Teacher : Hsien-Chu Wu Student : Hsiao-yun Tseng, Chen-ying Lai Speaker : Hsiao-yun Tseng Date...

21
Teacher Hsien-Chu Wu Student Hsiao-yun Tseng, Chen-yi ng Lai Speaker Hsiao-yun Tseng Date May 10, 2006 Database Temper Detection Techniques Based on Digital Watermarking

Transcript of Teacher : Hsien-Chu Wu Student : Hsiao-yun Tseng, Chen-ying Lai Speaker : Hsiao-yun Tseng Date...

Page 1: Teacher : Hsien-Chu Wu Student : Hsiao-yun Tseng, Chen-ying Lai Speaker : Hsiao-yun Tseng Date : May 10, 2006 Database Temper Detection Techniques Based.

Teacher: Hsien-Chu Wu

Student : Hsiao-yun Tseng, Chen-ying Lai Speaker: Hsiao-yun Tseng

Date :May 10, 2006

Database Temper Detection Techniques

Based on Digital Watermarking

Page 2: Teacher : Hsien-Chu Wu Student : Hsiao-yun Tseng, Chen-ying Lai Speaker : Hsiao-yun Tseng Date : May 10, 2006 Database Temper Detection Techniques Based.

2

Outline

IntroductionTwo proposed MethodsMethod I – for binary watermarksMethod II – for grey watermarks

Experimental results

Conclusions

Page 3: Teacher : Hsien-Chu Wu Student : Hsiao-yun Tseng, Chen-ying Lai Speaker : Hsiao-yun Tseng Date : May 10, 2006 Database Temper Detection Techniques Based.

3

Introduction

Data warehouse

Market analysis

innocent victim

payment

Stop

The purposes of database watermarking:

avoid data destroyed and tort

malicious damage

Page 4: Teacher : Hsien-Chu Wu Student : Hsiao-yun Tseng, Chen-ying Lai Speaker : Hsiao-yun Tseng Date : May 10, 2006 Database Temper Detection Techniques Based.

4

Produce the certification code

Database

PAtt.A1

Att.A2

…Att.AN

1 A11 A12 … A1N

2 A21 A22 … A2N

3 A31 A32 … A3N

… … … … …

feature C

watermarkWM'

⊕certification

code SKey

DB SD+

Internet

SDPKey

certification code

PAtt.A1

Att.A2

…Att.AN

1 A11 A12 … A1N

2 A21 A22 … A2N

3 A31 A32 … A3N

… … … … …

T

T feature C'

⊕ watermarkWM''

verification process

Page 5: Teacher : Hsien-Chu Wu Student : Hsiao-yun Tseng, Chen-ying Lai Speaker : Hsiao-yun Tseng Date : May 10, 2006 Database Temper Detection Techniques Based.

5

Method I – for binary watermarks

Produce the certification code(1/4)

1 0 0 1 0 1 ..

WM

WM = ECC(WM ,S ) WM

S=M×N

M: Total tuples of the database N: Total attributes of the database

Page 6: Teacher : Hsien-Chu Wu Student : Hsiao-yun Tseng, Chen-ying Lai Speaker : Hsiao-yun Tseng Date : May 10, 2006 Database Temper Detection Techniques Based.

6

Method I – for binary watermarks Produce the certification code(2/4)

group

NO. B_name Dateamoun

tprice

1Harry potter

11/10

100 200

2The Da Vinc

e Code 12/1

5150 400

3Digital fort

ess12/1

6240 410

4Little

women12/1

7260 380

5The world

is Flat12/1

6240 200

… … … … …

2005log = 5×200 mod 2

= 1000 mod 512= 4885

ANS. Ai,j = 488200

Ai,j =Pi Ai,j mod 2

jiiAP ,log

fetch the feature

Ci,j

Page 7: Teacher : Hsien-Chu Wu Student : Hsiao-yun Tseng, Chen-ying Lai Speaker : Hsiao-yun Tseng Date : May 10, 2006 Database Temper Detection Techniques Based.

