Cryptography Reference
In-Depth Information
limited in their power. The mimic functions described in Chapters
7 and 8 are theoretically hard to break. The classical work from Kurt
Goedel, Alan Turing, and others firmly establishes that there can't be
computer programs that analyze other computer programs.
Such theoretical guarantees are comforting, but they are rarely
as strong as they sound. To paraphrase Abraham Lincoln, “You can
fool all of the computer programs some of the time, and some of the
computer programs all of the time, but you can't fool all of the com-
puter programs all of the time.” Even if no program can be created
to crack mimic functions all of the time, there's no reason why some-
thing might not detect imperfections that happen most of the time.
For instance, the software for hiding information in the voice-over of
baseball games is going to become repetitive after some time. Some
form of statistical analysis may reveal something. It won't work all of
the time, but it will work some of the time.
The best the steganographer can do is constantly change the pa-
rameters and the locations used to hide information. The best the
steganalyst can do is constantly probe for subtle patterns left by mis-
take.
17.4 Visual and Aural Attacks
The simplest form of steganalysis is to examine the picture or sound
file with human eyes or ears. Our senses are often capable of com-
plex, intuitive analysis that can, in many ways, outstrip the power of
a computer. If the steganographic algorithm is any good, the changes
should not be apparent at first.
17.4.1 Visual Attacks
Hiding the information from human eyes is the first challenge. Some
basic algorithms will make mistakes and make large color changes
in the process of hiding the information, but most should produce a
functionally identical image or sound file.
But a bit of computer enhancement can quickly make a hidden
message apparent to our eyes. If the most important parts of the
image are stripped away, the eye can often spot encoded informa-
tion without any trouble. Figures 17.1 and 17.2 show the least signifi-
cant bits of an image before and after information is hidden with the
EzStego program.
The figures illustrate how the least significant bits in an image are
often far from random. Notice how the saturated areas of the image
