Watermarking model in database systems

Thủy vân trôn ảnh và trôn âm thanh đạt đuọc nhiểu kết quả, ở nước ngoài và trong nước. Figure 1.[r]

WATERMARKING MODEL IN DATABASE SYSTEMS

T u a n D o T r u n g

College o f Science, V ie tn a m N a tio n a l University, H anoi

T a o N g u y e n V a n

University o f Thai N g u yen

A b stra ct The paper aims at a watermarking model in databases, for copyright management and embedding information in data The model focuses on numeric data type and proposes solutions for the other multimedia data type such as text, image and video

Keywords : Model, watermark, database

1 I n t r o d u c t i o n

W a te r m a r k in g process is s tu d ie d in m ore t h a n ten y ears, for th e neecs of copyright protect, d a ta pro cessin g in th e larg e d a ta e n v iro n m e n t d is tr ib u te ! in com puter netw ork W a te r m a r k in g is in fo rm a tio n em b ed d in g in d a ta in keeping the data qualify a n d se cre t of em b ed ed d a ta [3, 4, 5, 6], T hey apply w a te rm a rk in g techniques for (l) a u th o r in g of d a ta ; th e y b a se on em b ed d in g in fo rm a tio n in dala in order to d e te rm in e th e ow ner of d a ta ; (ii) h id in g in fo rm atio n in d ata; th e y bas> on lim ited d a ta modification, b u t n o t c h a n g in g u sin g qualify of d a ta a n d n o t ap pealing embedded inform ation

R esearch r e s u lts focused on th e te c h n iq u e s on im ages, such as (i) F i r ie tra n s fo rm a tio n of sig n a ls by f a s t F u r ie tr a n s fo r m (FFT); discrete cosine transform (DCT); (ii) w avelet analysis; d isc re te w av ele t tr a n s fo r m (DWT); (ill) modification by

s in g u lar value decom position (SVD) for em b e d d in g info rm atio n on im age d a t' [4

5] The im ages h av e a lot of color levels, a re r e p r e s e n te d u n d e r different compressed sta n d a rd s The problem con cern in g w a t e r m a r k i n g on m u ltim e d ia d a ta have not studied yet, is proposed as a r e s e a r c h one [1], The w orking group of Institute of Inform ation Technology, V ie tn a m N a tio n a l I n s t i t u t e on Technology and Sciei'ce

h as obtained some re su lts

W a te r m a r k in g process h a s c h a r a c te r is tic s [2]:

- In fo rm atio n q u a n tity e m b e d d in g in d a ta a s s u re s t h a t th e modificatioi of

d a ta is a t p e rm itte d level T he level is in th e accepted in te rv a l of dita error The in fo rm a tio n IS so sm a ll t h a t i t does not influGnce to using quaify

of data U sers can not differ a tt a c h e d in fo rm atio n a n d m ay not take he

in fo rm atio n w h en n o t a p p ly in g w a te r m a r k in g algorithm Althourh

w a te r m a r k e d d a ta allow u s e r s to copy d a ta file in a s s u r in g th e data owrẻr

T his c h a r a c te r is called im p e rcep tib ility ; th e original d a ta should be unno ticeab le to th e h u m a n observer, i.e w a te r m a r k s should not interfire

w ith th e p ro tected m edia;

33

34 Tuan D oTrung, Tao N gu yenV an

W a t e r m a r k e d d a ta are c o n sisten t, i.e th e y keep th e s e m a r k s e ith e r in d a t a

m a n ip u la tio n o p e ra tio n s or in a tta c k s for ta k in g w a t e r m a r k e d in fo rm atio n from d a ta T h is c h a r a c te r is called ro b u stn ess; u n a u th o r iz e d one should n o t

be ab le to destroy th e w a te r m a r k w ith o u t technique; w a t e r m a r k should be

r o b u s t to common signal processing a n d in te n tio n a l a tta c k s ;

W a t e r m a r k e d in fo rm a tio n can be ta k e n by w a te r m a r k decom position It allow s to o b tain h id d e n in fo rm atio n from d a ta flow, is u se d in d a ta security

C oncerning in w a te r m a r k in g are tech n iq u e as follows :

E m b e d d in g in fo rm a tio n into d ata;

