2005/6/16 by pj 1 hiding biometric data hiding biometric data ieee transactions on pattern analysis...

Post on 18-Dec-2015

217 Views

Category:

Documents

2 Downloads

Preview:

Click to see full reader

TRANSCRIPT

2005/6/16 by pj 1

Hiding Biometric Hiding Biometric DataDataIEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 25, NO. 11, NOVEMBER 2003

Anil K. Jain, Fellow, IEEE, and Anil K. Jain, Fellow, IEEE, and Umut Uludag, Student Member,Umut Uludag, Student Member,

IEEE IEEE

2005/6/16 2by pj

OutlineOutline

IntroductionIntroduction Application scenariosApplication scenarios Skim through data hiding methodSkim through data hiding method Experimental resultsExperimental results

2005/6/16 3by pj

Introduction - Introduction - What’s shortcoming of What’s shortcoming of biometricbiometric The problem of ensuring the security aThe problem of ensuring the security a

nd integrity of biometric data is criticand integrity of biometric data is criticall Example: ID v.s. fingerprintExample: ID v.s. fingerprint

2005/6/16 4by pj

Introduction - Introduction - 8 basic sources of attacks8 basic sources of attacks

Fake biometric

Resubmission of digital stored biometric

Feature detector could be forced to produce feature values chosen by attacker

Synthetic feature setthe matcher

could be attacked to produced high or low scores

Attack databaseChannel

attack

Alter matching result

2005/6/16 5by pj

Skip…Skip…

Encryption v.s. steganographyEncryption v.s. steganography There have been only a few published There have been only a few published

papers on watermarking of fingerprint papers on watermarking of fingerprint images.images.

2005/6/16 6by pj

Application Application scenarios(1/2)scenarios(1/2) The biometric data (The biometric data (fingerprintfingerprint minutiaeminutiae) )

that need to bethat need to be transmitted istransmitted is hidden in a hidden in a host image, whose onlyhost image, whose only function is to function is to carry the data.carry the data. 7th attack7th attack Host: Host: synthetic fingerprintsynthetic fingerprint, face, …, face, …

R. Cappelli, A. Erol, D. Maio, and D. Maltoni, “SyntR. Cappelli, A. Erol, D. Maio, and D. Maltoni, “Synthetic Fingerprint Image Generation,” Proc. 15th Inthetic Fingerprint Image Generation,” Proc. 15th Int’l Conf. Pattern Recognition, vol. 3, pp. 475-478,Sept.’l Conf. Pattern Recognition, vol. 3, pp. 475-478,Sept. 2000. 2000.

Encrypt++Encrypt++

2005/6/16 7by pj

2005/6/16 8by pj

Application scenarios(2/2)Application scenarios(2/2) Hiding facial information (e.g. Hiding facial information (e.g.

eigen-face coefficientseigen-face coefficients) into fingerprin) into fingerprint imagest images Examine fingerprint & faceExamine fingerprint & face

2005/6/16 9by pj

2005/6/16 10by pj

Skim through data hiding mSkim through data hiding methodethod M. Kutter, F. Jordan, and F. Bossen, “Digital SiM. Kutter, F. Jordan, and F. Bossen, “Digital Si

gnature of Color Images Using Amplitude Modugnature of Color Images Using Amplitude Modulation,” Proc. SPIE, vol. 3022, pp. 518-526, 1997.lation,” Proc. SPIE, vol. 3022, pp. 518-526, 1997.

B. Gunsel, U. Uludag, and A.M. Tekalp, “RobusB. Gunsel, U. Uludag, and A.M. Tekalp, “Robust Watermarking of Fingerprint Images,” Pattert Watermarking of Fingerprint Images,” Pattern Recognition, vol. 35, no. 12, pp. 2739-2747, Den Recognition, vol. 35, no. 12, pp. 2739-2747, Dec. 2002.c. 2002.

2005/6/16 11by pj

Skim through data Skim through data hiding methodhiding method WatermarkWatermark

1th scenario: fingerprint minutiae 9-bit1th scenario: fingerprint minutiae 9-bit X[0,N-1], Y[0,M-1], orientattion[0,359]X[0,N-1], Y[0,M-1], orientattion[0,359]

2th scenario: eigenface coefficients 4-byte2th scenario: eigenface coefficients 4-byte Random seedRandom seed Embed watermark : repeat or notEmbed watermark : repeat or not Embed reference bits 0 & 1 ?Embed reference bits 0 & 1 ?

2005/6/16 12by pj

Skim through data hiding mSkim through data hiding methodethod Embedding functionEmbedding function

S : the value of watermark bitS : the value of watermark bit q : embedding strength (q : embedding strength ( 自訂自訂 )) PPAVAV, P, PSDSD: average and standard deviation of neighbor: average and standard deviation of neighbor

hood (ex. 5x5 square)hood (ex. 5x5 square) PPGMGM: gradient magnitude ?: gradient magnitude ? A, B : weightA, B : weight β: maskβ: mask

2005/6/16 13by pj

Skim through data hiding mSkim through data hiding methodethod Decoding functionDecoding function

5x5 cross-shaped neighborhood5x5 cross-shaped neighborhood

2005/6/16 14by pj

Experimental resultsExperimental results

Highlight Highlight decoding accuracydecoding accuracy and and matching performancematching performance

2005/6/16 15by pj

Experimental results -Experimental results - 1th scenario 1th scenario 1th scenario : 1th scenario :

Host : 5 synthetic fingerprint, 5 face, 5 othHost : 5 synthetic fingerprint, 5 face, 5 othersers

5 minutiae data sets, 5 seed keys5 minutiae data sets, 5 seed keys q= 0.1, A = 100, B = 1000 q= 0.1, A = 100, B = 1000 17% stego image pixels are changed17% stego image pixels are changed 100% accuracy100% accuracy

2005/6/16 16by pj

Experimental results –Experimental results – 2nd scenario 2nd scenario 2nd scenario :2nd scenario :

Fingerprint image : 300x300Fingerprint image : 300x300 Face : 150 x 130Face : 150 x 130

14 eigenface coefficients = 56 bytes14 eigenface coefficients = 56 bytes Face database : 4 x 10 face subjectsFace database : 4 x 10 face subjects MaskMask

Minutiae-based: 23x23 blockMinutiae-based: 23x23 block Ridge-based: 3x3 blockRidge-based: 3x3 block

q= 0.1, A = 100, B = 1000q= 0.1, A = 100, B = 1000 640 fingerprint images from 160 users640 fingerprint images from 160 users

2005/6/16 17by pj

Experimental results –Experimental results – 2nd scenario 2nd scenario

Origin fingerprint

Origin face Reconstruct eigenfa

ce

Mask minutiae

Reconstruct fingerprint from

watermarked minutiae-based

image

Mask ridge

Reconstruct fingerprint from

watermarked ridge-based

image

2005/6/16 18by pj

Experimental results –Experimental results – 2nd scenario 2nd scenario

2005/6/16 19by pj

The end…The end…

top related