Cryptography Reference
In-Depth Information
conclude that the JPEG algorithm doesn't really care what is in the
last 4 bits. The algorithm determined that those 4 bits could be set
to any value and the resulting image would still look “good enough.”
That means that 4 bits are available to hide information. Elsewhere
in the image, all 8 bits of
JPEG
−
1
(
f
JPEG
(
f
))
.
might agree with
Then no information can be hidden in these bits.
This algorithm makes it possible to identify the locations of im-
portant parts of the image. You can choose the right accuracy value
for JPEG as well. If you need a good final representation, then you
should use the best settings for JPEG and this will probably identify
a smaller number of bits available to hide data. A coarser setting for
JPEG should open up more bits.
There are many other compression algorithms being developed
to hide information. The fractal compression algorithms fromBarns-
ley's Iterated Systems [BH92] are some of the more popular tech-
niques around. Each could be used in a similar fashion to identify
sections of the image that can be successfully sacrificed.
Other solutions that are tuned to different types of images can
also be used successfully. For instance, there are some algorithms de-
signed to convert 24-bit color images into 8-bit color images. These
do a good job of identifying 256 colors that represent the image. You
can identify the number of free bits at each pixel by comparing the
24-bit value with the entry from the 256-color table that was chosen
to replace it. Some of these algorithms are tuned to do a better job
on faces. Others work well on natural scenes. Each is applicable in
its own way and can do a good job with the system.
If you use JPEG or a similar lossy algorithm to identify the high-
noise areas of an image, then you must change one crucial part of
the system. When a GIF file is used to hold information, then the
recipient doesn't need to have a copy of the original image. The
n
least significant bits can be stripped away and recovered. They
can be used verbatim. If JPEG is going to point out the corners and
crevices of the image waiting for more data, then both the sender
and the recipient must have access to the same list of corners and
crevices. Probably the easiest way to accomplish this is to make sure
that both sides have copies of the original image. This is a limitation
if you're going to communicate with someone whom you've never
met before. You must somehow arrange to get the image to them.
9.5.1 Hiding Information in JPEG Files
There is no doubt that the JPEG's lossy approach to hiding infor-
mation is a problem that confounds the basic approach to stega-
