Methods and systems for anomaly detection using internet protocol (IP) traffic conversation data

1. A computer-based method for detecting anomalies in the traffic passing through an internet protocol (IP) network, said method comprising:

extracting, from a database, a single instance of each unique packet header associated with a plurality of IP-to-IP packets, the IP-to-IP packets having been transmitted across the IP network over a predefined period of time;

analyzing by a computer the packet headers to identify anomalous conversations based on at least one of a conversation uniqueness, a time of week uniqueness, and a data quantity uniqueness; and

providing by the computer alerts corresponding to detected anomalous conversations.

2. The method of claim 1 , wherein said analyzing the packet headers to identify anomalous conversations based on a conversation uniqueness comprises determining whether a particular conversation has been conducted on the IP network over a historical and statistically relevant period based on an IP-to-IP-on-Port conversation tuple.

3. The method of claim 2 , wherein said determining whether a particular conversation has been conducted on the IP network over a historical and statistically relevant period comprises determining whether the IP-to-IP-on-Port conversation tuple was previously stored in the database.

4. The method of claim 1 , wherein said analyzing the packet headers to identify anomalous conversations based on a time of week uniqueness comprises determining whether a particular conversation has been conducted on the IP network in a specific hour of the week over a historical and statistically relevant period.

5. The method of claim 4 , wherein said determining whether a particular conversation has been conducted on the IP network in a specific hour of the week over a historical and statistically relevant period comprises determining whether the IP-to-IP-on-Port conversation tuple was previously conducted at the specific hour of the week.

6. The method of claim 1 , wherein said analyzing the packet headers to identify anomalous conversations based on a data quantity uniqueness comprises determining whether a particular conversation conducted on the IP network has used a statistically different amount of bandwidth in a given hour of the week over a historical and statistically relevant period.

7. The method of claim 6 , wherein said determining whether a particular conversation conducted on the IP network has used a statistically different amount of bandwidth comprises:

summing the number of bytes transferred for the particular conversation in the last relevant period; and

comparing the sum against an average bandwidth extracted from the database.

8. The method of claim 1 , wherein said providing alerts corresponding to detected anomalous conversations comprises outputting relevant security events based on the analysis of the packet headers to a security event management system.

9. The method of claim 1 further comprising filtering out known frequently false alarming IP addresses and ports.

10. The method of claim 9 , wherein said filtering out known frequently false alarming IP addresses and ports comprises at least one of:

filtering out all conversations between specific pair of IP addresses;

filtering out all conversations where a specific IP and port are found;

filtering out all conversations where a single IP address is involved; and

filtering out all conversations that occur on a specific port.

11. The method of claim 1 , wherein said extracting a single instance of each unique packet header associated with a plurality of IP-to-IP packets, the IP-to-IP packets having been transmitted across the IP network over a predefined period of time comprises:

selecting all of the conversation records for the predefined time period where the selection data includes at least one of both IP addresses, the Internet protocol, the port, and the number of bytes recorded; and

summing the selected data by the conversation tuple IP-to-IP-on-Port.

12. A system, comprising:

a plurality of communicatively coupled computers configured to communicate via an internet protocol (IP) across a computer network; and

a database containing a single instance of each unique packet header associated with a plurality of IP-to-IP packets, the IP-to-IP packets having been transmitted across the computer network over a predefined period of time, one of said computers programmed to:

analyze the packet headers in said database to identify anomalous conversations based on at least one of a conversation uniqueness, a time of week uniqueness, and a data quantity uniqueness; and

provide at least one alert to a network user, the at least one alert corresponding to detected anomalous conversations.

13. The system of claim 12 , wherein to analyze the packet headers in said database to identify anomalous conversations, the one of said computers is programmed to determine whether a particular conversation has been conducted on the IP network over a historical and statistically relevant period based on an IP-to-IP-on-Port conversation tuple.

14. The system of claim 13 , wherein to determine whether a particular conversation has been conducted on the IP network over a historical and statistically relevant period, the one of said computers is programmed to determine whether the IP-to-IP-on-Port conversation tuple was previously stored in the database.

15. The system of claim 12 , wherein to analyze the packet headers to identify anomalous conversations based on a time of week uniqueness, the one of said computers is programmed to determine whether a particular conversation has been conducted on the IP network in a specific hour of the week over a historical and statistically relevant period.

16. The system of claim 15 , wherein to determine whether a particular conversation has been conducted on the IP network in a specific hour of the week over a historical and statistically relevant period, the one of said computers is programmed to determine whether the IP-to-IP-on-Port conversation tuple was previously conducted at the specific hour of the week.

17. The system of claim 12 , wherein to analyze the packet headers to identify anomalous conversations based on a data quantity uniqueness, the one of said computers is programmed to determine whether a particular conversation conducted on the IP network has used a statistically different amount of bandwidth in a given hour of the week over a historical and statistically relevant period.

18. The system of claim 17 , wherein to determine whether a particular conversation conducted on the IP network has used a statistically different amount of bandwidth, the one of said computers is programmed to:

sum the number of bytes transferred for the particular conversation in the last relevant period; and

compare the sum against an average bandwidth extracted from the database.

19. The system of claim 12 , wherein to provide alerts corresponding to detected anomalous conversations, the one of said computers is programmed to output relevant security events based on the analysis of the packet headers to a security event management system.

20. The system of claim 12 , wherein the one of said computers is programmed to filter out known frequently false alarming IP addresses and ports through at least one of:

filter out all conversations between specific pair of IP addresses;

filter out all conversations where a specific IP and port are found;

filter out all conversations where a single IP address is involved; and

filter out all conversations that occur on a specific port.

21. The system of claim 12 , wherein to extract a single instance of each unique packet header associated with a plurality of IP-to-IP packets, the IP-to-IP packets having been transmitted across the IP network over a predefined period of time, the one of said computers is programmed to:

select all of the conversation records for the predefined time period where the selection data includes at least one of both IP addresses, the Internet protocol, the port, and the number of bytes recorded; and

sum the selected data by the conversation tuple IP-to-IP-on-Port.

22. A method for detecting anomalies in network traffic, said method comprising:

receiving a single instance of each unique packet header associated with a plurality of IP-to-IP packets;

analyzing the packet headers by a computer to identify anomalous conversations based on at least one of:

a determination, based on an IP-to-IP-on-Port conversation tuple, of whether a particular conversation has been conducted on the IP network over a historical and statistically relevant period,

a determination whether a particular conversation has been conducted on the IP network in a specific hour of the week over a historical and statistically relevant period, and

a determination whether a particular conversation conducted on the IP network has used a statistically different amount of bandwidth in a given hour of the week over a historical and statistically relevant period; and

providing alerts corresponding to detected anomalous conversations.