Method and system for real time detection of threats in high volume data streams
View Patent ↗A high speed detection system and method capable of generating audits of investigable patterns from log data using techniques for grouping and filtering the data so as to create vectors of patterns which can be then further analyzed by applying conditional filters to conclude that a threat may be active has been created to solve at least the above discussed problems.
1. A high speed threat detection system, comprising:
a plurality of network sources to generate network data comprising a plurality of raw events; and
a data analysis unit to receive the network data from the network sources, the data analysis unit comprising
a slicing unit to package the network data into a plurality of slices, wherein each of the slices comprises a predetermined chunk of the network data,
a relational database engine to
organize the slices into a plurality of windows, wherein each window is associated with a distinct pattern, comprises a rolling set of the slices that match a data relationship requirement for the distinct pattern associated with the particular window, and comprises a fixed time boundary over which the slices enter and exit the particular window and over which the distinct pattern associated with the particular window can emerge from the rolling set of slices, and
analyze, for each window, the corresponding rolling set of slices to detect an instance of the pattern associated with the particular window, and
a pattern identification unit to identify which of the detected pattern instances are to be reviewed by an operator, and
an output device to output the pattern instances to be reviewed by the operator.
2. The system of claim 1 , further comprising a memory to store the detected pattern instances.
3. The system of claim 1 , wherein the output device is a computer monitor.
4. The system of claim 1 , wherein at least one of the network sources is a firewall.
5. The system of claim 1 , wherein the relational database engine analyzes the rolling sets of slices based on an event filter.
6. The system of claim 1 , wherein the relational database engine analyzes the rolling sets of slices based on rate-detection.
7. The system of claim 1 , wherein the relational database engine analyzes the rolling sets of slices based on rare value detection.
8. The system of claim 1 , wherein the relational database engine analyzes the rolling sets of slices using a pattern filter.
9. The system of claim 1 , wherein the relational database engine analyzes the rolling sets of slices in real time.
10. The system of claim 1 , wherein the relational database engine analyzes the rolling sets of slices in polynomial time.
11. The system of claim 1 , wherein the data analysis unit is a processor.
12. The system of claim 1 , wherein the slices are implemented in one of a table in a memory or a memory based file.
13. A method of detecting network threats, comprising:
receiving network data from a plurality of network sources;
loading the network data into a plurality of slices, each of the slices comprising a predetermined chunk of the network data;
organizing the slices into a plurality of windows, wherein each window is associated with a distinct pattern, comprises a rolling set of the slices that match a data relationship requirement for the distinct pattern associated with the particular window, and comprises a fixed time boundary over which the slices enter and exit the particular window and over which the distinct pattern associated with the particular window can emerge from the rolling set of slices;
determining if an instance of the pattern associated with each of the windows is present in the corresponding rolling set of slices associated with each particular window; and
storing determined pattern instances.
14. The method of claim 13 , further comprising applying conditionals to the stored pattern instances to determine if a threat situation can be concluded.
15. The method of claim 13 , wherein each of the slices comprises an amount of data associated with a common predetermined period of time.
16. The method of claim 15 , wherein the predetermined period of time is one minute.
17. The method of claim 13 , wherein the network data is loaded into the slices so as to collate time-related data together.
18. The method of claim 13 , wherein at least two of the windows are of different sizes.
19. The method of claim 13 , wherein the slices may be incorporated into or removed from a window based on time.
20. The method of claim 13 , wherein a slice is deleted when it exits a window.
21. The method of claim 13 , wherein the determining if an instance of the pattern associated with each of the windows is present comprises applying at least one filter to each window to determine if an instance of the pattern is present.
22. The method of claim 13 , further comprising archiving all of the slices to aid in forensic analysis or reporting.
23. The method of claim 13 , further comprising persisting at least one audit, wherein the audit comprises at least one pattern instance.
24. The method of claim 13 , further comprising deleting all slices which do not contain pattern related information.
25. The method of claim 13 , wherein at least one of the network sources is a networked source.
26. The method of claim 13 , wherein the plurality of windows comprises at least one of a time based window, a value based window, or a count based window.
27. The method of claim 13 , wherein the slices are implemented as one of a table in a memory or a memory based file.
28. A non-transitory machine readable storage medium storing a program, wherein when the program is executed it performs a method comprising:
receiving network data from a plurality of network sources;
loading the network data into a plurality of slices, each of the slices comprising a predetermined chunk of the network data;
organizing the slices into a plurality of windows, wherein each window is associated with a distinct pattern, comprises a rolling set of the slices that match a data relationship requirement for the distinct pattern associated with the particular window, and comprises a fixed time boundary over which the slices enter and exit the particular window and over which the distinct pattern associated with the particular window can emerge from the rolling set of slices;
determining if an instance of the pattern associated with each of the windows is present in the corresponding rolling set of slices associated with each particular window; and
storing determined pattern instances in a pattern store.