D e tec tin g w a t e r m a r k in d ata;

W a t e r m a r k decomposition

W a te rm a rk in g re s e a rc h on im age, sound in some y e a rs ago h a s some resu lts,

•n [f» 6] How ever w a te r m a r k in g resea rch in d a ta b a s e s y s te m s w as in beginning [2] proposed c e rta in p rin cip al r e q u ire m e n ts a b o u t modeling a n d tec h n iq u e s In [7] there is a h a s h in g tec h n iq u e for th e purpose

The p a p e r aim s a t a developing th e tec h n iq u e proposed in [7], th e n p re s e n ts a lution for in te g r a tin g some w a te r m a r k in g techniques in a model allowing to erraark in m u ltim e d ia d a ta W hen a p a rtic u la r m u ltim e d ia d a t a m odel does not xi" ted yet, a re la tio n a l d a ta b a s e m a n a g e m e n t system is u se d as rep la ce d one The luticn in th e p a p e r focuses w a te r m a r k in g in (i) document; (ii) video d a ta

n'he r e s t of th e p a p e r : (i) th e second p a r t p re s e n ts r e q u ir e m e n ts a n d principal

pects of d ig ita l w a term a rk in g in th e rela tio n a l data m odel; (ii) th e third part IS rtipOied model and a solution for document and video data; (iii) the la st part is

c o n c l u s i o n a n d some re m a rk s

2 W a term a rk in g in r e la t io n a l d a t a b a s e s y s t e m s

The re la tio n a l d a ta model allows to organize d a ta by tu p le s, lin e s in two dim^rsion table D om ain d e te rm in e s r e p r e s e n ta tio n of a t t r i b u t e s P o p u la r d a ta tyofsm th e tab le a re c h a r a c te r a n d n u m b er Some d a ta b a s e m a n a g e m e n t system s

IL ojv to u se other ty p es of d ata, as m u ltim ed ia data (docum ent, im a g e, sound and

ldfO W a te r m a r k in g in re la tio n a l d a ta m odel is for c o p y rig h t p ro tectio n and ij-fcrriation hiding

Comparing th e difficulties in w a te r m a r k in g process w ith co n v en tio n a l d a ta tTpi with m u ltim e d ia d a ta type, in [2] are some r e m a r k s (i) m u ltim e d ia objects

h i \ e a lot of d ata; th e r e is d a ta r e d u n d a n t U s e rs are not co m fortable in selecting a lfCitpn for in fo rm a tio n hiding; therefore a tuple is se e n as a n object in

a e'm arking process; (ii) m u ltim e d ia d a ta m odification is a difficult one because of

m piex str u c tu r e s of m u ltim e d ia d ata; th e n decomposing h id in g in fo rm a tio n from

<£ta is difficult In re la tio n a l d a ta model, d eletin g a tu p le m ay d estro y w a te r m a r k ils)

C o n cern in g d ig ita l d a ta in a relatio n al d a ta b a s e system , [7, 9] pro])osod

w a te r m a r k in g a lg o r ith m The alg o rith m b ases on c o n te n t c h a ra c te rs of d ig ita l dat [7] and m e t a d a t a on t h e s e d a ta [9]

Dữ liệu vãn bân số hình ảnh âm thanh đoạn vỉdeo

i

7

Sinh viẻn Tẻn Tuổi ảnh Phát biểu VIDEO

ị &

Phuong án thủy vân trong

cơ sỏ dữ liệu quan hệ [7, 8, 9] trôn dữ liệu số đả đạt được một số kết quả

Đa số [2,3,4] cho rằng thủy vân trôn văn bản không bền

tức dẽ bị phát hiện Phần sau sẻ đế xuất giải pháp.

