Method and apparatus for analysis of electronic communications containing imagery

• US 20050216564A1
• Filed: 08/24/2004
• Published: 09/29/2005
• Est. Priority Date: 03/11/2004
• Status: Abandoned Application

• Current Assignee: SRI International, Inc.
• Original Assignee: SRI International, Inc.
• Inventors: Marcotullio, John P., Aradhye, Hrishikesh B., Myers, Gregory K., Mulgaonkar, Prasanna
• Application Number:
  US10/925,335
• Publication Number:
  US 20050216564A1
• Time in Patent Office: Days
• US Class Current: 709/206
• CPC Class Codes: G06V 2201/01 Solutions for problems related to non-uniform document background
• CPC Class Codes: G06V 30/10 Character recognition
• CPC Class Codes: G06V 30/413 Classification of content, e.g. text, photographs or tables
• CPC Class Codes: H04L 51/212 using filtering or selective blocking

Abstract

A method and apparatus are provided for analyzing an electronic communication containing imagery, e.g., to determine whether or not the electronic communication is a spam communication. In one embodiment, an inventive method includes detecting one or more regions of imagery in a received electronic communication and applying pre-processing techniques to locate regions (e.g., blocks or lines) of text in the imagery that may be distorted. The method then analyzes the regions of text to determine whether the content of the text indicates that the electronic communication is spam. In one embodiment, specialized extraction and rectification of embedded text followed by optical character recognition processing is applied to the regions of text to extract their content therefrom. In another embodiment, keyword recognition or shape-matching processing is applied to detect the presence or absence of spam-indicative words from the regions of text. In another embodiment, other attributes of extracted text regions, such as size, location, color and complexity are used to build evidence for or against the presence of spam.

Claims

1. A method for categorizing an electronic communication containing imagery, the method comprising the steps of:

locating portions of said imagery having text regions therein; and

analyzing said text regions to determine whether content of said text regions indicates that said electronic communication is likely to be unsolicited or unauthorized.

2. The method of claim 1, wherein said locating step comprises:

locating text regions that are distorted.

3. The method of claim 2, wherein distorted text regions comprise text regions that are superimposed over complex backgrounds, that include skewed text, or both.

4. The method of claim 1, wherein said analyzing step comprises:

identifying one or more words contained in said text regions; and

determining whether one or more of the identified words is a trigger word that indicates unsolicited and/or unauthorized information.

5. The method of claim 4, wherein said determining step comprises:

designating an identified word as a trigger word if said identified word substantially matches one or more words in a pre-defined library of trigger words.

6. The method of claim 5, wherein said designating step comprises:

applying a text-based spam identification tool to compare said identified word to words in said pre-defined library.

7. The method of claim 4, further comprising the step of:

designating said electronic communication as unsolicited and/or unauthorized if an occurrence of trigger words contained in said imagery satisfies a pre-defined criterion.

8. The method of claim 7, wherein said pre-defined criterion is a user-definable threshold defining a maximum acceptable quantity of trigger words for said imagery.

9. The method of claim 7, wherein said designating step comprises:

assigning a score to one or more identified words or phrases in said imagery, wherein said score indicates a likelihood that said identified words or phrases indicate that said electronic communication is unsolicited or unauthorized; and

concluding that said electronic communication is unsolicited and/or unauthorized if an aggregate score for said electronic communication exceeds a maximum acceptable score.

10. The method of claim 9, wherein said aggregate score is the sum of one or more scores for corresponding identified trigger words contained in one or more imagery elements in said electronic communication.

11. The method of claim 4, wherein said identifying step comprises:

applying optical character recognition (OCR) processing to said text regions to identify one or more words contained therein.

12. The method of claim 4, wherein said identifying step comprises:

applying keyword recognition processing to said text regions to identify one or more words contained therein.

13. The method of claim 12, wherein said keyword recognition processing comprises:

comparing the shape of at least a portion of a text region to the shapes of one or more keywords in a pre-defined keyword library; and

identifying said at least a portion of a text region as a trigger word if the shape of said at least a portion of a text region substantially matches the shape of one or more words contained in said keyword library.

14. The method of claim 12, wherein said keyword recognition processing comprises:

matching one or more features located in a text region to a hidden Markov model representing a keyword contained in a pre-defined keyword library; and

identifying said features as belonging to a trigger word.

15. A computer readable medium containing an executable program for categorizing an electronic communication containing imagery, where the program performs the steps of:

locating portions of said imagery having text regions therein; and

analyzing said text regions to determine whether content of said text regions indicates that said electronic communication is likely to be unsolicited or unauthorized.

16. The computer readable medium of claim 15, wherein said locating step comprises:

locating text regions that are distorted.

17. The computer readable medium of claim 16, wherein distorted text regions comprise text regions that are superimposed over complex backgrounds, that include skewed text, or both.

18. The computer readable medium of claim 15, wherein said analyzing step comprises:

identifying one or more words contained in said text regions; and

determining whether one or more of the identified words is a trigger word that indicates unsolicited and/or unauthorized information.

19. The computer readable medium of claim 18, wherein said determining step comprises:

designating an identified word as a trigger word if said identified word substantially matches one or more words in a pre-defined library of trigger words.

20. The computer readable medium of claim 19, wherein said designating step comprises:

applying a text-based spam identification tool to compare said identified word to words in said pre-defined library.

21. The computer readable medium of claim 18, further comprising the step of:

designating said electronic communication as unsolicited and/or unauthorized if an occurrence of identified trigger words contained in said imagery satisfies a pre-defined criterion.

