1. Advisor： Chin-Chen Chang Student： Thai-Son Nguyen
Reversible Information Hiding Techniques and Their Applications in Image Protection
Advisor： Chin-Chen Chang
Student： Thai-Son Nguyen
Department of Computer Science and Information Engineering,
FengChia University
June 29, 2015

2. Outline Introduction Reversible Data Hiding Schemes in spatial domain
Scheme 1: A Novel Reversible Data Hiding Scheme Based on Difference-Histogram Modification and Optimal EMD Algorithm
Scheme 2: An Efficient Reversible Data Hiding Scheme Based on Adaptive Rhombus Prediction and Pixel Selection
Reversible Data Hiding Schemes in compressed domain
Scheme 3: Reversible Data Hiding for Indices Based on Histogram Analysis
Scheme 4: Adaptive Lossless Data-Hiding and Compression Scheme for SMVQ Indices Using SOC
Applications of reversible data hiding
Scheme 5 Reversible Authentication Scheme for Digital Images with High-Quality Images
Scheme 6: A Blind Reversible Robust Watermarking Scheme for Relational Databases
Conclusions and future works
This is the outline of presentation.
First, I will introduce about the motivation and objective of this study.
Then, six RDH schemes have been presented.
Finally, Some conclusions are drawn.

3. Introduction-Motivation
Encryption: Meaningless
Information Hiding: Hide the secret data into a cover data (meaningful).
Irreversible data hiding
Reversible data hiding
Three domains: spatial domain, compression domain, frequency domain
Basic requirements: visual quality, hiding capacity, reversibility, compression rate, robustness
Cover data: images, videos, audios, written texts, database
Let start from the motivation:
Nowadays, a huge amount of digital data is transmitted and store in the network, such as Internet each second.
Such data are easily modified and illegal copied by attacker or unauthorized users.
To solve this shortcoming, several solutions have been proposed, and they can be classified into two categories
Encryption and Information hiding
Encryption transforms the secret data to meaningless form to ensure its security.
However, using meaningless form, it may raise suspicion of malicious attackers
In contrast, Information hiding hides the secret data into a cover data, meaningful form.
Information hiding is also divided into two types, irreversible and reversible data hiding
Irreversible: can obtain high embedding capacity but it will distort the cover image permanently after the secret data is extracted.
Due to the importance of the restoration of the cover images in some special fields, i.e., medical images and military images.
Therefore, reversible data hiding is the promising solution in this case because it allows the image can recover to its original version.
So in this study, we try to propose some new RDH schemes
They are performed in the spatial domain and compressed domain
To meet the basic requirements as image quality, hiding capacity compression rate. Here, the robustness is evaluated
when we apply RDH scheme to protect the ownership of database
The cover data used in this study are images and database.

4. Introduction-Research Objectives
To propose two RDH schemes in spatial domain with the high embedding capacity while maintaining the good image quality.
To provide two RDH schemes in compressed domain, to improve further embedding capacity, compression rate, as well as embedding efficiency.
To apply RDH for image authentication with high accuracy of tamper detection and high quality.
To develop a reversible watermarking for relational database to protect them from illegal copying and manipulation by malicious attackers.

5. Scheme 1: A Novel Reversible Data Hiding Scheme Based on Difference-Histogram Modification and Optimal EMD Algorithm
Here we begin with the scheme 1:
RDH based on histogram modification and optimal EMD

6. A Novel Reversible Data Hiding Scheme Based on Difference-Histogram Modification and Optimal EMD Algorithm
Embedding phase
First the original image is divided into two sub-images A and B,
Then the LSB of B and the secret message are used to construct the optimal EMD table
And that table is used to embed the secret message into the sub-image A to get the stego-image

7. A Novel Reversible Data Hiding Scheme Based on Difference-Histogram Modification and Optimal EMD Algorithm
Embedding phase - Generation of Optimal EMD Table
Divide the secret data into into sequence S = {s1, s2,… , s|R|/3}, si is 3 bits
Histogram of sequences S
Sorted histogram of sequences S
The question is how we construct the optimal EMD table?
Here, the secret data is divided into the sequence S of three bits.
Then, the histogram of S is generated and it is sorted to obtain the sorted histogram like this one
The optimal EMD table is constructed according to this guidance
EX: based on this histogram the optimal EMD is obtained as this one
According to this table, we can embed 3 bits each time
With minimum embedding distortion
Embed three bits each time
Minimum embedding distortion
Optimal EMD table

8. A Novel Reversible Data Hiding Scheme Based on Difference-Histogram Modification and Optimal EMD Algorithm
Embedding phase –Block classification
Compute complexity
The sub-image A is partitioned into block sized of 3x3
Here, the center pixel is defined as pivot pixel.
Then, the complexity of the current block is calculated based on its pivot pixel and the eight reference pivot pixels
If NV < T, smooth block. Otherwise, complex block

9. A Novel Reversible Data Hiding Scheme Based on Difference-Histogram Modification and Optimal EMD Algorithm
Embedding phase – Embedding procedure
150
151
149
152 -2 -3
152 1 2
Peak = 0
T= 15
NV = 10
Secret data =
Original block -3 -4
152 2 3
149
150
148
152 1
Stego block
Optimal EMD table