, 1 _

Thủy vân trôn dữ liệu video chưa đưọc nghiên cứu Phđn

sau sẻ đổ xuất giải pháp

Thủy vân trôn ảnh và trôn

âm thanh đạt đuọc nhiểu kết quả, ở nước ngoài và trong nước

Figure 1 W a t e r m a r k i n g w ith some kind of d a ta type in re la tio n a l d a t a laodel

2.1 A lg o r i th m f o r in fo rm a tio n e m b e d d in g in d a t a

The a lg o r ith m r u n s w ith th e a tt r ib u t e Ai in which is w aterm arking

Em bedding in f o r m a tio n affects to certain in fo rm atio n b it am o n g b its o f Dixi' j-y

re p re s e n ta tio n of a t t r i b u t e value Ai The process is r e p e a te d on tu p le s 3Í relatio n al tab le

In p u t to th e re la tio n a l table R(A1, A2, A n)

While n o t e x ists a tuple, do

I f ( r a n d o m (0,1) < threshold a) A N D (value A i ị N U L L ) then

T ra n sfo rm (va lu e A i) to binary value

Select c erta in b it fr o m the binary value

A tta ch selected b its into behind attribute value The a ttrib u te is ',ùll>d

w a term a rk in g a ttr ib u te

2.2 A lg o r i th m f o r w a te r m a r k in g d e te c tin g

W hen s u s p e c tin g som e d a ta a tta c h in g so m e th in g wrong, they d e m a n d (ì) 0

Because the h id d e n in fo rm a tio n does not affect d a ta qualify in an applicatior t IS not necessary to rec o v er t h e original d ata The w a te r m a r k detection alg o rith n vstg

(i) th resh o ld b, a llo w in g to detect w a term ark ; b is corresponding to threshold r (i)

th resh o ld c, a llo w in g to re fin in g experience values affecting to th e thresh o ld b

With the su sp ic io u s table R (A l, A2,, A n)

T l = n u m b e r o f su sp ic io u s line

I n i t a c o u n te r T 2 = 0

36 Tuan DoTrung, Tao N gu yenV an

W h ile exists a line o f the table, do

I f the value o f A i Ỷ N U L L , then

- T ra n sfo rm in g A i value to b in a ry value

■ E x tra c tin g certain bit (as in the w a term a rk in g a lg o rith m presented)

- I f these bits m a tch es the last p a r t o f attribute value o f

w a term a rk a ttr ib u te , let T 2 = T2 + 1

- I f T 2 / T1 > (a + c)/2 A N D a > c th en to inform "w aterm ark exists"

E lse i f T2 / T1 > d 2 A N D a <c then to in fo rm U w a term a rk existsU

E lse to in fo rm "w aterm ark does not exist".

Two a lg o rith m s le t us to d e te rm in e t h e r i g h t ow ner of the d a ta in th e

re la tio n a l d a ta b a s e T h e model proposed in n e x t p a r t sh o u ld develop th e alg o rith m s for d ig ital w a te r m a r k in g on complex d a ta ty p e of d a ta b a s e

3 M o d e l p r o p o s e d

Below are some a s s u m p tio n s in th e p ro p o se d m odel for w a te r m a r k in g in a

d atab ase :

- R e la tio n a l d a ta b a s e s h ave ta b le s R (A l, A2, An), in which a ttr ib u te s accept n u m b e r value, c h a r a c te r v alu e, so u n d , im a g e or video The tab le h as

N tu p le s (lines of table);

D a ta b a s e composes of conventional d a t a a n d im ag e, sound and video d ata

T h is a s s u m p tio n allows w a t e r m a r k in g is re a liz e d directly in d a ta , b u t not

in t h e ir r e p r e s e n ta tio n on e x te rn a l files w hich do n o t belong to d a ta b a se ;

A video clip is decomposed into key frani6S Some of th em IS d e m a n d ed in

im ag e w a te r m a r k in g process; i.e a p ro b lem of w a te r m a r k in g on video d a ta

is solved by w a te r m a r k in g on image;

- C o n cern in g digital w a te r m a r k in g , som e n o ta tio n s are (i) th re sh o ld a

p erm its to s e ls c t w a term a rk in g data; a t ca u se of experiencGS, le t the

th r e s h o ld a to se t to a p p ro x im a te v a lu e of 0.25; (ii) th re sh o ld b p e rm its to

d e te rm in e a w a te r m a r k in g tuple (line of tab le) W ith b = 0.12, i.e 12 lines

a re selected am o n g 100 lines It c o rre sp o n d s to th e probability to find a

w a t e r m a r k i n g line is 0.12; (iii) a h a s h i n g fu n ctio n h(x) p e rm its to select

c e r ta in b it am o n g th e b in a ry p r e s e n t a ti o n of a n u m b e r a ttrib u te ;

