.

8. The method of claim 3 , wherein each of the time differences in the probability distribution model is a time difference between a start time of each flow in the probability distribution model and a file stamp time of a corresponding network flow record.

9. A system for processing network flow data over a time series associated with a device-circuit pair, comprising:

a memory; and

at least one processor coupled to the memory and configured to:

receive flow data for one or more flows that correspond to the device-circuit pair;

calculate a time difference for each flow of the one or more flows that correspond to the device-circuit pair, wherein calculating the time difference for each flow is based on a start time and an end time of each flow at one of the device-circuit pair and a file stamp time of a network flow record indicating each flow received by the server;

based on the calculated time differences and the received flow data, update a probability distribution model associated with the device-circuit pair;

determine whether a time bucket, of the time series, is complete or open based on the updated probability distribution model; and

when the time bucket is determined to be complete, determine, based on the received flow data corresponding to the time bucket of the time series, that a denial of service attack is occurring based on a number of network flows being too high for a specific time period.

10. The system of claim 9 , wherein the at least one processor is further configured to:

when the time bucket is determined to be complete, ignore further flow data that corresponds to the time bucket; and

when the time bucket is determined to be open, incorporate further flow data that corresponds to the time bucket.

11. The system of claim 9 , wherein the probability distribution model comprises flow data that corresponds to the device-circuit pair and time differences for flows that correspond to the device-circuit pair.

12. The system of claim 11 , wherein the at least one processor is further configured to update the probability distribution model by:

incorporating the received flow data and the calculated time differences into the probability distribution model;

calculating a mean value based on the time differences and the flow data included in the probability distribution model; and

calculating a standard deviation value based the time differences and the flow data included in the probability distribution model.

13. The system of claim 12 , wherein the at least one processor is further configured to determine whether the time bucket is complete or open by:

calculating a time delay value based on the standard deviation value; and

determining whether the time bucket is complete or open based on the time delay value and a file stamp time value of a network flow record containing the received flow data.

14. The system of claim 13 , wherein the at least one processor is further configured to calculate the time delay value by calculating the time delay value based on the standard deviation value and the mean value.

15. The system of claim 13 , wherein the at least one processor is further configured to determine whether the time bucket is complete or open by:

creating an expiry time based on an end time of the time bucket and the calculated time delay value;

determining that the time bucket is complete if the file stamp time is beyond the created expiry time; and

determining that the time bucket is open if the file stamp time is not beyond the created expiry time.

16. The system of claim 11 , wherein each of the time differences in the probability distribution model is a time difference between a start time of each flow in the probability distribution model and a file stamp time of a corresponding network flow record.

17. A non-transitory computer-readable medium having instructions stored thereon that, when executed by at least one computing device, causes the at least one computing device to perform operations for processing network flow data over a time series associated with a device-circuit pair, comprising:

receiving flow data for one or more flows that correspond to the device-circuit pair;

calculating a time difference for each flow of the one or more flows that correspond to the device-circuit pair, wherein calculating the time difference for each flow is based on a start time and an end time of each flow at one of the device-circuit pair and a file stamp time of a network flow record indicating each flow received by the server;

based on the calculated time differences and the received flow data, updating a probability distribution model associated with the device-circuit pair;

determining whether a time bucket, of the time series, is complete or open based on the updated probability distribution model; and

when the time bucket is determined to be complete, determining, based on the received flow data corresponding to the time bucket of the time series, that a denial of service attack is occurring based on a number of network flows being too high for a specific time period.

18. The non-transitory computer-readable medium of claim 17 , further comprising:

when the time bucket is determined to be complete, ignoring further flow data that corresponds to the time bucket; and

when the time bucket is determined to be open, incorporating further flow data that corresponds to the time bucket.