7

Method I – for binary watermarks Produce the certification code(3/4)

Ci,j(Ai,j)

log21log2 , 1

1log 2 0 , 0

,,

,

,,

jiiji

jii

jiiji

APAAP

APA

Ci,j(Ai,j) =

24882 , 1

2488 0 , 098

8 ( 不成立 )( 成立 )

ANS. Ci,j(Ai,j) = 1 NO. B_name Date

amount

price

1Harry potter

11/10 100 200

2The Da Vince

Code 12/15 150 400

3Digital fortes

s 12/16 240 410

4Little

women12/17 260 380

5The world

is Flat12/16 240 200

… … … … …

Feature

1

Page 8: Teacher : Hsien-Chu Wu Student : Hsiao-yun Tseng, Chen-ying Lai Speaker : Hsiao-yun Tseng Date : May 10, 2006 Database Temper Detection Techniques Based.

8

Method I – for binary watermarks Produce the certification code(4/4)

WM

feature C

⊕certification code AK

0 0 0 1 0 0 ..

1 0 0 0 0 1 ..

SKey

Database

SD+

Internet

1 0 0 1 0 1 ..

Page 9: Teacher : Hsien-Chu Wu Student : Hsiao-yun Tseng, Chen-ying Lai Speaker : Hsiao-yun Tseng Date : May 10, 2006 Database Temper Detection Techniques Based.

9

Method I – for binary watermarks The verification process of database integrity

Database SD+

Internet

SD

PKey

certification code AK

T group fetch the feature

0 0 0 1 0 0 ..

feature C

1 0 0 0 0 1 ..

1 0 0 1 0 1 ..WM

Ai,j* =Pi

*Ai,j mod 2

jii AP ,*log

Ci,j (Ai,j*)

log

21log

2 , 1

1log

2 0 , 0

,**,

,*

,**,

jiiji

jii

jiiji

APA

AP

APA

Page 10: Teacher : Hsien-Chu Wu Student : Hsiao-yun Tseng, Chen-ying Lai Speaker : Hsiao-yun Tseng Date : May 10, 2006 Database Temper Detection Techniques Based.

10

Method I – for binary watermarks Experimental results(1/2)

NO. B_name Dateamoun

tprice

1Harry potter

11/10 100 200

2The Da Vinc

e Code 12/15 150 400

3Digital fort

ess 12/16 240 410

4Little

women12/17 260 380

5The world

is Flat12/16 240 200

Emulation experiment

Database : the trade of the network bookstore table : Book tuples : 10000 attributes : Book(NO., B_name, Date, amount, price)

the amount of values‧ : 100004 =40000

Page 11: Teacher : Hsien-Chu Wu Student : Hsiao-yun Tseng, Chen-ying Lai Speaker : Hsiao-yun Tseng Date : May 10, 2006 Database Temper Detection Techniques Based.

11

Method I – for binary watermarks Experimental results(2/2)

original watermark

Exp. 1 : revise the front 500 B_name

Exp. 2 : revise the last 500 price

Result : find out 390 values to be destroyed

Result : find out 498 values to be destroyed

The accuracy rate is 78%. The accuracy rate is 99%.

Page 12: Teacher : Hsien-Chu Wu Student : Hsiao-yun Tseng, Chen-ying Lai Speaker : Hsiao-yun Tseng Date : May 10, 2006 Database Temper Detection Techniques Based.

12

Method II – for grey watermarks Produce the certification code(1/3)

watermark WM with size ×

N N

N: total tuples of the database

original watermark

PAtt.A1

Att.A2

…Att.AN

1 A11 A12 … A1N

2 A21 A22 … A2N

3 A31 A32 … A3N

… … … … …

10000 tuples

N = 10000 = 100N

Page 13: Teacher : Hsien-Chu Wu Student : Hsiao-yun Tseng, Chen-ying Lai Speaker : Hsiao-yun Tseng Date : May 10, 2006 Database Temper Detection Techniques Based.

