Information Technology Reference
In-Depth Information
Cover
image
Y
Embedding
procedure
Stegodata
)
JPEG
com-
pression 1
JPEG
com-
pression 2
Extraction
procedure
Message
Z
(in
bmp
format)
S
S
S
S
(
X
,
)
JPEG
2
JPEG
1
S
(
X
,
)
JPEG
Channel dis-
tortion
1
JPEG
Intentional
distortion
2
S
(
X
,
Y
(in
bmp
format)
Messa
ge
X
Extracted
Message
Z
Message Sender
Transmission channel
Message receiver
Fig. 1.
Model of hidden communication and embedded message distortion
The attributes varied in the simulation procedure are as follows:
•
The size of the small block used in the cover image segmentation:
8
×
8,
12
×
12
, and
16
×
16;
each of such blocks is used for embedding a bit of data;
•
The attribute specifying the level of noise affecting the quality of
JPEG-
compression 1 (see Fig. 1):
10%, 20%, 40%, 60%, 80%
and
100%
;
•
The attribute specifying the extent of intentional distortion, i.e. the quality of
JPEG-
compression 2 (see Fig. 1):
10%, 20%, 40%, 60%, 80%
and
100%.
Statistical processing of the simulation results is done on the basis of 5 standard
pictures used in image-based information hiding research. They are Baboon, F16,
Lena, Brandyrose and Pepper, all of them reduced to the size
of
200
×
200
pixels.
4
Simulation Results and Discussion
×
The results of simulation averaged over 5 standard cover images of size
200
200
pixels are presented in Fig. 2 ,…, Fig. 6.
Fig. 2 illustrates the dependence of BER upon the quality (percentage) of
JPEG
compression 1, i.e. upon the distortion of steganography channel, for different seg-
mentation of the cover image. One can see that for segmentation of cover image into
blocks of size
16
×
16
and
for
JPEG
compression up to 20% the
averaged value of
BER is no more than 15% and for
JPEG
of 10%, which corresponds to an extremely
high distortion, BER is about 18%. The 15% limit of the BER value for blocks of size
12
×
12
is maintained up to
50%
of
JPEG
quality. It is natural that an increase of the
block size leads to the decrease of BER (i.e. to the increase of robustness). It should
be noticed that in the last case the “price of the increased robustness” is the decrease
of the embedded data capacity. This fact is demonstrated in Fig. 3 showing the de-
pendence of BER on the capacity rate of the embedded data. Let us notice that each
point of the plot presented in Fig. 3 implicitly corresponds to different percentages of
JPEG
compression and the latter leads to the modification of the transmission channel
payload due to the decrease of resulting length of the compressed file along with the
decrease of the
JPEG
compression quality while preserving the length of the embed-
ded data.