Cryptography Reference
In-Depth Information
Figure 17.3: The JPEG compression function often leaves strong vi-
sual artifacts around hard edges. This image was created by com-
pressing text with a very low quality, a process that encourages the
artifacts by limiting the number of coefficients kept.
In some cases, file compression algorithms can leave large areas
with the same least significant bits. The JPEG algorithm, for instance,
stores each image as the weighted sum of some cosine functions.
If the 64 pixels in the
8
8
block are sufficiently similar, the algo-
rithmwill simply store the average value. The GIF algorithmwill also
replace similar colors with the same value in the interest of saving
space. Both of these effects conspire to prevent the least significant
bits from being truly random.
The JPEG algorithm can also leave artifacts, a factor illustrated by
Figure 17.3. In this figure, the compression quality was deliberately
kept extremely low to amplify the artifacts that seem to echo around
the hard edges. If the coefficients of the JPEG image are tweaked to
encode the message, the echos and artifacts will be inconsistent and
this inconsistency may even be detectable by the eye. In most cases,
statistical techniques will be more powerful and many of the algo-
rithms described later in this chapter are sensitive to the existence of
these artifacts. The algorithms often become less adept at detecting
steganography when the JPEG algorithm is used with higher quality.
×
17.4.2 Aural Attacks
Skilled audio technicians and musicians can often hear changes ig-
nored by the average pair of ears. Many creators of watermarks for
music systems find it simple to inject extra echos in places that are
never heard by most people. The trained ears, however, can detect
them immediately.
Average ears can pick up information if the data is normalized.
Most techniques depend on the way that the human brain picks up
