System and method for keyword spotting using representative dictionary

• US 9,639,520 B2
• Filed: 01/29/2014
• Issued: 05/02/2017
• Est. Priority Date: 01/29/2013
• Status: Active Grant

• Current Assignee: Cognyte Technologies Israel Ltd. (Cognyte Software Ltd.)
• Original Assignee: Verint Systems Limited (Verint Systems Incorporated)
• Inventors: Yishay, Yitshak
• Primary Examiner(s): Yen, Eric
• Application Number:
  US14/167,052
• Publication Number:
  US 20140214407A1
• Time in Patent Office: 1,189 Days
• CPC Class Codes: G06F 16/334 Query execution G06F16/335 takes precedence
• CPC Class Codes: G06F 16/90344 by using string matching techniques
• CPC Class Codes: G06F 21/55 Detecting local intrusion or implementing counter-measures
• CPC Class Codes: G06F 40/242 Dictionaries
• CPC Class Codes: G06F 40/289 Phrasal analysis, e.g. finite state techniques or chunking

Abstract

Methods and systems for keyword spotting, i.e., for identifying textual phrases of interest in input data. In the embodiments described herein, the input data comprises communication packets exchanged in a communication network. The disclosed keyword spotting techniques can be used, for example, in applications such as Data Leakage Prevention (DLP), Intrusion Detection Systems (IDS) or Intrusion Prevention Systems (IPS), and spam e-mail detection. A keyword spotting system holds a dictionary of textual phrases for searching input data. In a communication analytics system, for example, the dictionary defines textual phrases to be located in communication packets—such as e-mail addresses or Uniform Resource Locators (URLs).

Claims

1. A method comprising operations performed by a computer, the operations comprising:

holding a first dictionary comprising first textual phrases for searching in data;

deriving from the first dictionary a second dictionary, which comprises second textual phrases and has a smaller data size than the first dictionary, such that occurrence of any of the first textual phrases in the data corresponds to the occurrence of at least one of the second textual phrases in the data;

searching input data with the second dictionary; and

in response to identifying in the input data a second textual phrase from the second dictionary, locating in the input data a first textual phrase from the first dictionary corresponding to the identified second textual phrase.

2. The method according to claim 1, wherein the second dictionary is derived such that any occurrence of one of the second textual phrases indicates the occurrence of one of the first textual phrases with a likelihood higher than a predefined threshold.

3. The method according to claim 1, wherein holding the first dictionary comprises storing the first dictionary in a first memory having a first access speed, and wherein deriving the second dictionary comprises storing the second dictionary in a second memory having a second access speed that is higher than the first access speed.

4. The method according to claim 3, wherein searching the input data comprises processing the input data using a processor, and wherein the second memory comprises a cache memory of the processor.

5. The method according to claim 1, wherein identifying the second textual phrase comprises specifying a partial subset of the first textual phrases based on the identified second textual phrase, and wherein locating the first textual phrase comprises searching only the partial subset of the first textual phrases.

6. The method according to claim 1, wherein identifying the second textual phrase comprises indicating a position in the input data where the first textual phrase is expected, and wherein locating the first textual phrase comprises accessing the input data in a vicinity of the indicated location.

7. The method according to claim 1, wherein deriving the second dictionary comprises excluding from the second dictionary one or more second textual phrases that are expected to cause false detections while searching the input data with the second dictionary.

8. The method according to claim 1, wherein deriving the second dictionary comprises defining one or more lists of the second textual phrases, such that, in each list, the second textual phrases are derived as the sub-strings of respective ones of the first textual phrases that begin at a given constant position in the first textual phrases and have a given constant length.

9. The method according to claim 8, wherein one or more of the first textual phrases comprise wildcard characters, and wherein defining the lists comprises choosing the lists such that the second textual phrases do not comprise any of the wildcard characters.

10. The method according to claim 8, wherein searching the input data with the second dictionary comprises applying a hash function to a portion of the input data, and comparing a result of the hash function to the sub-strings in a given list.

11. Apparatus, comprising:

a memory, which is configured to hold a first dictionary comprising first textual phrases for searching in data; and

a processor, which is configured to derive from the first dictionary a second dictionary, which comprises second textual phrases and has a smaller data size than the first dictionary, such that occurrence of any of the first textual phrases in the data corresponds to the occurrence of at least one of the second textual phrases in the data, to search input data with the second dictionary, and, in response to identifying in the input data a second textual phrase from the second dictionary, to locate in the input data a first textual phrase from the first dictionary corresponding to the identified second textual phrase.

12. The apparatus according to claim 11, wherein the processor is configured to derive the second dictionary such that any occurrence of one of the second textual phrases indicates the occurrence of one of the first textual phrases with a likelihood higher than a predefined threshold.

13. The apparatus according to claim 11, wherein the memory holding the first dictionary comprises a first memory having a first access speed, and comprising a second memory, which is configured to hold the second dictionary and which has a second access speed that is higher than the first access speed.

14. The apparatus according to claim 13, wherein the second memory comprises a cache memory of the processor.

15. The apparatus according to claim 11, wherein the processor is configured to specify a partial subset of the first textual phrases based on the identified second textual phrase, and to locate the first textual phrase by searching only the partial subset of the first textual phrases.

16. The apparatus according to claim 11, wherein, based on identifying the second textual phrase, the processor is configured to indicate a position in the input data where the first textual phrase is expected, and to locate the first textual phrase by accessing the input data in a vicinity of the indicated location.

17. The apparatus according to claim 11, wherein the processor is configured to exclude from the second dictionary one or more second textual phrases that are expected to cause false detections while searching the input data with the second dictionary.

18. The apparatus according to claim 11, wherein the processor is configured to derive the second dictionary by defining one or more lists of the second textual phrases, such that, in each list, the second textual phrases are derived as the sub-strings of respective ones of the first textual phrases that begin at a given constant position in the first textual phrases and have a given constant length.

19. The apparatus according to claim 18, wherein one or more of the first textual phrases comprise wildcard characters, and wherein the processor is configured to define the lists such that the second textual phrases do not comprise any of the wildcard characters.

20. The apparatus according to claim 18, wherein the processor is configured to search the input data with the second dictionary by applying a hash function to a portion of the input data, and comparing a result of the hash function to the sub-strings in a given list.