10. A Novel Reversible Data Hiding Scheme Based on Difference-Histogram Modification and Optimal EMD Algorithm
Extracting phase
149
150
148
152
T = 15 -2 -3
152 1 2
Peak = 0
Stego block
Optimal EMD -3 -4
152 2 3
150
151
149
152
Secret data = 001
011 1
[Peak -1, Peak +1]
Original block

11. A Novel Reversible Data Hiding Scheme Based on Difference-Histogram Modification and Optimal EMD Algorithm
Experimental results
Here comes the experimental results of the first scheme.
We compare this scheme to four previous schemes.
And as seen that the proposed scheme achieves the higher quality of stego images
and embedding capacity for different images
Lena
Boat
[8] X. Li, B. Yang, and T. Zeng, “Efficient reversible watermarking based on adaptive prediction-error expansion and pixel selection,” IEEE Trans. Image Process., vol. 20, no. 12, pp , Dec
[12] X. Li, B. Li, B. Yang, and T. Zeng, “General framework to histogram-shifting-based reversible data hiding,” IEEE Trans. on Image Process., vol. 22, no. 6, pp , Jun
[19] W. Hong, “Adaptive reversible data hiding method based on error energy control and histogram shifting,” Opt. Commun., vol. 285, no. 2, pp , 2012.
[20] V. Sachnev, H. J. Kim, J. Nam, S. Suresh, and Y.Q. Shi, “Reversible watermarking algorithm using sorting and prediction,” IEEE Trans. Circuits Syst. Video Technol., vol. 19, no. 7, pp , Jul

12. A Novel Reversible Data Hiding Scheme Based on Difference-Histogram Modification and Optimal EMD Algorithm
Experimental results
Baboon
Peppers

13. A Novel Reversible Data Hiding Scheme Based on Difference-Histogram Modification and Optimal EMD Algorithm
Experimental results
Table 2.1 Comparison of PSNR (dB) between the proposed scheme
four previous schemes [8, 12, 19, 20] for EC of 10,000 bits
Images
Sachnev et al. [20]
Li et al. [8]
Hong et al. [19]
Li et al. [12]
Proposed
Lena
58.18
58.12
58.64
59.37
F16
60.38
60.74
61.72
62.65
61.53
Baboon
54.15
54.21
53.29
54.41
55.35
Peppers
55.55
56.06
56.02
56.89
58.71
Sailboat
58.15
57.29
58.27
58.26
Boat
56.15
55.57
56.55
57.16
58.61
Average
57.09
57.14
57.25
58.13
On this table, we compare the image quality of the proposed scheme
and four other schemes for EC 10,000 bits

14. An optimal EMD table is generated for RDH Reversibility
A Novel Reversible Data Hiding Scheme Based on Difference-Histogram Modification and Optimal EMD Algorithm
Summary
An optimal EMD table is generated for RDH
Reversibility
High image quality and high embedding capacity

15. Scheme 2: An Efficient RDH Scheme Based on Adaptive Rhombus Prediction and Pixel Selection
Here is the scheme 2, we use adaptive rhombus prediction and pixel selection for RDH

16. An Efficient RDH Scheme Based on Adaptive Rhombus Prediction and Pixel Selection
Embedding phase
In this scheme, the image is also divided into sub-images A and B
The sub-image B is used to solve the overflow/underflow problem
While the smooth pixels in the image A is used for embedding data
Prediction errors are calculated by adaptive rhombus prediction.
Here, the pixels in cross set is predicted by the pixels in the dot set and vice versa.
Then, optimal pair of peak and zero points are determined to embed data
Rhombus prediction

17. An Efficient RDH Scheme Based on Adaptive Rhombus Prediction and Pixel Selection
Embedding phase- Pixel selection
Compute local complexity
If LVx < TH, select it for embedding data.
In this scheme, the pixel is selected if its complexity is smaller than threshold.
And its predicted value is calculated as this formula
Then, calculate predicted value

18. An Efficient RDH Scheme Based on Adaptive Rhombus Prediction and Pixel Selection
Embedding phase
250
252
197
190
253
223
185
254
180
255
170
Skip unchanged
Prediction errors e =
TH = 5

19. An Efficient RDH Scheme Based on Adaptive Rhombus Prediction and Pixel Selection
Embedding phase
Peak P
Zero Z e=
P = 0, Z = 2
W = 0 1
e= 2 1 1

20. An Efficient RDH Scheme Based on Adaptive Rhombus Prediction and Pixel Selection
Embedding phase
e=
250
252
197
190
253
223
185
254
180
255
170
254
253
Stego image

21. An Efficient RDH Scheme Based on Adaptive Rhombus Prediction and Pixel Selection
Extracting phase
250
252
197
190
253
223
185
254
180
255
170
250
252
197
190
253
223
185
254
180
255
170
P = 0, Z = 2
Original image
W = 0 1

