Biometric processing using random projection transforms

• US 8,384,515 B2
• Filed: 09/15/2008
• Issued: 02/26/2013
• Est. Priority Date: 09/15/2008
• Status: Active Grant

• Current Assignee: Accenture GmbH (Accenture PLC)
• Original Assignee: Accenture Global Services Limited (Accenture PLC)
• Inventors: Rachlin, Yaron
• Primary Examiner(s): BROWN, VERNAL U
• Application Number:
  US12/210,694
• Publication Number:
  US 20100066493A1
• Time in Patent Office: 1,625 Days
• Field of Search: 340/5.82, 340/3.1, 235/380, 705/44, 713/186
• US Class Current: 340/5.82
• CPC Class Codes: G06F 18/213 Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
• CPC Class Codes: G06F 21/32 using biometric data, e.g. fingerprints, iris scans or voiceprints
• CPC Class Codes: G06F 21/6245 Protecting personal data, e.g. for financial or medical purposes
• CPC Class Codes: G06V 10/7715 Feature extraction, e.g. by transforming the feature space, e.g. multi-dimensional scaling [MDS]; Mappings, e.g. subspace methods
• CPC Class Codes: G06V 40/10 Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
• CPC Class Codes: G07C 2209/12 Comprising means for protecting or securing the privacy of biometric data, e.g. cancellable biometrics
• CPC Class Codes: G07C 9/37 using biometric data, e.g. fingerprints, iris scans or voice recognition

Abstract

To provide cancelable biometrics, random projection transforms are used to produce transformed biometric data that may be used for authentication purposes. Each random projection transform provides a one-way or irreversible transformation such that the resulting transformed biometric data alone cannot be used to reconstruct the biometric data. Enrollment and backup transforms are produced based on enrollment biometric data. Authentication is possible by similarly transforming authentication biometric and then securely reconstructing the enrollment and authentication biometric data for comparison. In a similar manner, the enrollment and backup transforms may be used to reenroll biometric data when its corresponding transformed biometric data is compromised without requiring further user intervention.

Claims

1. A method for processing biometric data, the method comprising:

obtaining, by a processing device, enrollment biometric data of an enrolling user; and

encoding, by a processing device, the enrollment biometric data using an enrollment random projection transform to produce transformed enrollment biometric data, wherein the enrollment random projection transform comprises

a measurement function that is sufficiently incoherent to a basis in which the enrollment biometric data is sparse, and meeting the conditions of;

m_x+m_y≧

H(X,Y);

m_x≧

H(X|Y);

m_y≧

H(Y|X);

m_x<

H(X);

m_y<

H(Y);

where X is the enrollment biometric data, Y is at least one other biometric data, H(X,Y) is a joint entropy of X and Y, m_x is an encoding rate of the enrollment random projection transform, m_y is an encoding rate of another random projection transform applied to Y, H(X) is an entropy of X expressing a minimum encoding rate needed to reconstruct X, H(Y) is an entropy of Y expressing a minimum encoding rate needed to reconstruct Y, H(X|Y) is a conditional entropy expressing a minimum encoding rate needed to reconstruct X given knowledge of Y and H(Y|X) is a conditional entropy expressing a minimum encoding rate needed to reconstruct Y given X.

2. The method of claim 1, further comprising:

storing, by the processing device, the transformed enrollment biometric data and the enrollment random projection transform in association with an identification of the enrolling user in an authentication database.

3. The method of claim 1, wherein the enrollment random projection transform comprises a plurality of measurement vectors that are incoherent relative to a basis in which the biometric data is sparse.

4. The method of claim 1, wherein the enrollment biometric data comprises a feature vector that is substantially uniquely descriptive of a biometric characteristic of the enrolling user.

5. The method of claim 1, wherein the enrollment biometric data comprises at least one of an image of a face of the enrolling user, a recording of a voice of the enrolling user, an image of a fingerprint of the enrolling user or an image of an iris of the enrolling user.

6. The method of claim 1, further comprising:

encoding, by the processing device, the enrollment biometric data using a backup random projection transform to produce transformed backup biometric data, wherein knowledge of the backup random projection transform and the transformed backup biometric data is insufficient to reconstruct the enrollment biometric data.

