Beruflich Dokumente
Kultur Dokumente
Abstract: (iometric data )e.!. fin!er print, face ima!e* stored in database s#stem, this +ork mainl# discuss about
preserve the privac# of human face ima!e, i.e. biometric is ph#siolo!ical and behavioural characteristics. $n case of face ima!es, private face ima!e is split into t+o shares )i.e. t+o la#ered ima!e* and stored in database server. isual Cr#pto!raph# is secret sharin! method +here a secret face ima!e is encr#pted into the t+o shares +hich independentl# disclose no information about the ori!inal secret face ima!e. ,he private face ima!e can be retrieve onl# +hen both t+o shares are available durin! the content retrieval process. Content retrieval deal +ith identif#in! secret face ima!e based on biometric information )fin!er print*, durin! this process fin!er print ima!e are needed to retrieve ori!inal secret face ima!e )De-identification*, isual cr#pto!raph# Provide !reat securit# needs as +ell as e.tra la#er of authentication. $n this paper different visual cr#pto!raph# method is used
Keywords- Private face ima!e, /in!er print, Privac#, isual Cr#pto!raph#, De-identification
! "#$%D&C# %"
(iometrics is the automated use of ph#siolo!ical or behavioral characteristic to determine or verif# identit# 011. Ph#siolo!ical biometric is based on direct measurement of a part of the human bod#, such as fin!er-scan, facial-scan, irisscan, hand-scan and retina-scan. (ehavioral biometric is based on measurements and data derived from an action and therefore indirectl# measure characteristic of the human bod# such as voice-scan and si!nature-scan. $dentit# is an individual is a sin!ular, "ni2ue entit#, collo2uiall# or Person, +hich can have more than one identit#. /or E.ample, 3ohn Doe 4i!ht has an email identit# and a +ork identit#. (iometric benefit consists of increasin! the securit# and convenience, reducin! the fraud )or* deliver# of enhanced services. A biometric authentication process operates b# ac2uirin! ra+ biometric data from a sub5ect )e.!. private face ima!e*6 the template of a person in the database is !enerated durin! enrollment process. Protectin! privac# of individual enrolled in a biometric database storin! a transformed biometric template instead of the ori!inal biometric template in the database ,his +as referred to as a private template 021 or a cancelable biometric 071 and three step h#brid approach 081 that combined advanta!e of cr#ptos#stem and cancellable biometric. Apart from these methods various ima!e hidin! approach 091, 0:1, 0;1 have been su!!ested b# researcher to provide anon#mit# to the stored biometric data. Privac# to face ima!es present in surveillance videos0<1.A face de-identification al!orithm 0=1 that minimi>ed the chances of performin! automatic face reco!nition +hile preservin! details of the face such as e.pression, !ender and a!e. A h#brid transformation approach has a !ood cancelable abilit#, but the dra+back of this approach is that a trade-off bet+een performance and securit# is normall# re2uired. ?aive De-identification method is no !uarantee of privac# of people visible in the ima!e. ?either preserves privac# nor data utilit#. /ace s+appin! techni2ue and a!!ressive de-identification the ori!inal face ima!e can be lost 01@1. Proposed method to perform secure face identification b# representin! a private face ima!e +ith inde.ed facial components e.tracted from a public face database. ,he rest of the paper is or!ani>ed as follo+s. $n Section $$ deals +ith visual cr#pto!raph# and its e.tensions are presented. $n Section $$$, +e deal +ith h#brid approach and Section $ , +e deal +ith De-identification method for securin! fin!erprint, and face ima!es and +e conclude this paper in Section .
! V S&A' C$()#%*$A)+(
isual Cr#pto!raph# is a secret-sharin! method that encr#pts a secret ima!e into several shares but re2uires neither computer nor calculations to decr#pt the secret ima!e as sho+ in /i!. 1. $nstead, the secret ima!e is reconstructed visuall#A simpl# b# overla#in! the encr#pted shares the secret ima!e becomes clearl# visible. isual cr#pto!raph# techni2ue for black and +hite ima!es )(asic1* introduced b# ?aor and Shamir 0111. An# visual secret information )pictures, te.t, etc* is considered as ima!e and encr#ption is performed usin! simple al!orithm to !enerate n copies of shares dependin! on t#pe of access structure schemes.
a.
b.
c.
d.
,igure-! a. %riginal mage b. Share- c. Share/ d. Stac0ing the shares reveal the secret isual Cr#pto!raph# scheme represented in computer usin! n . m (asis matrices. 1/2/. ModelA 1. Construct t+o 2*2 basis matrices asA s@B 1 @ @ 1 s1B 1 1 @ @
2. "sin! the permutated basis matrices, each pi.el from the secret ima!e +ill be encoded into t+o sub pi.els on each participantCs share. A black pi.el on the secret ima!e +ill be encoded on the ith participantCs share as the ith ro+ of matri. S1, +here a 1 represents a black sub pi.el and a @ represents a +hite sub pi.el. Similarl#, a +hite pi.el on the secret ima!e +ill be encoded on the ith participantCs share as the ith ro+ of matri. S0. 7. (efore encodin! each pi.el from the secret ima!e onto each share, randoml# permute the columns of the basis matrices S0 and S1. ,his CS ) isual Cr#pto!raph# Scheme* divides each pi.el in the secret ima!e into mB2 sub pi.els. $t has a contrast of (m).mB1 and a relative contrast of (m)B1'2. (asic visual cr#pto!raph# is e.pansion of pi.els. /irst Continuous ima!e )Gra# scale* is converted into halftone ima!e )(inar#* usin! an# halftone techni2ue as error diffusion, thresholdin!, ordered ditherin!, etc. ,he error diffusion !ives better half tone. Each pi.el of half tone is broken into some sub-pi.els as sho+n in ,able 1. #able - Construction o3 /-out-o3-/ scheme
Dandom Pro5ection
CDP ,ransform
Hashin!
