System for and method of securing fingerprint biometric systems against fake-finger spoofing

• US 20070014443A1
• Filed: 07/10/2006
• Published: 01/18/2007
• Est. Priority Date: 07/12/2005
• Status: Active Grant

• Current Assignee: Apple Inc.
• Original Assignee: AuthenTec Incorporated (Apple Inc.)
• Inventors: Russo, Anthony
• Granted Patent: US 7,505,613 B2
• Time in Patent Office: Days
• US Class Current: 382/124
• CPC Class Codes: G06V 40/1388 using image processing
• CPC Class Codes: G06V 40/45 Detection of the body part being alive

Abstract

A biometric secured system grants a user access to a host system by classifying a fingerprint used to verify or authorize the user to the system as real or fake. The classification is based on a probability that fingerprint image data corresponds to characteristics that reliably identify the finger as real. The system includes a fingerprint sensor for capturing fingerprint image data coupled to a spoof detection module. In one embodiment, the spoof detection module is programmed to determine spoof probability based on a combination of metrics that include, among other metrics, pixel gray level average and the variance of pixels corresponding to a fingerprint ridge, pixel gray level average and the variance of pixels corresponding to a fingerprint valley, density of sweat pores, and density of sweat streaks, to name a few metrics.

Claims

1. A biometric secured system comprising:

a fingerprint sensor for capturing fingerprint image data; and

a spoof detection module programmed to determine a spoof probability from a combination of metrics derived from the fingerprint image data.

2. The secured system of claim 1, wherein the spoof detection module comprises:

a metric generator for generating the metrics; and

classifier logic programmed to generate from the metrics a raw probability that the fingerprint image data was generated from a synthetic material.

3. The secured system of claim 1, wherein the combination of metrics comprise any two or more of pixel gray level average and variance of pixels corresponding to a fingerprint ridge, pixel gray level average and variance of pixels corresponding to a fingerprint valley, density of sweat pores, density of streaks, and raw signal level.

4. The secured system of claim 1, wherein the combination of metrics comprise values corresponding to any two or more of ridge elasticity of a fingerprint image, electrical properties of a fingerprint, optical properties of a fingerprint, and vitality properties of a fingerprint.

5. The secured system of claim 2, wherein the spoof detection module also comprises an adjustor programmed to adjust the raw probability by a base probability to generate the spoof probability.

6. The secured system of claim 5, wherein the base probability is generated from stored metrics.

7. The secured system of claim 6, wherein the stored metrics are based on fingerprint image data captured during an enrollment step.

8. The secured system of claim 2, wherein the spoof detection module also comprises a filter for dividing the fingerprint sensor into multiple windows, and the classifier logic is also programmed to determine the spoof probability based on a comparison of the values of the metric in each of the multiple windows.

9. The secured system of claim 8, wherein dimensions of each of the windows are programmable.

10. The secured system of claim 2, wherein the classifier logic comprises a neural network for learning to generate the raw probability from the combination of metrics.

11. The secured system of claim 1, wherein the fingerprint sensor comprises a swipe sensor.

12. The secured system of claim 1, wherein the fingerprint sensor comprises a placement sensor.

13. The secured system of claim 1, further comprising an access module for granting access to a host system when the spoof probability is within a predetermined range.

14. The secured system of claim 13, further comprising a host system coupled to the access module.

15. The secured system of claim 14, wherein the host system comprises any one of a cell phone, a personal digital assistant, a digital camera, and a personal computer.

16. The secured system of claim 6, further comprising a storage for storing the stored metrics.

17. The secured system of claim 16, wherein the storage is coupled to the spoof detection module over a network.

18. The secured system of claim 6, wherein the stored metrics are encrypted.

19. A spoof detection module comprising:

a metric calculator programmed to receive fingerprint image data and calculate a combination of multiple metrics; and

classifier logic programmed to receive the multiple metrics and determine a spoof probability based on the combination of the multiple metrics.

20. The spoof detection module of claim 19, wherein the multiple metrics comprise any two or more of pixel gray level average and variance of pixels corresponding to a fingerprint ridge, pixel gray level average and variance of pixels corresponding to a fingerprint valley, density of sweat pores, density of streaks, and raw signal level.

21. The spoof detection module of claim 19, wherein the multiple metrics comprise values corresponding to any two or more of ridge elasticity of a fingerprint, electrical properties of a fingerprint, optical properties of a fingerprint, and vitality properties of a fingerprint.

22. The spoof detection module of claim 19, wherein at least two of the multiple metrics are interdependent.

23. The spoof detection module of claim 19, further comprising a probability adjustor programmed to receive the spoof probability and adjust it based on stored fingerprint data to generate an adjusted spoof probability.

24. The spoof detection module of claim 19, further comprising a filter for dividing the fingerprint sensor into multiple windows, wherein the classifier logic is also programmed to determine the spoof probability based on a comparison of the values of the metric in each of the multiple windows.

25. The spoof detection module of claim 24, wherein dimensions of each of the windows is programmable.

26. The spoof detection module of claim 19, wherein the classifier logic comprises a neural network for learning to generate the spoof probability from the multiple metrics.

27. The spoof detection module of claim 19, wherein the fingerprint image data corresponds to data captured using a fingerprint swipe sensor.

28. The spoof detection module of claim 19, wherein the fingerprint image data corresponds to data captured using a fingerprint placement sensor.

29. A method of determining the similarity of a finger to human skin, comprising:

capturing fingerprint image data; and

determining a spoof probability that the finger is real based on a combination of metrics derived from the fingerprint image data.

30. The method of claim 29, wherein the combination of metrics comprise any two or more of pixel gray level average and variance of pixels corresponding to a fingerprint ridge, pixel gray level average and variance of pixels corresponding to a fingerprint valley, density of sweat pores, density of streaks, and raw signal level.

31. The method of claim 30, wherein the combination of metrics includes density of streaks, the method further comprising dividing the fingerprint image data into multiple windows, wherein the spoof probability is based on a comparison of the values of the metric in each of the multiple windows.

32. The method of claim 29, wherein the combination of metrics comprise values corresponding to any two or more of ridge elasticity of a fingerprint, electrical properties of a finger, optical properties of a finger, and vitality properties of a finger.

33. The method of claim 29, wherein at least two of the metrics from the combination of metrics are interdependent.

34. The method of claim 29, further comprising adjusting the spoof probability using fingerprint image data captured during an enrollment step.

35. The method of claim 34, further comprising encrypting the fingerprint image data captured during the enrollment step.

36. The method of claim 34, further comprising determining the spoof probability at a first location and storing the fingerprint image data captured during an enrollment step at a second location, wherein the first location and the second location are coupled over a network.

37. The method of claim 39, further comprising using a neural network for learning to compute the spoof probability from the multiple metrics.

38. The method of claim 29, wherein the fingerprint image data is captured using a swipe sensor.

39. The method of claim 29, wherein the fingerprint image data is captured using a placement sensor.

40. The method of claim 29, wherein a threshold for the spoof probability depends on a level of a transaction.