22. An Efficient RDH Scheme Based on Adaptive Rhombus Prediction and Pixel Selection
Optimal pair of peak and zero points
Embedding capacity L
Two possible cases:
1. F(Pl)  L, (Pl, Zl) is the first candidate
2. F(Pr)  L, (Pr, Zr) is the second candidate
Here, we explain how to select the optimal pair of peak and zero point
First, from the value 0 search to left and right hand sides to find two zero points, Zl and Zr
There are two possible cases:
From the left zero point Zl, search toward to the center to determine the suitable peak point PL
such that frequency of Pl is larger than L. and the PL and Zl is the first candidate pair
Similarly, from the right zero point, search the suitable peak point Pr, such that frequency of Pr is larger than L.
And Pr and Zr is the second candidate.
Finally, the optimal one is selected with minimum embedding distortion.
3. Optimal one is selected with
the smaller embedding distortion

23. An Efficient RDH Scheme Based on Adaptive Rhombus Prediction and Pixel Selection
Experimental results
Here is the result of the scheme 2. On this table the proposed scheme can embed the higher EC with smaller embedding distortion.

24. An Efficient RDH Scheme Based on Adaptive Rhombus Prediction and Pixel Selection
Lena
Airplane
With the same EC, the image quality of stego images outperform than other four schemes.
Baboon
Peppers

25. An Efficient RDH Scheme Based on Adaptive Rhombus Prediction and Pixel Selection
Sailboat
Boat
Here is four schemes are used to compare to our scheme because of their high performance.
[12] X. Li, B. Li, B. Yang, and T. Zeng, “General framework to histogram-shifting-based reversible data hiding,” IEEE Trans. on Image Process., vol. 22, no. 6, pp , Jun
[13] J. Wang, J. Ni, and Y. Hu, “An efficient reversible data hiding scheme using prediction and optimal side information selection,” J. Vis. Commun. Image Represent., vol. 25, pp , 2014.
[20] V. Sachnev, H. J. Kim, J. Nam, S. Suresh, and Y.Q. Shi, “Reversible watermarking algorithm using sorting and prediction,” IEEE Trans. Circuits Syst. Video Technol., vol. 19, no. 7, pp , Jul
[21] L. Luo, Z. Chen, M. Chen, X. Zeng, and Z. Xiong “Reversible image watermarking using interpolation technique,” IEEE Trans. Inf. Forens. Secur., vol. 5, no. 1, pp , 2010.

26. An Efficient RDH Scheme Based on Adaptive Rhombus Prediction and Pixel Selection
Summary
Adaptive rhombus prediction and pixel selection techniques are used for RDH
Higher performance in terms of image quality and embedding capacity

27. Scheme 3: Reversible Data Hiding for Indices Based on Histogram Analysis
Here we describe the scheme 3

28. Reversible Data Hiding for Indices Based on Histogram Analysis
In this scheme, the image is encoded by VQ and then is transmitted by SMVQ
The question is Why we combine VQ and SMVQ?
Look at this table, if we used VQ coding only , the indices were distributed everywhere.
But if we used transform the VQ indices to SMVQ algorithm, most of indices are concentrated from 0 to 7.
We exploit this property to further compress the image and embed more secret data.
Encoding method
Frequency of indices
Tiffany
Goldhill
Peppers
VQ only
Indices [0,7]
0.00%
6.24%
16.84%
Indices [8,255]
100.00%
93.76%
83.16%
VQ and SMVQ
87.70%
59.72%
77.21%
12.30%
40.28%
22.79%

29. Reversible Data Hiding for Indices Based on Histogram Analysis
The number of zero frequency is U = 9
The largest mapping bits
Sorted histogram
Because most of indices are concentrated in the small value, so we analyze the histogram of indices, to embed more secret data.
For EX: the frequency of indices are obtained as this table.
Here the number of zero frequency is 9.
We used this formula to calculate the largest bits ca be embedded in our scheme.
The result is 3, meaning that, we can map at most three bits.
And we can determine mapping function by using this formula.
For this case, we can embed by mapping two bits once and mapping one bit 6 times.
And the mapping table for mapping two bits
Several mapping functions can be obtained. .…likes this.
And the optimal one with the highest embedding capacity is selected.
In this example, mapping function [3, 2, 0] is selected because it can achieve highest EC.
By doing so, we can embed higher EC while maintaining low CR as traditional VQ.
To make sure the reversibility, the sorted histogram is send to reciever
[2 0 1], [0 3 0], [3 2 0], and [6 1 0]
[3 2 0]
Achieve higher embedding capacity
Low compression rate as traditional VQ

30. Reversible Data Hiding for Indices Based on Histogram Analysis
Embedding phase
S =
Transformed index table IT 1 3 2 6 4 5 8
Here, we take an example for embedding process according to mapping table.
For this mapping table, we can embed 2 bits for each time the index 0 is encountered. 9 7 8
Stego index table

31. Reversible Data Hiding for Indices Based on Histogram Analysis
Experimental results
The CR of the proposed scheme is preserved the same as traditional VQ.
It is slightly higher than, Chang et al.’s scheme.
But
Figure 4.11 Compression rate results of our proposed scheme and some previous schemes

32. Reversible Data Hiding for Indices Based on Histogram Analysis
Experimental results
But when, in terms of EC, the proposed provides the superior performance.
Figure 4.12 Embedding rate (ER) results of our proposed scheme and some previous schemes

