Systems and methods for embedding media forensic identification markings

• US 20050154892A1
• Filed: 01/09/2004
• Published: 07/14/2005
• Est. Priority Date: 01/09/2004
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Venkatesan, Ramarathnam, Mihcak, Mehmet Kivanc, Kucukgoz, Mehmet
• Granted Patent: US 8,032,754 B2
• Time in Patent Office: Days
• US Class Current: 713/176
• CPC Class Codes: G06T 1/005 Robust watermarking, e.g. average attack or collusion attack resistant
• CPC Class Codes: G06T 2201/0052 Embedding of the watermark in the frequency domain
• CPC Class Codes: G06T 2201/0063 in relation to collusion attacks, e.g. collusion attack resistant
• CPC Class Codes: G06T 2201/0065 Extraction of an embedded watermark; Reliable detection
• CPC Class Codes: G06T 2201/0083 whereby only watermarked image required at decoder, e.g. source-based, blind, oblivious
• CPC Class Codes: H04N 1/32144 embedded in the image data, i.e. enclosed or integrated in the image, e.g. watermark, super-imposed logo or stamp
• CPC Class Codes: H04N 1/32187 with selective or adaptive application of the additional information, e.g. in selected frequency coefficients
• CPC Class Codes: H04N 1/32229 with selective or adaptive application of the additional information, e.g. in selected regions of the image H04N1/32213 takes precedence
• CPC Class Codes: H04N 2201/3205 of identification information, e.g. name or ID code H04N2201/3209 takes precedence
• CPC Class Codes: H04N 2201/323 for tracing or tracking, e.g. forensic tracing of unauthorized copies
• CPC Class Codes: H04N 2201/327 which are undetectable to the naked eye, e.g. embedded codes

Abstract

The present invention leverages a series of media watermark locations to provide information relating to media ownership and/or distribution source. This provides a forensic means to establish a source of media with a high degree of certainty despite malicious and/or non-malicious attacks. By providing embedded user unique identification, a source of copied media can be determined without the embedded information substantially interfering with the intended purpose of the media itself. In one instance of the present invention, digital media is transformed into an N-dimensional media signal. User unique keys are then utilized along with a master key to determine locations for embedding a watermark signal that possesses substantially similar quantized statistics as the transformed media signal. The watermark signal is then embedded into the media signal, fingerprinting the media.

Claims

1. A system that facilitates media forensics, comprising:

a component that transforms digital media into an N-dimensional media signal, where N represents any positive integer from one to infinity; and

a watermark component that embeds at least one unique identifier into the N-dimensional media signal.

2. The system of claim 1, the digital media comprising at least one selected from the group consisting of audio media, image media, and video media.

3. The system of claim 1, the N-dimensional media signal comprising at least a three-dimensional signal.

4. The system of claim 1, the watermark component utilizes a pseudo-randomly derived semi-global statistics approach to embed the unique identifier into the N-dimensional media signal.

5. The system of claim 4, the pseudo-randomly derived semi-global statistics approach utilizes, at least in part, non-linear statistics.

6. The system of claim 4, the pseudo-randomly derived semi-global statistics approach based, at least in part, upon a master key, K.

7. The system of claim 1, the watermark component embeds the unique identifier into the N-dimensional media signal at high entropy locations.

8. The system of claim 7, the high entropy locations based, at least in part, upon at least one hash function.

9. The system of claim 8, the hash function comprising:

10. The system of claim 8, the high entropy locations further based, at least in part, upon having at least one valid pole determined by a pole function.

11. The system of claim 10, the pole function comprising:

Σ

<sub>i=−

C2</sub>^C2Σ

<sub>l=−

C2</sub>^C2d(h_K(s_r+i), h_K(s_r+1))>

TH₂ 

 

(Eq.

     4)

where TH₂ is a user-specified threshold and C2 is a user-specified algorithm window size parameter.

12. The system of claim 10, the valid pole comprising a pole selected from a group of poles that number beyond a fractional threshold of a total number of poles targeted as possible valid poles.

13. The system of claim 12, the fractional threshold comprising a value of approximately 0.7.

14. The system of claim 1, further comprising:

a pre-processing component that applies at least one pre-processing technique before the watermark component embeds the unique identifier into the N-dimensional media signal; and

a post-processing component that applies at least one post-processing technique after the watermark component embeds the unique identifier into the N-dimensional media signal.

15. The system of claim 1, further comprising:

a watermark detection component that detects the unique identifier embedded in the N-dimensional media signal.

16. The system of claim 15, further comprising:

a pre-processing component that applies at least one pre-processing technique before the watermark detection component detects the unique identifier in the N-dimensional media signal.

47. A computer readable medium having stored thereon computer executable components of the system of claim 1.

49. A device employing the system of claim 1 comprising at least one selected from the group consisting of a computer, a server, and a handheld electronic device.

17. A method for facilitating media forensics, comprising:

transforming digital media into an N-dimensional media signal, where N represents any positive integer from one to infinity; and

embedding a unique identifier into the N-dimensional media signal.

18. The method of claim 17, further comprising:

creating a master key and at least one user key;

determining a hash function randomized by the master key;

selecting a plurality of locations within the digital media randomly based upon the master key;

calculating hash values utilizing the hash function for the plurality of locations;

determining neighborhood regions randomly selected based on the master key for the plurality of locations; and

replacing objects within the digital media with at least one user object based on the user key in selected neighborhood regions, creating the unique identifier.

