Cryptography Reference
In-Depth Information
•
If the data is hidden in the least significant bit of pixels or audio
file elements, then the canonical form can be found by setting
the least significant bit to zero.
•
If the information-hiding process modifies the elements by no
more than
then canonical points can be established as they
were above. Let
±
{
0
,
2
,
4
,
6
,...}
be the set of canonical points.
Only one will lie in the range
x
i
− <x
i
≤ x
i
+
.
•
Sentences can be put into canonical form. Mikhail Atallah and
Victor Raskin use natural language processing algorithms to
parse sentences.[ARC
+
01] The sentence “The dog chased the
cat” takes this form in their LISP-based system:
(S
(NP the dog)
(VP chased
(NP the cat)))
The letter 'S' stands for the beginning of a sentence, the letters
'NP' stands for a noun phrase, and the letters 'VP' stands for a
verb phrase. If the sentence can't be parsed, it is ignored.
Their solution hides information by applying a number of
transformations like switching between the active and passive
voice. One bit could be encoded by switching this example to
read “The cat was chased by the dog”. Others solutions they
use include moving the adjunct of a sentence, clefting the sen-
tence, and inserting extra unnecessary words and phrases like
“It seems that...”
A canonical form can be defined by choosing one version of the
transformation. In this example, the active voice might be the
canonical form for the sentence.
13.6 Packing inMultiple Messages
Sorting can also let you store multiple messages in a collection of
data. If
f
2
defines another order, then both
orderings can be used to hide information if the sizes are right. If the
number of elements in the camoflauging data ismuch larger than the
amount of information being hidden, then the chances of a collision
are small.
The chances of a collision are easy to estimate if you can assume
that the sorting functions,
f
1
defines one order and
f
1
and
f
2
, behave like random number
