Information Technology Reference
In-Depth Information
TaBLe 10.2
Example of Ternary Embedding
c
i
m
i
= 0
m
i
= 1
m
i
= 2
s
i
= 12
D
= 0
s
i
= 13
D
= 1
s
i
= 11
D
= 1
12
s
i
= 12
D
= 1
s
i
= 13
D
= 0
s
i
= 14
D
= 1
13
s
i
= 15
D
= 1
s
i
= 13
D
= 1
s
i
= 14
D
= 0
14
is more efficient than LSB embedding. After studying some message embed-
ding schemes, we propose a ±1 message embedding scheme using a DM
allocation method to minimize the impact of embedding.
10.4 Proposed Scheme
According to Willems and van Dijk [13], the impact of embedding becomes
detectable whenever the maximum allowable error is larger than one; thus,
we limit ourselves to ±1 embedding changes. The goal of message embed-
ding is to design schemes that have a high embedding rate but a low embed-
ding change rate. Thus, a DM allocation method is applied to the proposed
scheme to decrease the embedding change rate. It also enables improved
security in communication by eliminating the need to share the selection
channel. Let us now review a few basic concepts from the DM allocation
method that are relevant to explain the proposed scheme.
10.4.1 DM allocation Method
Throughout this paper, we use some standard concepts and results from DM
allocation technology. A file cannot reside in the memory in a large database;
thus, all records are divided into buckets and stored on disks. The task is to
allocate files among accessible disks to maximize disk access concurrency
and therefore minimize response time. The problem of disk allocation for
Cartesian product files on multiple disk systems was first considered by Du
and Sobolewski [10]. They proposed an allocation method that assigns all
buckets of a Cartesian product file to an
ND
-disk system and show that it is
strictly optimal. Before describing the DM allocation method, it is necessary
to state relevant definitions and assumptions.
Let
F
be an
N
-attribute binary Cartesian product file. Each bucket [
b
1
,
b
2
, …,
b
N
] is assigned to a disk unit