7. The method of claim 6, further comprising:

storing, by the processing device, the transformed backup biometric data and the backup random projection transform in association with the identification of the enrolling user in a backup database.

8. A method for processing biometric data, the method comprising:

obtaining, by a processing device, authentication biometric data of an authenticating user;

encoding, by the processing device, the authentication biometric data using an authentication random projection transform to produce transformed authentication biometric data, wherein the authentication random projection transform comprises a measurement function that

sufficiently incoherent to a basis in which the e authentication biometric data is sparse, and

meeting the conditions of;

m^x+m_y≧

H(X,Y);

m_x≧

H(X|Y);

m_y≧

H(Y|X);

m_x<

H(X);

m_y<

H(Y);

where X is the enrollment biometric data Y is at least one other biometric data, H(X,Y) is a joint entropy of X and Y, m_x is an encoding rate of the enrollment random projection transform, m_y is an encoding rate of another random projection transform applied to Y,H(X) is an entropy of X expressing a minimum encoding rate needed to reconstruct X, H(Y) is an entropy of Y expressing a minimum encoding rate needed to reconstruct Y, H(X|Y) is a conditional entropy expressing a minimum encoding rate needed to reconstruct X given knowledge of Y and H(Y|X) is a conditional entropy expressing a minimum encoding rate needed to reconstruct Y given X,

providing, by the processing device, the authentication random projection transform and the transformed authentication biometric data to an authenticator; and

receiving, by the processing device from the authenticator, an authentication decision concerning the authenticating user, wherein the authenticating decision is based on the authentication random projection transform and the transformed authentication biometric data.

9. The method of claim 8, wherein providing the authentication random projection transform and the transformed authentication biometric data to the authenticator further comprises:

receiving, by the processing device, an identification of an enrolling user; and

providing, by the processing device, the identification of the enrolling user to the authenticator,

wherein the authentication decision comprises an indication whether the authenticating user is the enrolling user.

10. A method for processing biometric data, the method comprising:

obtaining, by a processing device, transformed enrollment biometric data, transformed authentication biometric data, an enrollment random projection transform and an authentication random projection transform;

determining, by the processing device, reconstructed enrollment biometric data and reconstructed authentication biometric data based on the transformed enrollment biometric data, the transformed authentication biometric data, the enrollment random projection transform and the authentication random projection transform;

comparing, by the processing device, the reconstructed enrollment biometric data and the reconstructed authentication biometric data; and

providing, by the processing device, an authentication decision based on comparison of the reconstructed enrollment biometric data and the reconstructed authentication biometric data.

11. The method of claim 10, further comprising:

receiving, by the processing device from an authentication requester, an identification of an enrolling user; and

obtaining, by the processing device from an authentication database, the transformed enrollment biometric data and the enrollment random projection transform based on the identification of the enrolling user.

12. The method of claim 11, further comprising:

receiving, by the processing device from the authentication requester, the transformed authentication biometric data and the authentication random projection transform.

13. The method of claim 10, wherein the authentication decision comprises an indication that an authenticating user associated with the transformed authentication biometric data is an enrolling user associated with the transformed enrollment biometric data when the reconstructed enrollment biometric data compares favorably with the reconstructed authentication biometric data, or an indication that the authenticating user is not the enrolling user when the reconstructed enrollment biometric data does not compare favorably with the reconstructed authentication biometric data.

14. A method for processing biometric data, the method comprising:

obtaining, by a processing device, transformed enrollment biometric data, transformed backup biometric data, an enrollment random projection transform and a backup random projection transform;

determining, by the processing device, reconstructed biometric data corresponding to an enrolling user based on the transformed enrollment biometric data, the transformed backup biometric data, the enrollment random projection transform and the backup random projection transform; and

encoding, by the processing device, the reconstructed biometric data using a reenrollment random projection transform to produce transformed reconstructed biometric data, wherein knowledge of the reenrollment random projection transform and the transformed reconstructed biometric data is insufficient to reconstruct the reconstructed biometric data.

15. The method of claim 14, further comprising, prior to determining the reconstructed biometric data:

determining that security of the transformed enrollment biometric data has been compromised.

