Cryptography Reference
In-Depth Information
Steganography
For use
against detection
For use
against removal
For use
against forgery
fragile/semi-fragile
watermarking
data
hiding
robust
watermarking
Fig. 1.1.
The classification of watermarking [16].
the latter case, the inserted watermark (W ) is often necessary to check the
identity of the watermark.
These generic models are directly applicable to the chapters dealing with
data hiding topics.
Mathematical notions are also used to express the processes in Fig. 1.2.
We can view the
embedding
process as a function or mapping that maps the
inputs X, W and/or K to the output X
′
.Thatis,
′
X
= E(X, W , [K]),
(1.1)
where E() denotes the embedding process, and [K] indicates that K may
not be included. Similarly, the
decoding
or
extraction
process, D(), can be
denoted by
′
′′
W
= D(X
, [X], [K]).
(1.2)
The
detection
process, d(), is
′′
YesorNo= d(X
, [X], W , [K]).
(1.3)
Hence, [] means that the element in the bracket may be optional.
1.2.2 Some Requirements of Digital Watermarking
There are many metrics used to measure the effectiveness of a watermarking
algorithm. From algorithm design viewpoint, the three most critical require-
ments are
watermark imperceptibility
,
watermark robustness
,and
watermark
capacity
. Although these three requirements are all very desirable, as pointed
out in literature [17, 18, 19, 20, 21, 22], they influence, or even conflict, with
each other. After fixing one dimension, the remaining two may then have con-
flicts between one another. Some tradeoff must then be developed [23]. The
interrelationships may be threefold.
