INTEGRITY AUTHENTICATION METHOD FOR JPEG IMAGES
USING REVERSIBLE WATERMARKING
Hyun-Wu Jo, Dong-Gyu Yeo and Hae-Yeoun Lee
Dept. of Computer Software Engineering, Kumoh National Institute of Technology,
Sanho-ro 77, Gumi, Gyeongbuk 730-701, Korea
Keywords: Image Authentication, JPEG Compression, Reversible Watermark, DCT Coefficient.
Abstract: In these days, with increasing the importance of multimedia security, various multimedia security
techniques are studied. In this paper, we propose a content authentication algorithm based on reversible
watermarking which supports JPEG compression commonly used for multimedia contents. After splitting
image blocks, a specific authentication code for each block is extracted and embedded into the quantized
coefficients on JPEG compression which are preserved against lossy processing. At a decoding process, the
watermarked JPEG image is authenticated by extracting the embedded code and restored to have the
original image quality. To evaluate the performance of the proposed algorithm, we analyzed image quality
and compression ratio on various test images. The average PSNR value and compression ratio of the
watermarked JPEG image were 33.13dB and 90.65%, respectively, whose difference with the standard
JPEG compression were 2.44dB and 1.62%.
1 INTRODUCTION
Multimedia contents can be copied and manipulated
without quality degradation. Therefore, they are
vulnerable to digital forgery and illegal distribution.
In these days, with increasing the importance of
multimedia security, various multi-media security
techniques are studied.
Digital watermarking can be an efficient solution
to protect multimedia security and digital right
management, which inserts confidential information
called as the watermark into multimedia contents
themselves. By retrieving the inserted watermark
from the contents, it can be used for various
purposes including ownership verification, copyright
protection, broadcast monitoring, and contents
authentication, etc.
In this paper, we propose a content authentica-
tion algorithm based on reversible watermarking
which supports JPEG compression commonly used
for multimedia contents. After splitting image
blocks, a specific authentication code for each block
is ex-tracted and embedded into the quantized
coefficients on JPEG compression. At a decoding
process, the watermarked JPEG image is
authenticated by ex-tracting the embedded code and
restored to have the original image quality.
The paper is composed of as follows. In Sec. 2,
related researches are summarized. We present im-
age authentication algorithm for JPEG compression
in Sec. 3. Experimental results are shown in Sec. 4
and Sec. 5 concludes.
2 RELATED WORKS
The purpose of image authentication is to detect
tempering and to prove the integrity of the image. In
the various types of watermarking algorithms,
fragile watermarking is easily corrupted by slightest
modification and applicable for this purpose.
Yuan and Zhang proposed a fragile watermark-
ing method for image authentication based on statis-
tical analysis in the wavelet (Yuan, 2003). Hu and
Han suggested a semi-fragile watermarking algo-
rithm for image authentication (Hu, 2005), in which
image features are extracted from the low frequency
domain to generate two watermarks. One feature is
used to classify the intentional content modification
and the other is used to indicate the modified loca-
tion. A compressed-domain fragile watermarking
scheme with discrimination of tampers on image
content or watermark was studied (Wang, 2009).
Li proposed a transform-domain fragile
83
Jo H., Yeo D. and Lee H..
INTEGRITY AUTHENTICATION METHOD FOR JPEG IMAGES USING REVERSIBLE WATERMARKING.
DOI: 10.5220/0003826200830086
In Proceedings of the International Conference on Computer Vision Theory and Applications (VISAPP-2012), pages 83-86
ISBN: 978-989-8565-04-4
Copyright
c
2012 SCITEPRESS (Science and Technology Publications, Lda.)
watermarking algorithm for authentication. However,
this method uses a private key to generate binary-
sequence therefore it needs to share a private key
between sender and receiver. Furthermore, this
method doesn't consider about reversibility of
algorithm (Li, 2004).
In most previous researches, uncompressed im-
ages were considered and the quality of the original
image was degraded although it was not perceptible
to human eyes. Also, they were vulnerable against
JPEG compression since the JPEG compression
could be considered as a kind of attacks.
3 PROPOSED ALGORITHM
JPEG compression is a commonly used compression
standard for images. Therefore, the way to check the
integrity and ensure the original quality of JPEG
compressed images should be provided. In this
section, we present a JPEG image authentication
algorithm based on reversible watermarking which
is composed of two parts: insertion and
authentication. At the insertion process, the
authentication code is generated using the
characteristics of an image block and inserted into
image data itself during the JPEG compression
process. At the authentication process, the embedded
watermark is extracted and used to prove the
integrity. Also, the original quality is ensured by
removing the embedded watermark.
3.1 Watermark Insertion
The overall process to insert watermark is depicted
on Fig.
1
. Input image goes through lossy processing
step and lossless processing step on JPEG
compression. Quantized DCT (QDCT) blocks from
discrete cosine transform and quantization at the
lossy processing step do not make loss in the entropy
coding progress. Therefore, we insert the
authentication code (called the watermark) into this
QDCT blocks in a reversible way because they are
preserved against image compression.
The authentication code is generated as follows
(refer Fig. 2):
Make a 8x8 grayscale block through down-
sampling a 16x16 color block
Apply DCT to 8x8 grayscale block and
quantize DCT coefficients
Extract most significant 8 values from
quantized DCT coefficients in Zigzag order,
which is the 64 bits authentication code
Figure 1: Overall watermark insertion process.
Figure 2: Authentication code generation and splitting.
As shown in Fig. 2, the 64 bits authentication
code is split into four 16 bits sub-codes and inserted
into each 8x8 sub-blocks. One block is composed by
3 channels: Y, Cb, Cr. The 16 bits sub-code is split
again into 8, 4, and 4 bits to be inserted into each
channel.
The way to insert this authentication code is
based on histogram shifting for DCT coefficients as
shown in Fig. 3. We insert the authentication code
by shifting the zero value in DCT coefficients.
Usually, modifying the DC coefficient has effects on
quality degradation over other coefficients and its
possibility to have zero value is very low. Therefore,
Except the DC coefficient, the insertion proceeds
from most significant coefficients to least significant
VISAPP 2012 - International Conference on Computer Vision Theory and Applications
84
coefficients in zigzag order.
This watermark insertion using histogram
shifting can be formulated as equation (1). ZQDCT
is Zigzag-reordered QDCT Coefficient.

