Simulation-Based Exploration of SVD-Based Technique for Hidden Communication by Image Steganography Channel - Computer Network Security - page 357

100

95

Lower threshold

90

85

80

Rate of the message
to be embedded

75

0.010 0.015 0.020 0.025

0.005

Rate of the embedded bits

16x16 12x12 8x8

Fig. 6. Selection of an appropriate segmentation for a particular message transmitted by noisy
channel distorted by JPEG compression 1 of given quality

In practice of hidden communication it is important to provide a required quality of
message delivery given the message size. The recommendation can be formulated on
the basis of the simulation results discussed above. For example, Fig. 6 shows how
one can choose a type of SVD-based embedding procedure, i.e. how to choose the
size of block of cover image segmentation given both the length of the message and
the requirement message survivability if the latter is specified in terms of the lower
threshold of the rate of bits to be recovered correctly. For example, if the percentage
of the correctly transmitted bits of a message has to be no more than 92% and the rate
of message given cover image length is 0.009 than the admissible selection of the
cover image segmentation is 8x8 . The admissible selection area is shown in Fig. 6.

4 Conclusion

The paper presents results of an empirical study of a novel approach to information
hiding in digital images intended for hidden communication . The approach is based
on SVD transform of small blocks of the image matrix specifying it in bmp format.
This approach is blind, i.e. extraction of hidden message does not require possessing
the undistorted cover image. The approach is potentially robust since it embeds a bit
of data through slight modification of largest singular values of the image segmenta-
tion blocks, i.e. embeds a message into low frequency.
Indeed, an extended empirical study, the main subject of the paper, showed that the
approach provides survivability of hidden messages of a sufficiently high rate of ca-
pacity even if transmitted through a noisy steganography channel and subjected to
such a destructive distortion as JPEG compression. For example, it is capable of sup-
porting the payload rate of 0.005 (in respect to the size of the cover image in bmp
format) if the required BER is no more then 0.015 (see Fig. 6). A result of the study is
that intentional distortion imposed by the noisy steganography channel ( JPEG 2)
leads to the significant, higher than linear, BER increase.