- Know ledge on d a ta , i.e m eta d a ta is u se d to s u p p o r t info rm atio n q u an tity

h id in g in d a ta The m e ta d a ta is com posed of c h a ra c te ristic s a b o u t (i)

a t t r i b u t e v alu es of th e rela tio n (table of d a ta b a s e ); (li) expectation value TB

(a v era g e va.lu6)j lGt to know in fo rm a tio n q u a n tity hidden in dãta.Ị (ill)

in fo rm a tio n q u a n tity TT, p e rm its to kn o w c la rity level of a t t r i b u t e values

in co rre sp o n d in g to a t t r i b u t e se m a n tic , i.e th e conception t h a t th e

a t t r i b u t e h a s to p re s e n t

W a te r m a r k in g in d a ta b a s e sh o u ld resp o n d th e effect as following :

- In secret; th e u s e r s can n o t feel th e existence of h id in g in fo rm a tio n in

data N orm al; d a ta m a n ip u la tio n can re a liz e norm ally;

- R obust; w a t e r m a r k in fo rm a tio n is co n siste n t w ith d a ta m a n ip u la tio n

o p eratio n s;

- Security; only a u th o r iz e d u s e r c a n decompose h id d e n in fo rm a tio n from

d a ta ;

w a te r m a r k in g in fo rm a tio n ;

- W a te r m a r k in g process allows u p d a tin g o p e ra tio n on data;

- W a te r m a r k in g com m u n ication; h id d en in fo rm a tio n is transm itting in data

flow, on d a ta c o m m u n icatio n

Q u an h ệ đáu vào

R (A 1, A2 A n)

-►

Lâp lai dối Xấc định với c á c Ai các thuộc

tính thủy vân (Ai, Ạj)

Xác định hàm băm Thủy vân dữ liệu số

Thủy v ân trên van bản

Thủy vân th eo m ột s ố kĩ thuật SDV, biốn đổi cosin,

Thủy vân m ột s ố khung ảnh chinh trong đ o ạ n video

Lạp lại đổi với c ác dồng dử liệu c ủ a R

Figure 2 W a te r m a r k in g m odel for d ifferent d a ta types in d a ta b a s e

3.1 W a t e r m a r k i n g m o d e l

W a te r m a r k in g process in d a ta b a s e sy s te m s is rea liz ed a fte r two levels : (i)

w a te r m a r k in g for co pyright pro tectio n ; h id d e n in fo rm a tio n h av e not sem antic content; (ii) w a te r m a r k in g for d a ta a u th o r iz in g a n d d a t a hiding; h id d en inform ation

is decomposed a n d it h a s s e m a n tic c o n ten t; i.e it h a s o th e r u sin g role

bảng quan

Lựa chọn thuộc Unh Ihùy vàn

Tri thức vồ dử liệu

TB.TT

Khi cán giấu tim

m ậl trong dử liệu tạo thống tin giâu

trong dữ liệu

í>

c

Thủy vân, lức thay đổi

dữ liệu bằng thông tin nhúng vào

bang quăn

hệ R2 đâ thùy vân

Figure 3 W a te r m a r k i n g in d a ta b a s e w ith th e re la tio n a l one

38 Tuan DoTrung, Tao NguyenVan

A fter received d a ta in d a ta b a s e , u se rs a re informed t h a t exists or n o t lidden

i n f o r m a t i o n i n th e s e d a ta W a te r m a r k detection process b ases on t h e knowledge

a b o u t d a ta i.e m e t a d a t a via expectation value TB and info rm atio n q u a n tity IT

I b ả n g q u a n

ị h ê R

bảng quan

hệ R2 nghi

có thủy vân

í >

í >

p h á t hiện

P h á t hiên th ô n g tin thùy vân giẫu trong

d ữ liệu

Khi c á n g iấu tím

m ặ t trong d ữ liốu

■=>

= >

D ữ liệu R và R2

là n h ư n h a u ?

Dữ liệu R v à R2

là n h ư n h a u ? (và R2 có th ông tin giấu ?)

