Hierarchical and index based watermarks represented as trees

• US 8,750,630 B2
• Filed: 07/13/2012
• Issued: 06/10/2014
• Est. Priority Date: 07/13/2012
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Baughman, Aaron K., Darden, Richard L., Fan, James J., Kalyanpur, Aditya A.
• Primary Examiner(s): Alavi, Amir
• Assistant Examiner(s): CESE, KENNY A
• Application Number:
  US13/549,056
• Publication Number:
  US 20140016814A1
• Time in Patent Office: 697 Days
• Field of Search: 382/100, 382/226, 382/227, 713/176, 704/1, 704/9
• US Class Current: 382/226
• CPC Class Codes: G06F 21/16 Program or content traceability, e.g. by watermarking

Abstract

An approach that provides hierarchical and index based watermarks represented as trees is described. In one embodiment, a watermark tree is formed from feature watermarks generated from a natural language processing (NLP) stack having NLP analytics. The watermark tree represents a hierarchical relationship between each of the feature watermarks. In particular, the watermark tree defines hierarchical pointers that point out inherited watermarks that exist between the feature watermarks according to the hierarchical relationship. Further, the watermark tree includes a time stamp specifying a time that a data set content residing in a corpus was accessed.

Claims

1. A method, performed on a computer system, for watermarking content stored in a plurality of corpora each having a plurality of corpuses, the method comprising:

using the computer system to perform the following;

receiving a data set of content from a corpus within one of the plurality of corpora;

applying the data set to a natural language processing (NLP) stack including a plurality of NLP analytics each analyzing the data set;

generating a feature watermark from each of the plurality of NLP analytics; and

forming a watermark tree from each of the feature watermarks, the watermark tree representing a hierarchical relationship between each of the feature watermarks generated from each of the plurality of NLP analytics, the watermark tree defining hierarchical pointers that point out inherited watermarks that exist between the feature watermarks according to the hierarchical relationship, the watermark tree including a time stamp specifying a time that the data set was accessed from the corpus.

2. The method according to claim 1, further comprising storing the watermark tree in the corpus with a corpus time stamp and applying a corpora time stamp to the corpora associated with the corpus in response to storing the watermark in the corpus, the corpus including a plurality of watermark trees each formed at a different time stamp, and the corpora including a plurality of corpora time stamps each associated with storing one of the plurality of watermark trees in the corpus.

3. The method according to claim 2, further comprising using the time stamped plurality of watermark trees to validate content of the corpus in response to being updated with a new source of data.

4. The method according to claim 3, further comprising authenticating the corpus and corpora at a particular time as a function of the plurality of watermark trees.

5. The method according to claim 4, wherein the authenticating includes selecting a watermark tree from the plurality of watermark trees that has a time stamp that corresponds with the particular time designated in the authenticating of the corpus and corpora, recursively transversing the hierarchical pointers in the watermark tree to measure whether a feature watermark associated with each parent in the hierarchical relationship can be validated as having a valid feature watermark.

6. The method according to claim 5, further comprising invalidating a parent in the hierarchical relationship that fails to have a valid feature watermark, and invalidating all children in the hierarchical relationship that depend from the invalidated parent.

7. The method according to claim 6, further comprising invalidating all feature watermarks in a sub-chain of the hierarchical relationship that emanate from the invalidated parent.

8. The method according to claim 7, further comprising filtering all content associated with all invalidated feature watermarks from search results generated from the corpus in response to a search request made thereon.

9. The method according to claim 7, further comprising re-preprocessing all data content associated with parents and children in the hierarchical relationship that have invalidated feature watermarks.

10. The method according to claim 9, wherein the re-preprocessing comprises generating a new feature watermark for only parents and children in the hierarchical relationship that have invalidated feature watermarks, while parents and children in the hierarchical relationship having validated features retain current watermark features.

11. The method according to claim 10, further comprising forming an updated watermark tree for the hierarchical relationship that includes both the new feature watermarks and any unchanged watermarks, the updated watermark tree including updated hierarchical pointers that point out inherited watermarks that exist between the new feature watermarks and the unchanged watermarks according to the hierarchical relationship, the updated watermark tree including a new time stamp specifying a time that the updated watermark tree was formed.

12. The method according to claim 11, further comprising storing the updated watermark tree in the corpus with a new corpus time stamp and a new corpora time stamp.

13. The method according to claim 1, wherein each feature watermark is determined according to:

F(C_t,T, S_t)=W_t^x, wherein

C_t is a content vector at time t,

T is a time stamp at time t,

S_t is a supertype vector at time stamp T,

W_t^x is a watermark of feature or attribute x at time stamp T, and

F(*) is the visible watermark or hashing parameterized function.