33. Reversible Data Hiding for Indices Based on Histogram Analysis
Summary
Preserve image quality and compression rate the same as those of traditional VQ.
Improve embedding rate of previous schemes further
[46] Z. M. Lu, J. X Wang, and B. B. Liu, “An improved lossless data hiding scheme based on image VQ-index residual value coding,” Journal of Systems and Software, vol. 82, pp , 2009.
[47] C. H. Yang and Y. C. Lin, “Reversible data hiding of a VQ index table based on referred counts,” J. Vis. Commun. Image Represent., vol. 20, no. 6, pp , Aug
[48] J. X. Wang and Z. M. Lu, “A path optional lossless data hiding scheme based on VQ joint neighboring coding”, Information Sciences, vol. 179, pp , 2009.
[49] C. F. Lee, H. L. Chen, and S. H. Lai, “An adaptive data hiding scheme with high embedding capacity and visual image quality based on SMVQ prediction through classification codebooks,” Image and Vision Computing, vol. 28, no. 8, pp , 2010.
[50]C. C. Chang, T. S. Nguyen, and C. C. Lin, “A novel VQ-based reversible data hiding scheme by using hybrid encoding strategies,” Journal of Systems and Software, vol. 86, pp , 2013.
These five schemes are used in comparisons.

34. Scheme 4: Adaptive Lossless Data-Hiding and Compression Scheme for SMVQ Indices Using SOC
Here come schemes 4.

35. Adaptive Lossless Data-Hiding and Compression Scheme for SMVQ Indices Using SOC
Figure 5.2 Distribution of indices by using VQ compression and SOC algorithm
SOC is designed for further compressing VQ indices.
However, the distribution of VQ indices are not very concentrated.
So, the property of SOC can not fully be exploited.
It is observed that, if applying SOC for SMVQ the better compression result will be obtained.
Because of the concentrated distribution of SMVQ indices.
Meaning that the more embedding space is generated for hiding secret data.
Figure 5.3 Distribution of indices by using SMVQ compression and SOC algorithm

36. Figure 5.5 The main processes of the proposed embedding algorithm
Adaptive Lossless Data-Hiding and Compression Scheme for SMVQ Indices Using SOC
Here the flowchart of the proposed schemes, five cases are determined for embedding data
Figure 5.5 The main processes of the proposed embedding algorithm

37. Table 5.3 Encoding rule of the proposed scheme
Adaptive Lossless Data-Hiding and Compression Scheme for SMVQ Indices Using SOC
Table 5.3 Encoding rule of the proposed scheme
Cases
Encoding rule
Compression code
Under-hiding
m1 = 5 bits
m2 = 3 bits
Normal-hiding
m1 = 6 bits
m2 = 4 bits
Over-hiding
m1 = 7 bits
m2 = 5 bits
Case 1 00
00|| m1 secret bits
Case 2
01|| SOC-2bit
01||SOC-2bit || m2 secret bits
00||SOC-2bit || m2 secret bits
Case 3
10||OIV of P -
Case 4
11||indicatori||0||log2(thr) bits of |d|
Case 5
11||indicatori||1||log2(thr) bits of |d|
They are described in this table
The first two cases are used for embedding data.
And the last three cased are used for compression.
Here, we divide the proposed schemes into four different schemes,
The first one is compression scheme.
The second is under-hiding, it controls the compression rate smaller than 0.5
The third one is equal to 0.5
The last one is larger than 0.5

38. Adaptive Lossless Data-Hiding and Compression Scheme for SMVQ Indices Using SOC4 11
Case 3
Case 4
Case 1
Case 2
SMVQ index table
Secret message
Threshold thr = 8, m1 = 6, m2 = 4
Here is an example of embedding data of the scheme 4
Code steam CS
10||
11||00||0||111
00||101110
01||01||1010

39. Adaptive Lossless Data-Hiding and Compression Scheme for SMVQ Indices Using SOC
Images
Parameters
Under-hiding
Qin et al. [64]
Pan et al.[65]
Lena EC
43,463
10,292
16,384 CR
0.46
0.40
0.52 EF
36.40%
9.72%
12.02%
Airplane
60,488
11,182
0.45
0.39
0.51
51.15%
10.91%
12.25%
Tiffany
71,707
11,566
0.44
0.49
62.19%
11.46%
12.76%
Peppers
43871
11,295
0.54
36.57%
11.08%
11.57%
Sailboat
42042
7,247
0.53
34.63%
6.09%
11.79%
Boat
44886
9,301
0.42
37.51%
8.41%
Average
51,076
10,147
0.41
43.08%
9.61%
12.10%
Here is the result of scheme 4