Figure 4 W a te rm a rk detection process

3.2 W a te r m a r k in g on d i g i t a l d a ta

Em bedding inform ation into digital data dem ands som e assum ptions as follows:

- D a ta b a se a n d know ledge a b o u t d a ta (m etad ata), TB an d TT, for em bedding

h id d en in fo rm a tio n in to data;

H a sh in g function h e re is h(X, N, M) on bit strin g X; in which (i) X is binary

p re s e n ta tio n of a n a tt r ib u t e value; N is n u m b er of th e first selected; M is

n u m b er of selected bit a t th e end of strin g of N bit W hen th e re is not enough N b it form X, bit 0 can be in se rte d M is a m ark bit, is selected successively or a fte r o th er selection m an n er;

While w a te r m a r k in g w ith th e a tt r ib u t e A l, select th e nex t a tt r ib u t e A2 in the sch em a R as w a te r m a r k in g a ttr ib u te The selection is su itab le with th e

re q u ire m e n t of only copyright protection W hen needing to hide secret inform ation, th e y use knowledge TB and TT in o rder to cre ate a secret inform ation D a ta o b tain ed w ith h id d en inform ation h a s to belong to the

a ttr ib u te dom ain; it a s s u r e s t h a t d a ta have only accepted error Hidden inform ation is em bedded in the a tt r ib u te value A2;

The th re s h o ld a, b a n d c used for w a te rm a rk in g a n d w a te r m a r k detection are experience n u m b er A t the beginning, m ay be a=0.25, b^o.12 and

c=0.12.

- The em b ed d in g function nh (Y, TBi, TTi) is for to c re ate secret inform ation The function b a ses on th e in fo rm a tio n Y th e u se rs w a n t to embed in to data The know ledge TBi and TTi ab o u t th e a ttr ib u te v alu e Ai are n ecessary to create a se cre t in fo rm a tio n corresponding to Ai value

Input the table R(A1, A2, A n)

Input the know ledge about d ig ita l {TBi, TTil

While exists a d a ta tuple, do

I f (ra n d o m (0,1) < threshold a) A N D (value A i ị N U L L ) th en

1 I n i t the line counter k =1

2 T ra n sfo rm (value Ai) into the binary fo rm X

3 N l = length (X); N is created from N l ; M - N 14'

4 R u n the function h(X, N, M) h a v in g the m a rk T V

5 D eterm ine a w a term a rkin g a ttrib u te A j'

6 M a r k in g the m ark T V at the e n d o f the a ttrib u te value A j

7 I f they w a n t to hide a secret inform ation, then Generate a embedding information with the help o f nh (Y, Tbi Tti)-Select a em b ed d in g inform ation for the current d a ta l i n e ■ thp lino (tuple) is noted N H k ;

In se rt N H k into the w a term a rk in g a ttribute value Aj.

The w a te r m a r k detection process uses th e th re sh o ld c I n p u t of th e process is original d a ta b a s e , h a v in g th e table R, and a suspicious d a ta b a s e , h a v in g th e table R2 R esu lt of th e process is yes/ no about the sim ila r betw een two d a ta b a s e s W hen exists hidden in fo rm atio n , let them to decompose th e se cre t inform ation

In p u t the table R (A l, A2, An)

I n p u t the suspicious table R 2 (A l, A2„ A n)

In p u t the values N, M used in w a term a rkin g alg o rith m

In p u t the know ledge about data, TB T T

Let T I = n u m b e r o f suspicious data lines

I n it the counter T2 - 0

While exists a line in the table R2, do

I f value A i ị N U L L then

1 D eterm ine w a term a rkin g a ttribute A j

2 Let X = binary form o f the value A i

3 R u n the function (X, N, M), obtain T V

4 C ompare T V w ith M last it o f the a ttrib u te value Aj I f results is

m atching, then increase the counter T2 = T2

+1-I f (T 2 / T1 > (a +c)/2 A N D a > c) OR (T2 / T1 > d 2 A N D a < c) then to inform

"watermarking exists"

Else to in fo rm "w aterm arking does not exist".

In th e case of w a te r m a r k in g existing, they can decompose hidden information from th e w a te r m a r k in g a ttr ib u te

40 T uan DoTrung, Tao N guyenV an

Ì.Ì W a te r m a r k in g on im age, so u n d a n d vid eo

3ỉ.sing on physical p re s e n ta tio n of im a g e a n d sound d a ta , th e y have a bit .nitric presentation, A W a te rm a rk in g process on th e s e type of d a ta h a v e achieved

a ot, p trm its to tra n s fo rm A into A’ In th e d a ta b a s e model, th e m eth o d SVD [5] is

pnpo^ed.