13

Method II – for grey watermarks Produce the certification code(2/3)

PAtt.A1

Att.A2

…Att.AN

1 A11 A12 … A1N

2 A21 A22 … A2N

3 A31 A32 … A3N

… … … … …

T Mi = MD5 (t

i)

1 0 0 1 0 1 1 1 0 0 1 … 0 1 0 1 1 1 1 0 1 0 0 …Mi =

1 0 0 1 0 1 1 1 0 0 1 … 0 1 0 1 1 1 1 0 1 0 0 …⊕bi= {0 ~ 63 bits of Mi}

1 1 0 0 1 0 0 1 1 0 1 …Xi Xi mod 256Ci

fi= {64 ~127 bits of M

i}

Page 14: Teacher : Hsien-Chu Wu Student : Hsiao-yun Tseng, Chen-ying Lai Speaker : Hsiao-yun Tseng Date : May 10, 2006 Database Temper Detection Techniques Based.

14

Method II – for grey watermarks Produce the certification code(3/3)

feature C

⊕ certification image R

SD Database

SD+

Internet

SKey

Page 15: Teacher : Hsien-Chu Wu Student : Hsiao-yun Tseng, Chen-ying Lai Speaker : Hsiao-yun Tseng Date : May 10, 2006 Database Temper Detection Techniques Based.

15

Method II – for grey watermarks The verification process of database integrity

Database SD+

Internet

SD

PKey

certification image R

T fetch the feature C

feature C

Page 16: Teacher : Hsien-Chu Wu Student : Hsiao-yun Tseng, Chen-ying Lai Speaker : Hsiao-yun Tseng Date : May 10, 2006 Database Temper Detection Techniques Based.

16

Method II – for grey watermarks Experimental results(1/4)

Emulation experiment Database Source : ProQuest Digital Dissertations (PQDD) Build a table in Microsoft SQL 2000 Sever Attributes : (Index, Publication number, Title, Author, Degree, School, Pages, Date, Digital formats)

original watermark(30 × 30)

designed watermark(100 × 100)

Page 17: Teacher : Hsien-Chu Wu Student : Hsiao-yun Tseng, Chen-ying Lai Speaker : Hsiao-yun Tseng Date : May 10, 2006 Database Temper Detection Techniques Based.

17

Method II – for grey watermarks Experimental results(2/4)

Exp. 1 : revise 30 Publication numbers

Result : find out 3 clear hashes and 30 tuples to be destroyed

The accuracy rate is 100%.

Exp. 2 : revise the front 2000 Author

Result : find out the clear hashes on the top half fetched watermark and 1,991 tuples to be altered

The accuracy rate is 99.55%.

Page 18: Teacher : Hsien-Chu Wu Student : Hsiao-yun Tseng, Chen-ying Lai Speaker : Hsiao-yun Tseng Date : May 10, 2006 Database Temper Detection Techniques Based.

18

Method II – for grey watermarks Experimental results(3/4)

Exp. 3 : revise the last 3000 Pages Result : find out the clear hashes on the foot half fetched watermark and 2,991 tuples to be alteredThe accuracy rate is 99.7 %.

Exp. 4 : delete Digital Formats attribute and replace with Degree attribute Result : find out the clear hashes on the whole fetched watermark and 9,965 tuples to be altered

The accuracy rate is 99.65 %.

Page 19: Teacher : Hsien-Chu Wu Student : Hsiao-yun Tseng, Chen-ying Lai Speaker : Hsiao-yun Tseng Date : May 10, 2006 Database Temper Detection Techniques Based.

19

Method II – for grey watermarks Experimental results(4/4)

Attack Accuracy rate

small alteration 100%

character alteration 99.55%

numerical alteration 99.7%

huge alteration 99.65%.

Average 99.725%

Page 20: Teacher : Hsien-Chu Wu Student : Hsiao-yun Tseng, Chen-ying Lai Speaker : Hsiao-yun Tseng Date : May 10, 2006 Database Temper Detection Techniques Based.

20

The proposed methods prove the integrity of the database and preserve a lossless database

In the future, we can hide SD in the database and restore an lossless database. It means to design a reversible database watermarking technique.

Conclusions

Page 21: Teacher : Hsien-Chu Wu Student : Hsiao-yun Tseng, Chen-ying Lai Speaker : Hsiao-yun Tseng Date : May 10, 2006 Database Temper Detection Techniques Based.

21

~ Thank You ~