40. Adaptive Lossless Data-Hiding and Compression Scheme for SMVQ Indices Using SOC
Images
Parameters
Normal-hiding
Over-hiding
Lee et al.[63]
Wang et al.[66]
Lena EC
56,322
69,181
46,962
46,694 CR
0.50
0.55
0.52
0.49 EF
42.59%
47.67%
34.45%
36.35%
Airplane
74,064
87,640
57,933
59,473
0.53
56.18%
60.28%
42.50%
42.81%
Tiffany
87,522
103,337
72,177
74,875
0.56
0.51
66.75%
70.33%
53.67%
51.00%
Peppers
56470
69,069
45516
49,480
42.60%
47.58%
33.33%
37.75%
Sailboat
53,396
64,750
44046
49,152
40.22%
44.93%
32.07%
37.50%
Boat
57,578
70,270
39042
48,634
43.51%
48.45%
28.70%
36.38%
Average
64,225
77,375
50,946
54,718
48.64%
53.21%
37.45%
40.30%

41. High EC and high EF while guaranteeing a low CR.
Adaptive Lossless Data-Hiding and Compression Scheme for SMVQ Indices Using SOC
Summary
High EC and high EF while guaranteeing a low CR.
Further improve the performance of four previous schemes [63-66]
Here is four schemes are used for comparisons
[63] J. D. Lee, Y. H. Chiou, and J. M. Guo, “Lossless data hiding for VQ indices based on neighboring correlation,” Information Sciences, vol. 221, pp , Dec
[64] C. Qin, C. C. Chang, Y. P. Ping, “A novel joint data hiding and compression scheme based on SMVQ and image inpainting,” IEEE Trans. Image Process., vol. 23, no. 3, pp , Mar
[65]Z. B. Pan, X. X. Ma, X. M. Deng, S. Hu, “Low bit-rate information hiding method based on search-order-coding technique,” Journal of Systems and Software, vol. 86, pp , 2013.
[66] L. F. Wang, Z. B. Pan, X. X. Ma, S. Hu, “A novel high-performance reversible data hiding scheme using SMVQ and improved locally adaptive coding method,” J. Vis. Commun. Image Represent., vol. 25, pp , 2013.

42. Scheme 5: A Reversible Authentication Scheme for Digital Images with High-Quality Images
In scheme 5, we apply RDH for image authentication

43. Figure 6.1 Framework of the proposed authentication scheme
A Reversible Authentication Scheme for Digital Images with High-Quality Images
The image blocks are classified into the smooth block and the complex blocks
Then, the authentication code is generated based on the seed K and embedded into the image
with minimum modifications.
Figure 6.1 Framework of the proposed authentication scheme

44. A Reversible Authentication Scheme for Digital Images with High-Quality Images
For each image block with the size of 3x3 , its complexity is computed.
The authentication code is embedded into the left or the right pixels of the block for verification.
Figure 6.2 Example of an image block and its satellite reference pixels

45. A Reversible Authentication Scheme for Digital Images with High-Quality Images25 24 26 23 22
dL = L- C = 0
dL’ = 2 x = 1
dR’ = 2 x = 2
dR = R - C = 1
Original block 25 24 26 27 23 22
T* = 2
W = 1 0
Here is an example to embed the authentication code into the block.
We have the original image block, and the prediction errors of left and right pixels are calculated as 0 and 1
Then according to this formula the authentication code W is embedded.
Here T* is equal to 2. then, the new prediction errors are obtained as this ones.
And the stego block is obtained.
Stego block

46. Figure 6.6 Image quality of the embedded images of the proposed scheme
A Reversible Authentication Scheme for Digital Images with High-Quality Images
Image quality of various image under different values of T*
Figure 6.6 Image quality of the embedded images of the proposed scheme

47. A Reversible Authentication Scheme for Digital Images with High-Quality Images
(b) T* = 1
(d) T* = 2
(e) T* = 3
For tamper detection test, the image butterfly is added on the wall of the image Lena
Figure 6.9 Tampered images “Lena” of the proposed scheme with various values of T*

48. A Reversible Authentication Scheme for Digital Images with High-Quality Images
(a) Refined detected image with T* = 0, NC =
(b) Refined detected image with T* = 1,
NC =
(c) Refined detected image with T* = 2, NC =
(d) Refined detected image with T* = 3,
NC =
Here is the tampered detection results

49. Authentication code bits per block
A Reversible Authentication Scheme for Digital Images with High-Quality Images
Table 6.5 Comparison of the proposed scheme and Lo and Hu’s scheme
Schemes
Block size
Authentication code bits per block
Average PSNRs
Average NC
Clear tamper area
Lo and Hu
4 × 4
2.76
51.48 dB
0.9106
Yes
Proposed
3 × 3
1.35
51.72 dB
0.9429
We compared the proposed scheme to Lo and Hu’s scheme because this scheme obtained the reversibility as the proposed scheme.
Here, the proposed scheme can achieved more accuracy of tamper detection although embedding less authentication code.
[86] C. C. Lo, Y. C. Hu, “A novel reversible image authentication scheme for digital images,” Signal Process., vol. 98, pp , 2014.

50. Achieve reversibility.
A Reversible Authentication Scheme for Digital Images with High-Quality Images
Summary
Obtain high accuracy of tamper detection and preserve high quality of the stego images.
Achieve reversibility.