(
)
=

(
)
,=0

(
)
,
(
)
<0

(
)
+1,
(
)
>0

(
)
+
(
)
,
(
)
=0
0
(

(

)
−1
)
,
=
(1)
Figure 3: Histogram shifting for watermarking.
3.2 Image Authentication
The overall process to authenticate images and
restore the original quality is depicted on Fig. 4. The
main idea of image authentication and tamper
detection is originated from the characteristic of
JPEG compression method.
Figure 4: Overall image authentication process.
The authentication code is regenerated by using
the same steps in the watermark insertion. The way
to extract the watermark and to restore the original
quality is depicted in Fig. 5 and can be modelled as
equation (2) and (3).

()
=
, =0

()
,
()
{0,1}
0
(

(

)
−1
)
,
=

−

(2)

()
=

()
,=0

(
)
,
(
)
<0

(
)
−1,
(
)
>0
=

−

(3)
Figure 5: Image restoring by inverse histogram shifting.
4 EXPERIMANTAL RESULTS
We evaluated the proposed algorithm using 8 images
which are 512x512 8 bits color images in USC-SIPI
image database (see Fig. 6).
First, the image quality between an original
JPEG compressed image and its watermarked JPEG
compressed image is compared and summarized in
Table 1. The PSNR value is calculated against a
non-compressed original image.
Table 1: PSNR comparison (unit: dB).
Images JPEG w.JPEG Difference
Airplane 37.68 34.33 3.35
Baboon 31.64 30.31 1.33
House 35.66 32.95 2.71
Lena 36.34 33.75 2.59
Peppers 35.16 33.02 2.14
Sailboat 33.55 31.82 1.73
Splash 38.42 35.18 3.24
Tiffany 36.13 33.68 2.45
Average 35.57 33.13 2.44
Although the watermarked JPEG images showed
tiny noisy pattern on flat area, there were not easily
recognizable. Also, note that the original quality can
be restored in our algorithm. The PSNR degradation
is 2.44dB on average and it supports that the water-
mark insertion does not degrade the image quality.
INTEGRITY AUTHENTICATION METHOD FOR JPEG IMAGES USING REVERSIBLE WATERMARKING
85
Figure 6: 8 Test images for experiments.
Table 2: Compression ratio comparison (Unit: %).
Images JPEG w.JPEG Difference
Airplane 93.77 92.27 1.50
Baboon 86.69 84.79 1.90
House 91.84 90.26 1.58
Lena 93.77 92.11 1.66
Peppers 92.90 91.80 1.10
Sailboat 91.07 89.37 1.70
Splash 94.59 93.19 1.40
Tiffany 93.55 92.00 1.55
Average 92.27 90.65 1.62
We compared image compression ratio between
an original JPEG compressed image and its
watermarked JPEG compressed image (refer Table
2). The compression ratio of the original JPEG and
the watermarked JPEG was 92.27% and 90.65% on
average, respectively. The difference of each
compression ratio is 1.63% which is small enough.
5 CONCLUSIONS
This paper presented a JPEG authentication
algorithm using reversible watermarking, which can
authenticate the integrity of JPEG compressed
images because it is sensitive against attacks. Also,
it has no large difference with an original JPEG
compressed image in aspect of the image quality and
the compression ratio. The proposed algorithm
considers the re-saving or re-compression without
any modification as a kind of attacks because it
modifies the quantization DCT coefficients. Some
applications require robustness against these
processing. Future works is studying the algorithm
to support this requirement.
ACKNOWLEDGEMENTS
This research project was supported by Ministry of
Culture, Sports and Tourism (MCST) and from
Korea Copyright Commission in 2011 and Basic
Science Research Program through the National
Research Foundation of Korea (NRF) funded by the
Ministry of Education, Science and Technology
(2011-0005129).
REFERENCES
Yuan, H., Zhang, X.-P., 2003, Fragile watermark based on
the Gaussian mixture model in the wavelet domain for
image authentication, Proc. of Int. Conf. on Image
Processing, vol. 1, pp. I-505-8.
Hu, Y.-P., Han, D.-Z., 2005, Using two semi-fragile
watermark for image authentication, Proc. of Int. Conf.
on Machine Learning and Cybernetics, pp. 5484-5489.
Wang, H., Liao, C, 2009, Compressed-domain fragile
watermarking scheme for distinguishing tampers on
image content or watermark, Proc. of Int. Conf. on
Communications, Circuits and Systems, pp. 480-484.
C. T. Li, 2004. Vision, Image and Signal Processing, IEE
Proceedings, 2004, "Digital fragile watermarking
scheme for authentication of JPEG images".
VISAPP 2012 - International Conference on Computer Vision Theory and Applications
86