16. The method of claim 14, further comprising:

receiving, by the processing device from a reenrollment requester, an identification of the enrolling user; and

obtaining, by the processing device, the transformed enrollment biometric data and the enrollment random projection transform from an authentication database based on the identification of the enrolling user; and

obtaining, by the processing device, the transformed backup biometric data and the backup random projection transform from a backup database based on the identification of the enrolling user.

17. The method of claim 14, further comprising:

storing, by the processing device, the transformed reconstructed biometric data and the reenrollment random projection transform in association with an identification of the enrolling using in another authentication database.

18. A system for processing biometric data comprising:

an enrollment encoder operative to process enrollment biometric data of an enrolling user using an enrollment random projection to transform to produce transformed enrollment biometric data, wherein the enrollment random projection transform comprises a measurement function that is sufficiently incoherent to a basis in which the enrollment biometric data is sparse, and meeting the conditions of

m^x+m_y≧

H(X,Y);

m_x≧

H(X|Y);

m_y≧

H(Y|X);

m_x<

H(X);

m_y<

H(Y);

where X is the enrollment biometric data, Y is at least one other biometric data, H(X,Y) is a joint entropy of X and Y, m_x is an encoding rate of the enrollment random projection transform, m_y is an encoding rate of another random projection transform applied to Y, H(X) is an entropy of X expressing a minimum encoding rate needed to reconstruct X, H(Y) is an entropy of Y expressing a minimum encoding rate needed to reconstruct Y, H(X|Y) is a conditional entropy expressing a minimum encoding rate needed to reconstruct X given knowledge of Y and H(Y|X)is a conditional entropy expressing a minimum encoding rate needed to reconstruct Y given X, and

an authentication database, in communication with the enrollment encoder, operative to store the transformed enrollment biometric data and the enrollment random projection transform in association with an identification of the enrolling user.

19. The system of claim 18, further comprising:

an enrollment biometric input device, in communication with the enrollment encoder, operative to process a biometric feature of the enrolling user and provide the enrollment biometric data as output, wherein the biometric feature comprises at least one of the enrolling user'"'"'s;

face, voice, fingerprint or iris.

20. The system of claim 18, further comprising:

a backup encoder operative to process the enrollment biometric data using a backup random projection transform to produce transformed backup biometric data, wherein knowledge of the backup random projection transform and the transformed backup biometric data is insufficient to reconstruct the enrollment biometric data; and

a backup database, in communication with the backup encoder, operative to store the transformed backup biometric data and the backup random projection transform in association with the identification of the enrolling user.

21. The system of claim 20, wherein the enrollment random projection transform and the backup random projection transform each comprises a plurality of measurement vectors that are incoherent relative to a basis in Which the biometric data is sparse.

22. An system for processing biometric data comprising:

an authentication encoder operative to process authentication biometric data of an authenticating user using an authentication random projection transform to produce transformed authentication biometric data, wherein knowledge of the authentication random projection transform and the transformed authentication biometric data is insufficient reconstruct the authentication biometric data;

an authentication reconstructor, in communication with the authentication encoder, operative to determine reconstructed enrollment biometric data and reconstructed authentication biometric data based on transformed enrollment biometric data, the transformed authentication biometric data, an enrollment random projection transform and the authentication random projection transform; and

a comparator, in communication with the authentication reconstructor, operative to compare the reconstructed enrollment biometric data and the reconstructed authentication biometric data and provide an authentication decision based on comparison of the reconstructed enrollment biometric data and the reconstructed authentication biometric data.

23. The apparatus of claim 22, further comprising:

an authentication biometric input device, in communication with the authentication encoder, operative to process a biometric feature of the authenticating user and provide the authentication biometric data as output, wherein the biometric feature comprises at least one of the authenticating user'"'"'s;

face, voice, fingerprint or iris.

24. The apparatus of claim 22, wherein the authentication random projection transform comprises a plurality of measurement vectors that are incoherent relative to a basis in which the biometric data is sparse.

25. The apparatus of claim 22, wherein the authentication reconstructor further comprises:

an input interface, in communication with an authentication database, for obtaining the transformed enrollment biometric data and the enrollment random projection transform.