51. Scheme 6: A Blind Reversible Robust Watermarking Scheme for Relational Databases
In the scheme 6, we apply RDH to protect the ownership of relational database

52. A Blind Reversible Robust Watermarking Scheme for Relational Databases
Embed
Watermark
Relational database
Several techniques have been proposed to protect the ownership of database.
However, most of them are irreversible.
We try to design new RDH scheme for relational database with high robustness
Irreversible
Reversible and more robustness

53. A Blind Reversible Robust Watermarking Scheme for Relational Databases

54. A Blind Reversible Robust Watermarking Scheme for Relational DatabasesID A1 A2 A3 A4
PK1
295 52 75 48
PK8
256 94 25
PK13
126
451
455
PK17 55 15 11
512
PK23
452
964
PK30 58
254 54
Database
Assume that
selected attribute:
ATT(PK1) = 2
ATT(PK8) = 1
ATT(PK13) = 4
ATT(PK17) = 1
ATT(PK23) = 3
ATT(PK30) = 3
Get two last digits of each selected attributes
Seq = 52, 52, 54, 55, 55, 56
For each tuple, the attribute is selected.
For example, the attributes are selected in this database table as
Then, two last digits are extracted and sorted.
Later on, the mid value is determined. And the different values are computed by using this formula.
The different results are obtained.
Mid = 54
Diff_Seq [i] =Seq [i] - mid
Diff_Seq = -2, -2, 0, 1, 1, 2

55. A Blind Reversible Robust Watermarking Scheme for Relational Databases
PP1
PP2
Diff _Seq= -2, -2, 0, 1, 1, 2
ZP1
ZP2
We construct the histogram of different values
Two pairs of peak and zero points are determined in the histogram
And the watermark is embedded by histogram shifting
The values between the peak and zero points, PP1 and ZP1, are shifted toward to left
And the values between the peak and zero points, PP2 and ZP2, are shifted toward to right.
Different values are obtained as this one
PP1
PP2
Diff _Seq= -2, -2, 0, 1, 1, 3
Histogram after shifting

56. A Blind Reversible Robust Watermarking Scheme for Relational Databases
Diff _Seq= -2, -2, 0, 1, 1, 3
PP1
PP2
Watermark b =
Hide ‘1’ in – 2  – 2 – b = – 2 – 1 = – 3
Hide ‘0’ in – 2  – 2 – b = – 2 – 0 = – 2
Hide ‘1’ in 1  1 + b = = 2
Histogram after shifting
Hide ‘0’ in 1  1 + b = = 1 ID A1 A2 A3 A4
PK1
295 51 75 48
PK8
257 94 25
PK13
126
451
456
PK17 55 15 11
512
PK23
452
964
PK30 58
254 54
We can embed the watermark into the different values and the watermarked database is generated.
Diff _Seq becomes -3, -2, 0, 2, 1, 3

57. A Blind Reversible Robust Watermarking Scheme for Relational Databases
Experimental results
Here, the result of the proposed scheme in comparison to two previous schemes.
Figure 7.7 Comparison of the results for resilience to an alteration attack by the two proposed schemes and two other schemes

58. A Blind Reversible Robust Watermarking Scheme for Relational Databases
Experimental results
The proposed scheme provided stronger robustness than other two schemes under different data attacks, such as alternation and deletion
Figure 7.8 Comparison of the results for the resilience to the detection attack of the two proposed schemes and two other schemes

59. A Blind Reversible Robust Watermarking Scheme for Relational Databases
Summary
Achieves reversibility.
Resilient to various attacks.
[89] M. Shehab, E. Bertino, and A. Ghafoor, “Watermarking relational databases using optimization-based techniques,” IEEE Trans. Knowledge Data Engineer., vol. 20, no. 1, pp. 116, 129, Jan
[92] M. E. Farfoura, S. J. Horng, J. L. Lai, R. S. Run, R. J. Chen, and M. K. Khan, “A blind reversible method for watermarking relational databases based on a time-stamping,” Expert Systems with Applications, vol. 39, pp , 2012.

60. Conclusions
Six RDH techniques have been proposed in this study with high performances:
Reversibility.
Good image quality.
Higher embedding capacity.
Lower compression rate.
High accuracy of tampered detection.
More robustness.

61. Future works
For the RDH schemes of Chapter 2 and Chapter 3, only one bit is embedded into each pixel. Therefore, we try to improve embedding capacity as well as image quality (i.e., larger than 1 bit per pixel and greater than 60 dB).
Enhance the performance of my proposed schemes in Chapters 4 and 5, i.e., higher embedding rate (up to 6 bits per index) and lower compression rate (less than 0.4 bit per pixel) while guaranteeing good image quality of reconstructed image (PSNRs > 30dB) .
Improve the Scheme 6 in Chapter 7 to achieve more security against some malicious attacks in relational database (e.g. deletion, sorting, modification, addition attacks)
Develop RDH schemes in the frequency domain by using Discrete Wavelet Transform (DWT) and Discrete Cosine Transform (DCT) to achieve high robustness.
Develop reversible data hiding techniques for encrypted images and high dynamic range (HDR) images.
Address data hiding algorithms for 3D images, stereo images, and videos.
In the future work, we intend to improve the proposed schemes further in terms of image quality and embedding capacity.
We also want to develop RDH scheme in the frequency domains and apply RDH for different cover data, such as encrypted image, 3D image, stereo image and videos.