Database
Hashin!
4atchin!
#rans3orm-based Approach
$nput $ma!es
/AC,ID$JA,$I?
?on-$dentit# /actors
$dentit# /actors
DE-identification
DECIG?$,$I?
DECI?S,D"C,$I?
Figure 1
De-$dentified $ma!es
V! C%"C'&S %"S
$n this paper, surve#s of isual Cr#pto!raph# and /ace De-identification methods have been presented. (ased on the above summar#, three step h#brid approach consists of three parts are namel# cancelable transform, discriminabilit# enhancement transform and template protection. Each part provides the cancelable abilit#, discriminabilit# and securit#, respectivel#. $n /ace De-identification method, multi factor model preservin! data utilit# and multi-factor de-identification al!orithm described here operates on sin!le ima!es. (ut biometric private face ima!e method to protect the privac# of a face database b# decomposin! an input private face ima!e into t+o independent sheet ima!es such that the private face ima!e can be reconstructed onl# +hen both sheets are simultaneousl# available. GE CS is then used to encr#pt the private ima!e in the selected host ima!es. $t is observed that the reconstructed ima!es are similar to the ori!inal private ima!e. Since the# have respective advanta!es and ma# complement each other. 4an# research need to be done in isual Cr#pto!raph#.
$4,4$4"C4S
011 A. 3ain, P. /l#nn, and A. Doss, Handbook of Biometrics. ?e+ KorkA Sprin!er, 2@@;. 021 G. $. Davida, K. /rankel, and (. 3. 4att, FIn enablin! secure applications throu!h off-line biometric identification,G in Proc. IEEE S m!. Sec"rit and Pri#ac , 1==<, pp. 18<L19;. 071 ?. Datha, 3. Connell, and D. (olle, FEnhancin! securit# and privac# in biometrics-based authentication s#stems,G IB$ S st. %., vol. 8@, no. 7, pp. :18L:78, 2@@1. 081 K. /en!, P. Kuen, and A. 3ain, FA h#brid approach for face template protection,G in Proc. SPIE &onf. Biometric 'echno(o) for H"man Identification, Irlando, /M, 2@@<, vol. :=88. 091 A. 3ain and ". "luda!, FHidin! biometric data,G IEEE 'rans. Pattern *na(. $ach. Inte((., vol. 29, no. 11, pp. 18=8L18=<, ?ov. 2@@7. 0:1 3. Don! and ,. ,an, FEffects of +atermarkin! on iris reco!nition performance,G in Proc. 10th Int. &onf. &ontro(, *"tomation, +obotics and ,ision, -00. (I&*+&, -00.), 2@@<, pp.119:-11:1. 0;1 ?. A!ra+al and 4. Savvides, F(iometric data hidin!A A 7 factor authentication approach to verif# identit# +ith a sin!le ima!e usin! ste!ano!raph#, encr#ption and matchin!,G in Proc. &om!"ter ,ision and Pattern +eco)nition /orksho!, 2@@=, vol. @, pp. <9L=2. 0<1 E. 4. ?e+ton, M. S+eene#, and (. 4alin, FPreservin! privac# b# de-identif#in! face ima!es,G IEEE 'rans. 0no1(. 2ata En)., vol. 1;, no. 2, pp. 272L287, /eb. 2@@9. 0=1 D. Gross, M. S+eene#, /. De la ,orre, and S. (aker, F4odel-based face de-identification,G in IEEE /orksho! on Pri#ac +esearch in ,ision, Mos Alamitos, CA, 2@@:. 01@1 D. (itouk, ?. Numar, S. Dhillon, P. (elhumeur, and S. N. ?a#ar, F/ace s+appin!A Automaticall# replacin! faces in photo!raphs,G *&$'rans. 3ra!h., vol. 2;, no. 7, pp. 1L<, 2@@<. 0111 4. ?aor and A. Shamir, F isual cr#pto!raph#,G in Proc. E4+5&+6P', 1==8, pp. 1L12. 0121. A. ,eoh, A. Goh, and D. ?!o, FDandom 4ultispace Ouanti>ation as an Anal#tic 4echanism for (ioHashin! of (iometric and Dandom $dentit# $nputs,G IEEE 'ransactions on Pattern *na( sis and $achine Inte((i)ence, vol. 2<, no. 12, pp. 1<=2-1=@1, Dec. 2@@:. 0171. K C /en! and P C Kuen, FClass-Distribution Preservin! ,ransform for /ace (iometric Data Securit#,G Proceedin)s of IEEE Internationa( &onference on *co"stics, S!eech, and Si)na( Processin) (I&*SSP), pp. 181-188, 2@@;. 0181. K C /en! and P C Kuen, FSelection of Distin!uish Points for Class Distribution Preservin! ,ransform for (iometric ,emplate Protection,G Proceedin)s of IEEE Internationa( &onference on Biometrics (I&B), 2@@;. 0191. D. M. Divest, F,he 4D9 4essa!e-Di!est Al!orithm,G +7&18-1, 9et1ork /orkin) 3ro"!, $I' :aborator for &om!"ter Science and +S* 2ata Sec"rit , Inc., 1==2. 01:1. Arun Doss, and Asem Ithman, ; <,9o. 1,$arch -011. isual Cr#pto!raph# for (iometric Privac#, G IEEE 'rans, 5n information 7orensics and Sec"rit ,,o(.