26. A system for processing biometric data comprising:

an authentication database comprising transformed enrollment biometric data and an enrollment random projection transform associated with an identification of an enrolling user;

a backup database comprising transformed backup biometric data and a backup random projection transform associated with the identification of the enrolling user;

a reenrollment reconstructor, in communication with the authentication database and the backup database, operative to determine reconstructed enrollment biometric data based on the transformed enrollment biometric data, the transformed backup biometric data, the enrollment random projection transform and the backup random projection transform; and

a reenrollment encoder operative to process the reconstructed enrollment biometric data using a reenrollment random projection transform to produce transformed reconstructed biometric data, wherein knowledge of the reenrollment random projection transform and the transformed reconstructed biometric data is insufficient to reconstruct enrollment biometric data.

27. The system of claim 26, further comprising:

another authentication database, in communication with the reenrollment encoder, operative to store the transformed reconstructed biometric data and the reenrollment random projection transform in association with the identification of the enrolling user.

28. The system of claim 26, wherein the reenrollment random projection transform comprises a plurality of measurement vectors that are incoherent relative to a basis in which the biometric data is sparse.

29. An apparatus for processing biometric data comprising:

a processor; and

a processor-readable medium having stored thereon executable instructions that, when executed by the processor, cause the processor to;

obtaining enrollment biometric data of an enrolling user; and

encode the enrollment biometric data using an enrollment random projection transform to produce transformed enrollment biometric data, wherein the enrollment random projection transform comprises a measurement function that is sufficiently incoherent to a basis in which the enrollment biometric data is sparse, and meeting the conditions of

meeting the conditions of;

m^x+m_y≧

H(X,Y);

m_x≧

H(X|Y);

m_y≧

H(Y|X);

m_x<

H(X);

m_y<

H(Y);

where X is the enrollment biometric data, Y is at least one other biometric data, H(X,Y) is a joint entropy of X and Y, m_x is an encoding rate of the enrollment random projection transform, m_y an encoding another random projection transform applied to Y, H(X)is an entropy of X expressing a minimum encoding rate needed to reconstruct X, H(Y) is an entropy of Y expressing a minimum encoding rate needed to reconstruct Y, H(X|Y) is a conditional entropy expressing a minimum encoding rate needed to reconstruct X given knowledge of Y and H(Y|X) is a conditional entropy expressing a minimum encoding rate needed to reconstruct Y given X.

30. The apparatus of claim 29, wherein the processor-readable medium further comprises executable instructions that, when executed by the processor, cause the processor to store the transformed enrollment biometric data and the enrollment random projection transform in association with an identification of the enrolling user in an authentication database.

31. The apparatus of claim 29, wherein the enrollment random projection transform comprises a plurality of measurement vectors that are incoherent relative to a basis in which the biometric data is sparse.

32. The apparatus of claim 29, wherein the processor-readable medium further comprises executable instructions that, when executed by the processor, cause the processor to:

encode the enrollment biometric data using a backup random projection transform to produce transformed backup biometric data, wherein knowledge of the backup random projection transform and the transformed backup biometric data is insufficient to reconstruct the enrollment biometric data.

33. The apparatus of claim 32, wherein the processor-readable medium further comprises executable instructions that, when executed by the processor, cause the processor to:

store the transformed backup biometric data and the backup random projection transform in association with the identification of the enrolling user in a backup database.

34. An apparatus for processing biometric data comprising:

a processor; and

a processor-readable medium having stored thereon executable instructions that, when executed by the processor, cause the processor to;

obtain authentication biometric data of an authenticating user;

encode the authentication biometric data using an authentication random projection transform to produce transformed authentication biometric data, wherein the authentication random projection transform comprises a measurement function that is sufficiently incoherent to a basis in which the authentication biometric data is sparse, and meeting the conditions of;

m^x+m_y≧

H(X,Y);

m_x≧

H(X|Y);

m_y≧

H(Y|X);

m_x<

H(X);

m_y<

H(Y);

where X is the enrollment biometric data, Y is at least one other biometric data, H(X,Y) is a joint entropy of X and Y, m_x is a rate of the enrollment random projection transform, m_y is an encoding rate of another random projection transform applied to Y, H(X) is an entropy of X expressing a minimum encoding rate needed to reconstruct X, H(Y) is an entropy of Y expressing a minimum encoding rate needed to reconstruct Y, H(X|Y) is a conditional entropy expressing a minimum encoding rate needed to reconstruct X given knowledge of Y and H(Y|X) is a conditional entropy expressing a minimum encoding rate needed to reconstruct Y given X,

providing the authentication random projection transform and the transformed authentication biometric data to an authenticator; and

receive, from the authenticator, an authentication decision concerning the authenticating user, wherein the authenticating decision is based on the authentication random projection transform and the transformed authentication biometric data.