For video d a ta , th e model re-uses th e te c h n iq u e of im age w a te rm a rk in g , w ith only certain fram es; i.e w ith key fram es of video clip

Figure 5 Re-use th e im age w a t e r m a r k i n g tec h n iq u e for video d a ta

3.4 W a te rm a r k in g on text

A larg e p a r t of d o cu m en ts is w ith th e p r e s e n t a ti o n of c h a ra c te rs Very m uch ef'ort on do cu m en t w a te rm a rk in g , b u t t h e r e is n o t c o n siste n t a n d it is difficult to hide information I t is easy to find a w rong s e n te n c e in docum ents

For th e tex t type of d a ta , th e pro p o sed m odel b a ses on know ledge of dccunents to g e n e ra te (i) re p e a te d c h a ra c te rs ; (ii) new c h a ra c te rs The location for

w atem arking in th e d o cu m en ts is d e te r m in e d by a location function A simple solution for th e function is to select th e fixed po sitio n s in th e docum ent For example positions in th e docum ent s t a r t a t i, i+k, i+ 2 k , At th e selected position, a new character is in s e rte d or in te g ra te d into e x is tin g c h a ra c te r

4 C o n clu sio n a n d r e m a r k s

The p a p e r p r e s e n ts w a te r m a r k in g m odel in d a ta b a s e sy stem s A rela tio n a l cUtahase m a n a g e m e n t is re p r e s e n ta tiv e one T h e m odel p e rm its w a te r m a r k in g a n d waterm ark detection on different k in d s of m u ltim e d ia d a ta The image

w aterm arking is worth, is p re s e n te d in a lot of re s e a rc h e s

The th re sh o ld p a ra m e te r s , as a, b, c, N a n d M a re experience p a ra m e te r s In

W a te r m a r k in g model should be a p p lie d for d a ta b a s e for t r a i n in g purpose Then they need (i) p a r a m e te r s for w a te r m a r k in g ; (ii) e v a lu a tio n a b o u t consistency

of w a te rm a rk in g process; (iii) w a te r m a r k in g pay-off T hese r e m a r k s should be discussed in o th e r paper

P a p e r a u th o rs p r e s e n t th e ir a ck n o w le d g e m e n t to th e group on w a te r m a r k in g

in I n s t i tu t e of In fo rm atio n Technology, V ie tn a m N a tio n a l I n s t i t u t e on Technology ar.d Sciences for scientific m ateria ls

Phân đoạn, chọn các khung chính

thủy vân ảnh đối với các khung hinh chính các ảnh

1 T uan DoTrung, Cuong LuongXuan, Khun Piseth, Tao NguyenVan, Abcut

video data m anipulation, N a tu ra l sciences a n d technology Journal, VhU,

VLDB Conference, 2002.

3 M andal p., T hakral A., Verma s., W aterm ark based digital riglts

management, Proc of the int conference on Information Technology : (odiig

a n d Computing, 2005.

4 P arask ev i Bassia, Ioannis Pitas, Robust audio w aterm aking in thf tin e

domain, IE E E transaction on M u ltim e d ia , Vol 3, No 2(2001)

rightful ownership, IE E E transaction on M ultimedia, Vol 4 No.l(ỈOOỈ),

p 121- 128

6 S anka Basu et Al., Introduction to the special issue on m ultimedia datibase,

IE E E transaction on M ultim edia, Vol 4, No.2(2002), p 141- 145.

7 Yong Zhang et al., Relational d atab ases w aterm ark technique basỉd on

content characteristic, Proc o f the first Conference on Computirẽp

Information a n d Control, 2006.

8 Yong Zhang, Bian Yang, Xia-mu Niu, Reversible w aterm arking for realtioral

d atab ase authentication, Journal o f Computers, V 17, N.2(2006) p 59 - 66

d atab ases copyright with cloud w aterm ark , Transactions on engin&rirg,

computing a n d technology, V.3(2004), p 170 - 174.