14. A computer system for watermarking content stored in a plurality of corpora each having a plurality of corpuses, the computer system comprising:

at least one processing unit;

memory operably associated with the at least one processing unit; and

a feature watermark tool storable in memory and executable by the at least one processing unit, the tool comprising;

an input component configured to receive a data set of content from a corpus within one of the plurality of corpora;

a natural language processing (NLP) stack including a plurality of NLP analytics each configured to extract features from the data set; and

a feature watermark generator configured to generate a feature watermark for each of the plurality of NLP analytics for features extracted therefrom, the feature watermark generator further configured to form a watermark tree from each of the feature watermarks, the watermark tree representing a hierarchical relationship between each of the feature watermarks generated from each of the plurality of NLP analytics, the watermark tree defining hierarchical pointers that point out inherited watermarks that exist between the feature watermarks according to the hierarchical relationship, the watermark tree including a time stamp specifying a time that the data set was accessed from the corpus.

15. The computer system according to claim 14, wherein the feature watermark generator is further configured to store the watermark tree in the corpus with a corpus time stamp and apply a corpora time stamp to the corpora associated with the corpus in response to storing the watermark in the corpus, the corpus including a plurality of watermark trees each formed at a different time stamp, and the corpora including a plurality of corpora time stamps each associated with storing one of the plurality of watermark trees in the corpus.

16. The computer system of claim 14, wherein the feature watermark generator generates the feature watermarks from a plurality of hierarchical hashes each being a component of a feature watermark function defined as:

F(C_t,T, S_t)=W_t^x, wherein

C_t is a content vector at time t,

T is a time stamp at time t,

S_t is a supertype vector at time stamp T,

W_t^x is a watermark of feature or attribute x at time stamp T, and

F(*) is the visible watermark or hashing parameterized function.

17. The computer system of claim 14, wherein the feature watermark generator is further configured to use the time stamped plurality of watermark trees to validate content of the corpus in response to being updated with a new source of data.

18. A non-transitory computer readable medium storing computer instructions, which when executed, enables a computer system to watermark content stored in a plurality of corpora each having a plurality of corpuses, the executed computer instructions causing the computer system to perform a method comprising:

receiving a data set of content from a corpus within one of the plurality of corpora;

applying the data set to a natural language processing (NLP) stack including a plurality of NLP analytics each analyzing the data set;

generating a feature watermark from each of the plurality of NLP analytics; and

forming a watermark tree from each of the feature watermarks, the watermark tree representing a hierarchical relationship between each of the feature watermarks generated from each of the plurality of NLP analytics, the watermark tree defining hierarchical pointers that point out inherited watermarks that exist between the feature watermarks according to the hierarchical relationship, the watermark tree including a time stamp specifying a time that the data set was accessed.

19. The non-transitory computer readable medium according to claim 18, wherein the method performed by the computer system operating according to the executed computer instructions further comprises storing the watermark tree in the corpus with a corpus time stamp and applying a corpora time stamp to the corpora associated with the corpus in response to storing the watermark in the corpus, the corpus including a plurality of watermark trees each formed at a different time stamp, and the corpora including a plurality of corpora time stamps each associated with storing one of the plurality of watermark trees in the corpus.

20. The non-transitory computer readable medium according to claim 19, wherein the method performed by the computer system operating according to the executed computer instructions further comprises using the time stamped plurality of watermark trees to validate content of the corpus in response to being updated with a new source of data.

21. The non-transitory computer readable medium according to claim 19, wherein the method performed by the computer system operating according to the executed computer instructions further comprises authenticating the corpus and corpora at a particular time as a function of the plurality of watermark trees, wherein the authenticating includes selecting a watermark tree from the plurality of watermark trees that has a time stamp that corresponds with the particular time designated in the authenticating of the corpus and corpora, recursively transversing the hierarchical pointers in the watermark tree to measure whether a feature watermark associated with each parent in the hierarchical relationship can be validated as having a valid feature watermark.

22. The non-transitory computer readable medium according to claim 21, wherein the method performed by the computer system operating according to the executed computer instructions further comprises invalidating a parent in the hierarchical relationship that fails to have a valid feature watermark, and invalidating all children in the hierarchical relationship that depend from the invalidated parent.

23. The non-transitory computer readable medium according to claim 22, wherein the method performed by the computer system operating according to the executed computer instructions further comprises invalidating all feature watermarks in a sub-chain of the hierarchical relationship that emanate from the invalidated parent.

24. The non-transitory computer readable medium according to claim 23, wherein the method performed by the computer system operating according to the executed computer instructions further comprises re-preprocessing all data content associated with parents and children in the hierarchical relationship that have invalidated feature watermarks, the re-preprocessing including generating a new feature watermark for only parents and children in the hierarchical relationship that have invalidated feature watermarks, while parents and children in the hierarchical relationship having validated features retain current watermark features.

25. The non-transitory computer readable medium according to claim 24, wherein the method performed by the computer system operating according to the executed computer instructions further comprises forming an updated watermark tree for the hierarchical relationship that includes both the new feature watermarks and any unchanged watermarks, and storing the updated watermark tree in the corpus with a new corpus time stamp and a new corpora time stamp, the updated watermark tree including updated hierarchical pointers that point out inherited watermarks that exist between the new feature watermarks and the unchanged watermarks according to the hierarchical relationship, the updated watermark tree including a new time stamp specifying a time that the updated watermark tree was formed.