35. The apparatus of claim 34, wherein the executable instructions that, when executed by the processor, cause the processor to provide the authentication random projection transform and the transformed authentication biometric data to the authenticator further comprise executable instructions that, when executed by the processor, cause the processor to:

receive an identification of an enrolling user; and

provide the identification of the enrolling user to the authenticator,

wherein the authentication decision comprises an indication whether the authenticating user is the enrolling user.

36. An apparatus for processing biometric data comprising:

a processor; and

a processor-readable medium having stored thereon executable instructions that, when executed by the processor, cause the processor to;

obtain transformed enrollment biometric data, transformed authentication biometric data, an enrollment random projection transform and an authentication random projection transform;

determine reconstructed enrollment biometric data and reconstructed authentication biometric data based on the transformed enrollment biometric data, the transformed authentication biometric data, the enrollment random projection transform and the authentication random projection transform;

compare the reconstructed enrollment biometric data and the reconstructed authentication biometric data; and

provide an authentication decision based on comparison of the reconstructed enrollment biometric data and the reconstructed authentication biometric data.

37. The apparatus of claim 36, wherein the processor-readable medium further comprises executable instructions that, when executed by the processor, cause the processor to:

receive, from an authentication requester, an identification of an enrolling user; and

obtain, from an authentication database, the transformed enrollment biometric data and the enrollment random projection transform based on the identification of the enrolling user.

38. The apparatus of claim 37, wherein the processor-readable medium further comprises executable instructions that, when executed by the processor, cause the processor to:

receive, from the authentication requester, the transformed authentication biometric data and the authentication random projection transform.

39. The apparatus of claim 36, wherein the authentication decision comprises an indication that an authenticating user associated with the transformed authentication biometric data is an enrolling user associated with the transformed enrollment biometric data when the reconstructed enrollment biometric data compares favorably with the reconstructed authentication biometric data, or an indication that the authenticating user is not the enrolling user when the reconstructed enrollment biometric data does not compare favorably with the reconstructed authentication biometric data.

40. An apparatus for processing biometric data comprising:

a processor; and

a processor-readable medium having stored thereon executable instructions that, when executed by the processor, cause the processor to;

obtain transformed enrollment biometric data, transformed backup biometric data, an enrollment random projection transform and a backup random projection transform;

determine reconstructed biometric data corresponding to an enrolling user based on the transformed enrollment biometric data, the transformed backup biometric data, the enrollment random projection transform and the backup random projection transform; and

encode the reconstructed biometric data using a reenrollment random projection transform to produce transformed reconstructed biometric data, wherein knowledge of the reenrollment random projection transform and the transformed reconstructed biometric data is insufficient to reconstruct the reconstructed biometric data.

41. The apparatus of claim 40, wherein the processor-readable medium further comprises executable instructions that, when executed by the processor, cause the processor to:

determine that security of the transformed enrollment biometric data has been compromised.

42. The apparatus of claim 40, wherein the processor-readable medium further comprises executable instructions that, when executed by the processor, cause the processor to:

receive, from a reenrollment requester, an identification of the enrolling user; and

obtain the transformed enrollment biometric data and the enrollment random projection transform from an authentication database based on the identification of the enrolling user; and

obtain the transformed backup biometric data and the backup random projection transform from a backup database based on the identification of the enrolling user.

43. The apparatus of claim 40, wherein the processor-readable medium further comprises executable instructions that, when executed by the processor, cause the processor to:

store the transformed reconstructed biometric data and the reenrollment random projection transform in association with an identification of the enrolling using in another authentication database.