Cryptography Reference
In-Depth Information
domly. This would add a significant amount of texture that might
mitigate some of the effects. 3
Another significant hurdle for image- and sound-based stegano-
graphy is the compression function. The digital representations of
these data are so common that many specialized compression func-
tions exist to pack the data into smaller files for shipping across the
network. JPEG (Joint Photographic Experts Group) is one popular
standard algorithm for compressing photographs. MPEG is a similar
standard designed for motion pictures.
Both of these are dangerous for bit-level infopacking because
they are lossy compression functions. If you take a file, compress it
with JPEG, and then uncompress it later, the result will not be exactly
the same as the original. It will look similar, but it won't be the same.
This effect is quite different than the lossless compression used on
many other forms of data like text. Those functions reproduce the
data verbatim. Lossy compression functions are able to get signif-
icantly more compression because they take a devil-may-care atti-
tude with the details. The end result looks close enough. The JPEG
algorithm itself is adjustable. You can get significantly more com-
pression if you're willing to tolerate more inaccuracies. If you turn
up the compression significantly, the pixels begin to blend into big
blocks of the same color.
JPEG compression can
also help. See page 183.
9.2.5 Deniability
Deniability is one of the greatest features of hiding information in
images fromWeb pages. If you structure your information correctly,
you can spread it out among a number of unrelated locations. If the
information is discovered, it will be impossible to tell exactly where
it came from.
Imagine that you have some bits that youwant to distribute to the
world. You could hide these bits in an image file and place it on your
home page for all to download. If unintended people discover the
bits, however, they know the information came from you because it
is on your Web page.
Instead, you can split up the information into
n
parts using the
files, when they're XORed to-
gether, will reveal the hidden data. Ordinarily, you would create
basic tricks from Chapter 4. These
n
1
files at random and then compute the last file so that everything adds
up. But why bother using files at randomwhen there is a great source
of randomness on the network? You could snarf
nāˆ’
1 different GIF
images from the Net and use them. One might be a picture of Socks
n āˆ’
3 Random noise has been used to make quantization look more realistic. Too many
discrete levels look artificial [Rob62].
Search WWH ::




Custom Search