62. Publication list International Journal Papers:
1. C. C., Chang, T. S. Nguyen, and C. C. Lin, “A reversible data hiding scheme for VQ indices using locally adaptive coding,” Journal of Visual Communication and Image Representation (JVCI), vol. 22, no. 7, pp , Oct  (SCI) 
2.  W. X. Tian, C. C. Chang, T. S. Nguyen, and M. C.  Li, “Reversible data hiding for high quality image exploiting interpolation and direction order mechanism,” Digital Signal Processing, vol. 23, no. 2, pp , Mar  (SCI) 
3.  C. C., Chang, T. S. Nguyen, and C. C. Lin, “A novel VQ-based reversible data hiding scheme by using hybrid encoding strategies,” Journal of Systems and Software, (JSS) vol. 86, no. 2, pp , Feb   (SCI) 
4.   C. C., Chang, T. S. Nguyen, and C. C. Lin, “Distortion-free data hiding for high dynamic range images,” Journal of Electronic Science and Technology, (JEST) vol. 11, no. 1, pp , Mar   (EI)
5. C. C., Chang, T. S. Nguyen, and C. C. Lin, “Reversible image hiding for high image quality based on histogram shifting and local complexity,”  International Journal of Network and Security (IJNS), vol. 16, no. 3, pp , May 2014 . (EI, Scopus)
6. C. C., Chang, T. S. Nguyen, and C. C. Lin, “A blind reversible robust watermarking scheme for relational databases,” Scientific World Journal (SWJ), volume 2013. (SCI)
7. C. C., Chang, T. S. Nguyen, and C. C. Lin, “A novel compression scheme based on SMVQ and Huffman coding,” International Journal of Innovative Computing, Information and Control, vol. 10, no. 3, June 2014.  (EI, Scopus)
8.   C. C., Chang, T. S. Nguyen, and C. C. Lin, “A reversible data hiding scheme for VQ indices based on absolute difference trees,” KSII Transactions on Internet and Information Systems, vol. 8, no. 7, Jul   (SCI) 
9.   C. C., Chang, T. S. Nguyen, and C. C. Lin, “Reversible data embedding for indices based on histogram analysis,” Journal of Visual Communication and Image Representation (JVCI), vol. 25, no. 7, pp , Oct   (SCI, EI) 
This is the results that I worked in three years

63. Publication list
10. T. S. Nguyen, C. C. Chang, and T. F. Chung, “A tamper-detection scheme for BTC-compressed images with high-quality images,” KSII Transactions on Internet and Information Systems, vol. 8, no. 6, Jun   (SCI)
11. T. S. Nguyen, C. C. Chang, and M. C.  Lin, “Adaptive lossless data-hiding and compression scheme for SMVQ indices using SOC,” Smart Computing Review, vol. 4, no. 3, Jun
12. W. L. Lyu, C. C. Chang, T. S. Nguyen, and C. C. Lin, “Image watermarking scheme in areas of interest using scale-invariant feature transform,” KSII Transactions on Internet and Information Systems, vol. 8, no. 10, 2014. (SCI) 
13. C. C., Chang, T. S. Nguyen, and C. C. Lin, “A new distortion-free data embedding scheme for high dynamic images,” Multimedia Tools and Applications, Available on 28/9/2014 (SCI)  
14. C. C. Chang and T. S. Nguyen, “A reversible data hiding scheme for SMVQ indices,” Informatica, vol. 25, no. 4, pp. 523–540, 2014. (SCI) 
15. C. C., Chang, T. S. Nguyen, and C. C. Lin, “A reversible compression code hiding using SOC and SMVQ indices,” Information Sciences, vol. 300, pp , (SCI)
International Conference Papers
16. C. C. Chang, Y. J. Liu, and T. S. Nguyen, “A novel turtle shell based scheme for data hiding,” Proceedings of the 10th International Conference on Intelligent Information Hiding and Multimedia Signal Processing (IIHMSP14),Kitakyushu, Japan (EI) (Best paper award).
17. C. C., Chang, T. S. Nguyen, and C. C. Lin, “A virtual primary key for reversible watermarking textual relational databases,” Proceedings of The International Computer Symposium 2014, Dec., 2014, Taichung, Taiwan, pp
18. C. C. Chang, Y. J. Liu, and T. S. Nguyen, “Hiding secret information in block truncation code using dynamic programming strategy,” 6th International Conference on Graphic and Image Processing (ICGIP 2014), Beijing, China, 2014/10/ (Best paper award).
19. C. C., Chang, T. S. Nguyen, and C. C. Lin, “A blind robust reversible watermark scheme for textual relational databases with virtual primary key,” 13th International Workshop on Digital-Forensics and Watermarking (IWDW 2014), October 1-4, 2014, Taipei, Taiwan.

