Cryptography Reference
In-Depth Information
the Cat from the White House page. Another might be a quilt from
the page of some quilting club that is the picture of innocence. Any-
one who wanted to recover the information would get all of these
GIFs from the Net, recover the secret bits from all of them, and add
them up to recover the hidden data.
The net effect of this trick is deniability. No one can be sure who
was the one who was hiding the secret bits. Could it be the White
House? They've been known to sponsor covert missions based in the
Old Executive Office Building. Could it be the quilting circle? No one
knows who injected the secret bits into the file. Even the person who
recovers the bits might not know who was sending the message. It's
quite a ruse.
The section beginning
on page 99 describes
sophisticated ways of
matching patterns in
the least significant bits.
There are some practical problems associated with this tech-
nique. First, you must keep the file creation dates secret. The one
GIF that actually contains the message will be the newest file. HTTP
doesn't usually ship this information toWeb browsers so there is little
problem with keeping the information secret. But you can also fake
it by resetting the clock on your machine.
Second, you should search out GIF files that seem to have the
right structure for storing secret bits. This will prevent someone from
examining the files and discovering that only one of them has the
right structure to hide bits. That is, all but one of the
n
files are 8-bit
color with color tables filled with 256 different shades.
Third, you should worry about one of the images disappearing
from the Net. It's tempting to use images from other web sites for
parts because it will deflect attention and hide the source of themes-
sage, but this could be thwarted if someone redesigns a web site.
You can add some error-correcting features to this scheme if you
want to create, say, three different sets of files. When each set of the
three sets of files are combined, then three versions of the hidden
bits emerges. Any disparities between the files can be resolved by
choosing the value of the bit in question that is correct in two out of
three files.
More complicated error-correcting schemes like the ones de-
scribed in Chapter 3 can also be used successfully. For instance, a
file to be hidden could be encodedwith an error-correcting code that
converts every 8 bits into, say, a 12-bit block that can recover errors.
One bit from each of the 12-bit blocks could be placed into 12 sepa-
rate files that were then hidden in 12 different GIFs sprinkled around
the network. If someone could not recover all 12 GIFs because of
network failures, then the error-correcting code will allow the infor-
mation to be recovered.