19. The method of claim 18, further comprising:

choosing a sub-band of the N-dimensional media signal;

selecting sub-band objects within the sub-band in a random fashion;

weighting entities within the sub-band objects that are randomly selected;

determining statistics of the entities;

quantizing the statistics of the entities;

creating a watermark signal that, when combined with the N-dimensional media signal at one of the plurality of locations, exhibits substantially similar quantized statistics as the N-dimensional media signal alone; and

combining the watermark signal with the N-dimensional media signal to formulate the user object.

20. The method of claim 19, the sub-band objects comprising at least one selected from the group consisting of rectangular prisms, spheres, and cylinders.

21. The method of claim 19, the entities comprising pixels.

22. The method of claim 17, the unique identifier comprising:

at least one watermark that indicates a particular user associated with the digital media to a substantially high degree of certainty.

23. The method of claim 17, transforming digital media into an N-dimensional media signal comprising:

utilizing a transformation operator that compactly captures energy of the digital media into the N-dimensional media signal.

24. The method of claim 23, the transformation operator comprising at least one selected from the group consisting of separable filters, non-separable filters, linear shift invariant filters, linear shift filters, and P-dimensional wavelet transforms, where P is any positive integer from one to infinity.

25. The method of claim 17, further comprising:

applying at least one pre-processing technique before embedding the unique identifier into the N-dimensional media signal; and

employing at least one post-processing technique after embedding the unique identifier into the N-dimensional media signal.

26. The method of claim 25, the pre-processing and post-processing techniques comprising:

utilizing a subset determination technique to separate a plurality of pixels into at least two subsets and making a first subset group and a second subset group available for unique identifier detection;

applying histogram equalization to the pixels of the first subset group before embedding the unique identifier into the N-dimensional media signal; and

inversely mapping the histogram equalization that was applied to the first subset group after embedding the unique identifier into the N-dimensional media signal.

27. The method of claim 26, further comprising:

employing histogram equalization to pixels of the first subset group made available for unique identifier detection;

replacing pixels corresponding to the second subset group with their original unmarked values; and

detecting the unique identifier embedded in the N-dimensional media signal.

28. The method of claim 17, further comprising:

detecting the unique identifier embedded in the N-dimensional media signal.

29. The method of claim 28, further comprising:

applying at least one pre-processing technique before detecting the unique identifier in the N-dimensional media signal.

30. The method of claim 17, embedding a unique identifier into the N-dimensional media signal further comprising, at least in part:

utilizing a pseudo-randomly derived semi-global statistics approach to embed the unique identifier into the N-dimensional media signal.

31. The method of claim 30, the pseudo-randomly derived semi-global statistics approach utilizing, at least in part, non-linear statistics.

32. The method of claim 30, the pseudo-randomly derived semi-global statistics approach based, at least in part, upon a master key, K.

33. The method of claim 17, embedding a unique identifier into the N-dimensional media signal further comprising, at least in part:

embedding the unique identifier into the N-dimensional media signal at high entropy locations.

34. The method of claim 33, the high entropy locations based, at least in part, upon at least one hash function.

35. The method of claim 34, the hash function comprising:

36. The method of claim 34, the high entropy locations further based, at least in part, upon having at least one valid pole determined by a pole function.

37. The method of claim 36, the pole function comprising:

Σ

<sub>i=−

C2</sub>^C2Σ

<sub>l=−

C2</sub>^C2d(h_K(s_r+i), h_K(s_r+1))>

TH₂ 

 

(Eq.

     4)

where TH₂ is a user-specified threshold and C2 is a user-specified algorithm window size parameter.

38. The method of claim 36, the valid pole comprising a pole selected from a group of poles that number beyond a fractional threshold of a total number of poles targeted as possible valid poles.

39. The method of claim 38, the fractional threshold comprising a value of approximately 0.7.

40. The method of claim 17, embedding the unique identifier into the N-dimensional media signal further comprising:

utilizing shape adaptation of object groupings along a time axis dependent on a motion search to select locations for embedding the unique identifier.

41. The method of claim 17, embedding the unique identifier into the N-dimensional media signal further comprising:

employing image hash values based, at least in part, on iterative geometric methods for selecting multiple mark embedding points and for achieving time synchronization of the unique identifier.

48. A device employing the method of claim 17 comprising at least one selected from the group consisting of a computer, a server, and a handheld electronic device.

42. A system that facilitates media forensics, comprising:

means for transforming digital media into an N-dimensional media signal; and

means for embedding at least one unique identifier into the N-dimensional media signal.

43. A method of facilitating media forensics, comprising:

detecting from digital media a unique identifier, the unique identifier incorporated into the digital media via embedding at least one watermark into an N-dimensional media signal.

44. The method of claim 43, the detecting from the digital media further comprising:

correlating pseudo-randomly chosen subsets of object groupings relating to statistical information contained in the embedded watermark; and

sorting resulting correlations and utilizing an average correlation of highest X subsets as an overall correlation value to minimize an effect of at least one specific object grouping hurt by an attack;

where X is any positive integer from one to infinity.

45. The method of claim 43, further comprising:

determining at least one particular user associated with the digital media with a substantially high degree of certainty by utilizing, at least in part, the unique identifier.

46. A data packet transmitted between two or more computer components that facilitates media forensics, the data packet comprising, at least in part, information relating to at least one media source, the information including, at least in part, at least one unique identifier embedded into an N-dimensional media signal via transformation of digital media into an N-dimensional media signal.