22. The computer readable medium of claim 21, wherein said pre-defined criterion is a user-definable threshold defining a maximum acceptable quantity of trigger words for said imagery.

23. The computer readable medium of claim 21, wherein said designating step comprises:

assigning a score to one or more identified words or phrases in said imagery, wherein said score indicates the likelihood that said identified words or phrases indicate that said electronic communication is unsolicited or unauthorized; and

concluding that said electronic communication is unsolicited and/or unauthorized if an aggregate score for said electronic communication exceeds a maximum acceptable score.

24. The computer readable medium of claim 21, wherein said aggregate score is the sum of one or more scores for corresponding identified trigger words contained in one or more imagery elements in said electronic communication.

25. The computer readable medium of claim 18, wherein said identifying step comprises:

applying optical character recognition (OCR) processing to said text regions to identify one or more words contained therein.

26. The computer readable medium of claim 18, wherein said identifying step comprises:

applying keyword recognition processing to said text regions to identify one or more words contained therein.

27. The computer readable medium of claim 26, wherein said keyword recognition processing comprises:

comparing the shape of at least a portion of a text region to the shapes of one or more keywords in a pre-defined keyword library; and

identifying said at least a portion of a text region as a trigger word if the shape of said at least a portion of a text region substantially matches the shape of one or more words contained in said keyword library.

28. The computer readable medium of claim 15, wherein said keyword recognition processing comprises:

matching one or more features located in a text region to a hidden Markov model representing a keyword contained in a pre-defined keyword library; and

identifying said features as belonging to a trigger word.

29. Apparatus for categorizing an electronic communication containing imagery, the apparatus comprising:

means for locating portions of said imagery having text regions therein; and

means for analyzing said text regions to determine whether content of said text regions indicates that said electronic communication is unsolicited and/or unauthorized.

30. A method for categorizing an electronic communication containing imagery, the method comprising the steps of:

applying pre-processing techniques to said imagery in order to locate regions of text in said imagery;

measuring one or more characteristics of sets of image pixels within said regions of text; and

determining if one or more measured characteristics indicates that said electronic communication is likely to be unsolicited or unauthorized.

31. The method of claim 30, wherein said characteristics to be measured are one or more of the following:

text superimposition over said imagery, distribution of colors in said imagery, distribution of intensity in said imagery, a number of text regions, positions of text regions, sizes of text regions, fonts used in text regions, the presence of random noise or distorting or interfering patterns, text overlap, text distortion and the presence of cursive text.

32. The method of claim 30, wherein said one or more measured characteristics indicate that said electronic communication is likely to be unsolicited or unauthorized if attributes of said characteristics are common in unsolicited or unauthorized communications but not common in legitimate electronic communications.

33. The method of claim 32, further comprising the step of:

concluding that said electronic communication is unsolicited or unauthorized if the incidence of characteristics indicating that said electronic communication is likely to be unsolicited or unauthorized satisfies a pre-defined criterion.

34. The method of claim 33, wherein characteristics indicating that said electronic communication is likely to be unsolicited or unauthorized are assigned a score associated with a degree of likelihood that the presence of said characteristics indicates that said electronic communication is in fact unsolicited or unauthorized.

35. The method of claim 34, wherein said pre-defined criterion is a maximum acceptable score representing an aggregate of scores of said characteristics.

36. The method of claim 30, wherein said pre-processing techniques comprise:

locating regions of text in said imagery that are superimposed over complex backgrounds, that are distorted, or both.

37. A computer readable medium containing an executable program for categorizing an electronic communication containing imagery, where the program performs the steps of:

applying pre-processing techniques to said imagery in order to locate regions of text in said imagery;

measuring one or more characteristics of sets of image pixels within said regions of text; and

determining if one or more measured characteristics indicates that said electronic communication is likely to be unsolicited or unauthorized.

38. The computer readable medium of claim 37, wherein said characteristics to be measured are one or more of the following:

text superimposition over said imagery, distribution of colors in said imagery, distribution of intensity in said imagery, positions of text regions, sizes of text regions, fonts used in text regions, the presence of random noise, text overlap text, text distortion and the presence of cursive text.

39. The computer readable medium of claim 37, wherein said one or more measured characteristics indicate that said electronic communication is determining if one or more measured characteristics indicates that said electronic communication is likely to be unsolicited or unauthorized if attributes of said characteristics are common in unsolicited or unauthorized communications but not common in legitimate electronic communications.

40. The computer readable medium of claim 39, further comprising the step of:

concluding that said electronic communication is unsolicited or unauthorized if the incidence of characteristics indicating that said electronic communication is determining if one or more measured characteristics indicates that said electronic communication is likely to be unsolicited or unauthorized satisfies a pre-defined criterion.

41. The computer readable medium of claim 40, wherein characteristics indicating that said electronic communication is determining if one or more measured characteristics indicates that said electronic communication is likely to be unsolicited or unauthorized are assigned a score associated with a degree of likelihood that said characteristics indicate that said electronic communication is in fact unsolicited or unauthorized.

42. The computer readable medium of claim 41, wherein said pre-defined criterion is a maximum acceptable score representing an aggregate of scores of said characteristics.

43. The computer readable medium of claim 37, wherein said pre-processing techniques comprise:

locating regions of text in said imagery that are superimposed over complex backgrounds, that are distorted, or both.

44. Apparatus for categorizing an electronic communication containing imagery, the apparatus comprising:

means for applying pre-processing techniques to said imagery in order to locate regions of text in said imagery;

means for measuring one or more characteristics of sets of image pixels within said regions of text; and

means for determining if one or more measured characteristics indicates that said electronic communication is likely to be unsolicited or unauthorized.