64. Publication list Submitted Papers
20. C. C., Chang, T. S. Nguyen, and C. C. Lin, “New compression algorithms based on SOC and SMVQ,” submitted to Informatica (Submitted 2014/12/25) (SCI / EI, Impact Factor: 0.901) (Reviewing).
21. C. C. Chang and T. S. Nguyen, “A reversible data hiding scheme based on the Sudoku technique,” submitted to Displays (Submitted: 2013/01/27) (SCI / EI, Impact Factor: 1.390) (Reviewing).
22. C. C. Chang and T. S. Nguyen, “A reversible data hiding scheme for image interpolation based on reference matrix,” submitted to Journal of Imaging Science and Technology (Submitted: 2013/10/17). (SCI / EI, Impact Factor: 1.390) (Reviewing).
23. C. C., Chang, T. S. Nguyen, and C. C. Lin, “A blind reversible robust watermarking scheme for categorical relational databases,” submitted to Computers & Security (Submitted: 2015/04/15) (SCI , Impact Factor: 1.488)
24. T. S. Nguyen, C. C. Chang, W. C. Chang, C. C. and Lin, “High capacity reversible data hiding for BTC images,” submitted to ACM Transactions on Multimedia Computing, Communication and Applications (Submitted: 2015/3/24). (SCI, Impact Factor: 0.48)
25. C. C. Chang, T. S. Nguyen, M. C. Lin, and C. C. Lin, “A novel data-hiding and compression scheme based on block classification of SMVQ indices,” submitted to Digital Signal Processing (Submitted: 2014/07). (SCI, Impact Factor: 2.018) (Revision)
26. T. S. Nguyen, C. C. Chang, and T. H. Shih, “A high-quality reversible image authentication based on adaptive PEE for digital images,” submitted to KSII Transactions on Internet and Information Systems (Submitted: 2015/04/09). (SCI, Impact Factor: 0.56)
27. T. S. Nguyen, C. C. Chang, and T. H. Shih, “A reversible authentication scheme for digital images with high-quality images,” submitted to Multimedia Tools and Applications (Submitted: 2015/01/20). (SCI, Impact Factor: 1.058)
28. Y. J. Liu, C. C. Chang, and T. S. Nguyen, “A high capacity data hiding scheme based on turtle shell,” submitted to IET Image Processing (Submitted: 2014/12/21). (SCI, Impact Factor: 0.676)

65. Publication list
29. T. S. Nguyen, C. C. Chang, and H. S. Hsueh, “High capacity data hiding for binary image based on block classification,” submitted to Multimedia Tools and Applications (Submitted: 2014/12/28). (SCI, Impact Factor: 1.058)
30. H. L. Wu, C. C. Chang, and T. S. Nguyen, “Reversible data hiding using pixel order exchange,” submitted to Journal of Visual Languages & Computing (Submitted: 2015/05/20). (SCI, Impact Factor: 0.8)
31. T. S. Nguyen, C. C. Chang, and N. T. Huynh, “A novel reversible data hiding scheme based on difference-histogram modification and optimal EMD algorithm,” submitted to Journal of Visual Communication and Image Representation (JVCI) (Submitted: 2015/1/30). (SCI, Impact Factor: 1.430)
32. T. S. Nguyen, C. C. Chang, and X. Q. Yang, “High-quality semi-fragile watermarking for image authentication and tamper detection,” submitted to Computers & Security (Submitted: 2015/2). (SCI, Impact Factor: 1.488)
33. T. S. Nguyen, C. C. Chang, and H. L. Wu, “Adaptive image sharing scheme based on quadri-directional-search-strategy with meaningful shadows,” submitted to Multimedia Tools and Applications (SCI, Impact Factor: 1.058)
34. C. C. Chang, Nguyen, T. S., and C. C. Lin, “A blind, reversible, robust watermarking scheme for textual and numerical relational databases,” submitted to Systems, Man, and Cybernetics: Systems, IEEE Transactions on (Submitted: 2015/4/15)
35. T. S. Nguyen, C. C. Chang, and W. C. Chang,  “High capacity reversible data hiding scheme for encrypted images,” submitted to Signal Processing: Image Communication (Submitted: 2015/04/13). (SCI, Impact Factor: 1.28)
36. T. S. Nguyen, C. C. Chang, and C. C. Lin, “High capacity reversible data hiding scheme based on AMBTC for encrypted images,” submitted to IEEE Trans. on Cybernetics (Submitted: 2015/4/20).
37. T. S. Nguyen, C. C. Chang, and W. C. Chang,  “An efficient reversible data hiding scheme based on rhombus prediction and pixel selection,” submitted to Information Sciences (Submitted: 2015/05/14). (SCI, Impact Factor: 3.969)

66. Publication list
38. T. S. Nguyen, C. C. Chang, and H. S. Hsueh, “High-quality Data Hiding Algorithm for H.246/AVC Video Stream without Intra-Frame Distortion Drift,” Neurocomputing (Submitted: 2015/05/19). (SCI)
39. T. S. Nguyen, C. C. Chang, and T. H. Shih, “Effective reversible image authentication based on rhombus prediction and local complexity,” submitted to Journal of Visual Languages & Computing (Submitted: 2015/05/26). (SCI)
40. T. S. Nguyen, C. C. Chang, and Y. Q. Yang, “Fragile watermarking for image authentication based on DWT-SVD-DCT features with High-quality,” submitted to Signal Processing (Submitted: 2015/06/09). (SCI)

67. Thanks for your listening !