1. New Framework for Reversible Data Hiding in Encrypted Domain
Source： IEEE Transactions on Information Forensics and Security ,
vol.11, no.12, pp:  , 2016.
Authors：Fangjun Huang, Jiwu Huang, Yun-Qing Shi
Speaker ：Xiaozhu Xie
Date ： /01/19

2. Outline Motivation Proposed scheme Experimental results Conclusions
I will present in four parts. So first let us turn to introduction.

3. Correlation between the neighboring pixels does not exist any more.
Motivation(1/2)
RDH methods in plain images(PI)
(Original) histogram shifting(HS)
Difference histogram shifting(DHS)
Prediction-error histogram shifting(PEHS)
RDH in encrypted images.
Correlation between the neighboring pixels does not exist any more. …
Simple computation.
High embedding capacity.
Good visual quality.

4. Motivation(2/2) Specific encryption algorithm is designed.
Preserve the correlation between the neighboring pixels.
本Paper的motivation
Therefore, this paper proposes a novel framework for reversible data hiding in encrypted image (RDH-EI) based on reversible image transformation (RIT).

5. Proposed scheme(1/6)- Specific encryption algorithm
Original Image I
154
155
158 ⨁
Original
sub-block
key1 …
𝑅 1
𝑅 2
𝑅 𝑁 … 13 12 9
N non-overlapping blocks
Encrypted
sub-block
Step 1: Specific stream encryption.
Step 2: Permutation.
key2
Permute all the N encrypted sub-blocks.

6. Proposed scheme(2/6)-Preservation of Statistical Characteristics
Two neighboring pixels: (𝑃 𝑖,𝑥 , 𝑃 𝑖,𝑦 )
Case 1: 𝑃 𝑖,𝑦 − 𝑃 𝑖,𝑥 =0 , difference remains 0 after encryption.
Case 2: 𝑃 𝑖,𝑦 − 𝑃 𝑖,𝑥 =−1.
Case 3: 𝑃 𝑖,𝑦 − 𝑃 𝑖,𝑥 =1
{(𝛼,𝛼+1)|0≤𝛼≤254} different pixel pairs.
K=0
𝑃 𝑖,𝑥 =
𝑃 𝑖,𝑦 =
K=1
𝑃 𝑖,𝑥 =
𝑃 𝑖,𝑦 =
𝑃 𝑖,𝑥 =
𝑃 𝑖,𝑦 =
K=2
𝑃 𝑖,𝑥 =
𝑃 𝑖,𝑦 =
𝑃 𝑖,𝑥 =
𝑃 𝑖,𝑦 =
𝑃 𝑖,𝑥 =
𝑃 𝑖,𝑦 =
𝑃 𝑖,𝑥 =
𝑃 𝑖,𝑦 =
K=3,…,7 …

7. Proposed scheme(3/6)- Preservation of Statistical Characteristics
𝑁 #1 : Number of original pixel pairs with difference value 1.
𝑁 𝐸 #1 : Number of encrypted pixel pairs with difference value 1.
K=7
𝑃 𝑖,𝑥 = 𝑏 1 𝑏 2 𝑏 3 𝑏 4 𝑏 5 𝑏 6 𝑏 7 0
𝑃 𝑖,𝑦 = 𝑏 1 𝑏 2 𝑏 3 𝑏 4 𝑏 5 𝑏 6 𝑏 7 1
𝑅 𝑖 ⨁
Difference
1,-1
K=6
𝑃 𝑖,𝑥 = 𝑏 1 𝑏 2 𝑏 3 𝑏 4 𝑏 5 𝑏 6 01
𝑃 𝑖,𝑦 = 𝑏 1 𝑏 2 𝑏 3 𝑏 4 𝑏 5 𝑏 6 10
3,-3,1,-1 …
Assumption:
Key stream bits are uniformly distributed.
The difference may remain 1 or be changed to -1,
with 50 percentage, respectively .

8. Proposed scheme(4/6)-Preservation of Statistical Characteristics
Deduce :
At least 𝑁 #1 pixel pairs’ difference remains 1, and 𝑁 #1 of them will change to -1.
After encryption, the neighboring pixel pairs with the difference value 0 are preserved, and more than 2/3 pixel pairs with the difference values 1 and -1 are still preserved.

9. Proposed scheme(5/6)- DHS embedding
𝑪 𝒊,𝟏
𝑪 𝒊,𝟐
𝑪 𝒊,𝟑
𝑪 𝒊,𝟒
Overflow/ underflow
𝑪 𝒊,𝟏 ′
𝑪 𝒊,𝟐 ′
𝑪 𝒊,𝟑 ′
𝑪 𝒊,𝟒 ′
𝑪 𝒊,𝟏 ′
𝑫 𝒊,𝟐
𝑫 𝒊,𝟑
𝑫 𝒊,𝟒
𝑪 𝒊,𝟏 ′
𝑪 𝒊,𝟐 ′′
𝑪 𝒊,𝟑 ′′
𝑪 𝒊,𝟒 ′′
b: bit to be embedded.
Encrypted
sub-block
preprocessed
sub-block
Difference
sub-block
Marked
sub-block
Encrypted
image
(Location map L will
also be embedded.)
Procedure of DHS embedding
253
255
252
253
254
252
253 1 -1
253
255
254
252
Secret message: 10
Encrypted
sub-block
preprocessed
sub-block
Difference
sub-block
Marked
sub-block
Encrypted
image
Example of DHS embedding

10. Proposed scheme(6/6)- DHS extraction & recovery
253
255
254
252
Extraction
Secret message: 10
Marked
image
Marked
sub-block
Location map L
Recovery
253
254
252
253
255
252
Key 1
154
155
158
Key 2
Original
image
permute
Encrypted
sub-block
Original
sub-block
Example of DHS extraction & recovery

11. Experimental Results(1/5)
Different sub-blocks of the encrypted image.
(a) DHS_1. (b) DHS_2. (c) DHS_3.
Three prediction algorithms. (a) MED. (b) GAP. (c) rhombus.
(a) PEHS_1. (b) PEHS_2. (c) PEHS_3.
PEHS_1: M. J. Weinberger, G. Seroussi, and G. Sapiro, “The LOCO-I lossless image compression algorithm: Principles and standardization into JPEG-LS,” IEEE Trans. Image Process., vol. 9, no. 8, pp. 1309–1324, Aug
PEHS_2: X.Wu and N. Memon, “Context-based, adaptive, lossless image coding,” IEEE Trans. Commun., vol. 45, no. 4, pp. 437–444, Apr. 1997
PEHS_3: B. Ou, X. Li, Y. Zhao, R. Ni, and Y.-Q. Shi, “Pairwise predictionerror expansion for efficient reversible data hiding,” IEEE Trans. Image Process., vol. 22, no. 12, pp. 5010–5021, Dec

12. Experimental Results(2/5)
Look like Mosaic [moʊˈzeɪɪk] images.

13. Experimental Results(3/5)

14. Experimental Results(4/5)

15. Experimental Results(5/5)
Zhang [12]: X. Zhang, “Reversible data hiding in encrypted image,” IEEE Signal Process. Lett., vol. 18, no. 4, pp. 255–258, Apr
Zhang [14]: X. Zhang, “Separable reversible data hiding in encrypted image,”IEEE Trans. Inf. Forensics Security, vol. 7, no. 2, pp. 826–832, Apr

16. Conclusions
Present a new framework which allows the numerous RDH schemes developed before for non-encrypted images be conducted in the encrypted domain directly.
A new specific stream encryption algorithm is proposed to preserve some correlation between the neighboring pixels.

17. Thank you!