Detecting network anomalies by probabilistic modeling of argument strings with markov chains

• US 10,063,576 B2
• Filed: 12/29/2015
• Issued: 08/28/2018
• Est. Priority Date: 05/27/2008
• Status: Active Grant

• Current Assignee: Trustees Of Columbia University In The City Of New York (Columbia University)
• Original Assignee: Trustees Of Columbia University In The City Of New York (Columbia University)
• Inventors: Song, Yingbo, Keromytis, Angelos D., Stolfo, Salvatore J.
• Primary Examiner(s): Patel, Haresh N
• Application Number:
  US14/982,888
• Publication Number:
  US 20160366169A1
• Time in Patent Office: 973 Days
• Field of Search: 726 23
• CPC Class Codes: H04L 63/1416 Event detection, e.g. attack signature detection
• CPC Class Codes: H04L 63/1425 Traffic logging, e.g. anomaly detection
• CPC Class Codes: H04L 63/1466 Active attacks involving interception, injection, modification, spoofing of data unit addresses, e.g. hijacking, packet injection or TCP sequence number attacks
• CPC Class Codes: H04L 67/02 based on web technology, e.g. hypertext transfer protocol [HTTP]
• CPC Class Codes: H04L 69/16 Implementation or adaptation of Internet protocol [IP], of transmission control protocol [TCP] or of user datagram protocol [UDP]

Abstract

Systems, methods, and media for detecting network anomalies are provided. In some embodiments, a training dataset of communication protocol messages having argument strings is received. The content and structure associated with each of the argument strings is determined and a probabilistic model is trained using the determined content and structure of each of the argument strings. A communication protocol message having an argument string that is transmitted from a first processor to a second processor across a computer network is received. The received communication protocol message is compared to the probabilistic model and then it is determined whether the communication protocol message is anomalous.

Claims

1. A method for detecting network anomalies, the method comprising:

receiving, using a hardware processor, a communication protocol message having an argument string that is transmitted from a first processor to a second processor across a computer network;

applying, using the hardware processor, a probabilistic model to the received communication protocol message to determine whether the communication protocol message is anomalous based on determining that at least one n-gram in the communication protocol message is anomalous, wherein the probabilistic model uses at least one Markov chain specified by one or more parameters to determine a probability that the argument string is anomalous based on n-grams in the argument string; and

performing, using the hardware processor, a predetermined action in response to determining that the communication protocol message is anomalous.

2. The method of claim 1, wherein the one or more parameters includes gram size, and wherein the probabilistic model was trained based on the gram size.

3. The method of claim 1, wherein the one or more parameters includes mixture size that specifies a number of Markov chains to use in the probabilistic model and wherein the at least one Markov chain is a mixture of Markov chains specified by the mixture size.

4. The method of claim 1, further comprising:

receiving a plurality of packets transmitted from the first processor to the second processor across the computer network;

assembling the plurality of packets to form the communication protocol message; and

extracting the argument string from the communication protocol message.

5. The method of claim 1, wherein the communication protocol message is a Hypertext Transfer Protocol (HTTP) request message.

6. The method of claim 1, wherein the probabilistic model was trained based on content and structure of the argument string included in each of a plurality of communication protocol messages.

7. The method of claim 6, wherein the content and the structure associated with each of the argument strings further comprises one or more variable names, a corresponding argument value for each of the variable names, and a layout with respect to each of the variable names.

8. The method of claim 1, wherein the at least one Markov chain comprises one or more Markov models of transition probabilities in n-grams of the communication protocol message.

9. The method of claim 1, wherein the predetermined action comprises issuing an alert.

10. The method of claim 1, wherein applying the probabilistic model further comprises calculating a normality score for the received communication protocol message using the probabilistic model.

11. A system for detecting network anomalies, the system comprising:

a memory; and

a hardware processor that, when executing computer-executable instructions stored in the memory, is configured to;

receive a communication protocol message having an argument string that is transmitted from a first processor to a second processor across a computer network;

apply a probabilistic model to the received communication protocol message to determine whether the communication protocol message is anomalous based on determining that at least one n-gram in the communication protocol message is anomalous, wherein the probabilistic model uses at least one Markov chain specified by one or more parameters to determine a probability that the argument string is anomalous based on n-grams in the argument string; and

perform a predetermined action in response to determining that the communication protocol message is anomalous.

12. The system of claim 11, wherein the one or more parameters includes gram size, and wherein the probabilistic model was trained based on the gram size.

13. The system of claim 11, wherein the one or more parameters includes mixture size that specifies a number of Markov chains to use in the probabilistic model and wherein the at least one Markov chain is a mixture of Markov chains specified by the mixture size.

14. The system of claim 11, wherein the hardware processor is further configured to:

receive a plurality of packets transmitted from the first processor to the second processor across the computer network;

assemble the plurality of packets to form the communication protocol message; and

extract the argument string from the communication protocol message.

15. The system of claim 11, wherein the communication protocol message is a Hypertext Transfer Protocol (HTTP) request message.

16. The system of claim 11, wherein the probabilistic model was trained based on content and structure of the argument string included in each of a plurality of communication protocol messages.

17. The system of claim 16, wherein the content and the structure associated with each of the argument strings further comprises one or more variable names, a corresponding argument value for each of the variable names, and a layout with respect to each of the variable names.

18. The system of claim 11, wherein the at least one Markov chain comprises one or more Markov models of transition probabilities in n-grams of the communication protocol message.

19. The system of claim 11, wherein the predetermined action comprises issuing an alert.

20. The system of claim 11, wherein the hardware processor further configured to calculate a normality score for the received communication protocol message using the probabilistic model.

21. A non-transitory computer-readable medium containing computer-executable instructions that, when executed by a processor, cause the processor to perform method for detecting network anomalies, the method comprising:

receiving a communication protocol message having an argument string that is transmitted from a first processor to a second processor across a computer network;

applying a probabilistic model to the received communication protocol message to determine whether the communication protocol message is anomalous based on determining that at least one n-gram in the communication protocol message is anomalous, wherein the probabilistic model uses at least one Markov chain specified by one or more parameters to determine a probability that the argument string is anomalous based on n-grams in the argument string; and

performing a predetermined action in response to determining that the communication protocol message is anomalous.