![](bg2.png)
operation of losing compression (Hsu, 1999). This
motivates researchers to realize the importance of
perceptual modeling of the human visual system and
the need to embed a signal in perceptually
significant regions of an image, especially if the
watermark is to survive losing compression (Cox,
1997). In the spatial domain block based approach
this perceptually significant region is synonymous of
low variance blocks of the cover image.
Since the meaning of multimedia data is based
on its content, it is necessary to modify the
multimedia bit-stream to embed some codes, i. e.
watermarks, without changing the meaning of the
content. The embedded watermark may represent
either a specific digital producer identification label,
or some content-based codes generated by applying
a specific rule. Because the watermarks are
embedded in the data content, once the data is
manipulated, these watermarks will also be modified
such that the authenticator can examine them to
verify the integrity of the data. For complete
verification of uncompressed raw multimedia data,
watermarking may work better than digital signature
methods because: (a) the watermarks are always
integrated with the data such that the authenticator
can examine them conveniently, and (b) there are
many spaces in the multimedia data to embed the
watermarks without degrading the quality too much
(Yeung, 1997). However, there is no advantage to
use the watermarking method in a compressed
multimedia data for complete verification.
Compression standards e.g. JPEG or MPEG have
user-defined sections where digital signatures can be
placed. Because multimedia data are stored or
distributed in the file format instead of pixel values,
therefore the digital signature can be considered as
being “embedded” in the data. For content
verification, a watermarking method that can
reliably distinguish compression from other
manipulations still has not been found. The
watermarks are either too fragile for compression or
too flexible for manipulation.
So far, the robust watermarking systems found in
the literature can only withstand some of the
possible external attacks but not all. The attacks
against the watermark try to neutralize the
watermark, without damaging the image too much.
The watermark is neutralized if: (a) the detector
cannot detect the watermark (distortion, attenuation
etc.), (b) the detector cannot recognize the
watermark in the image from another one, and (c)
the watermark is no longer in the image. The attacks
can be very different: (a) in the spatial domain, it can
be scaling, cropping, rotation, noise addition. The
open source software STIRMARK available on the
Web generates many of these attacks, (b) in the
frequential domain it can be filtering, (c)
compression and (d) adding another watermark over
the first one. The STIRMARK software generates
random rotations and distortions on blocks of the
image. The STIRMARK software simulates JPEG
coding, filtering operations, rotation, scaling and
cropping. The result is very slight alterations on
image, but watermarks are usually heavily damaged.
The present paper describes a computationally
efficient block-based spatial domain authentication
technique for a one level watermark symbol. The
selection of the required pixels is based on variance
of the block and watermark insertion exploits
average color of the blocks. The proposed
algorithms were tested on a few of the most usual
transformations of images and the obtained results
showed that the proposed method is efficient. The
authentication method developed below works for
all types of digital image and it can be applied in
medical domain because it can be inserted into
images immediately when the image is obtained by a
medical apparatus.
2 AUTHENTICATION
ALGORITHMS
All watermarking methods share the same building
blocks – an embedding system and the watermark
extraction or recovery system (
Hsu et al., 1999). Any
generic embedding system should have as inputs: a
cover data/image (I), a watermark symbol (W) and a
key (k) to enforce security. The output of the
embedding process is always the watermarked
data/image (I’). The generic watermark recovery
process needs the watermarked data (I’), the secret
key (K) and depending on the method, the original
data (I) and/or the original watermark (W) as input
while the output is the recovered watermark with
some kind of confidence measure for the given
watermark symbol or an indication about the
presence of watermark in the cover image under
inspection. The original cover image I is a standard
image of size N
N where N = 2
p
with a 24 bit RGB
format (in medical domain the image is read in a
.bmp format). In the proposed work a binary image
of size 256
256 or 512
512 is considered. Each
image is marked with a watermark coefficient. That
means, for each pixel, the value of the pixel is
changed according to the given formula:
D(i,j) = C(i,j) + a*M*W (1)
SECRYPT 2006 - INTERNATIONAL CONFERENCE ON SECURITY AND CRYPTOGRAPHY
304