Information Technology Reference
In-Depth Information
14
Bound
Bound of proposed method
LSB embedding
Ternary embedding
12
10
8
6
4
2
0
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
Embedding rate
Figure 10.6
Performance comparison between the proposed method, LSB embedding, and ternary
embedding.
changes on average. We also observe that an increase of
N
results in a low
change density. Low change density preserves the quality of the stego-image
and allows the embedding to remain imperceptible to bad actors, resulting
in improved steganographic security. As a result, the proposed method is
best among these conventional ±1 message embedding techniques.
Finally, we close this section with a note on how the proposed method
improves security. In the proposed method, the impact of embedding becomes
undetectable since it is limited to ±1 embedding changes. Further, the DM
allocation function enables communication without sharing the placement of
embedding changes. The warden is unable to read the secret messages because
the placement of the embedding changes is not communicated directly. Thus,
the proposed method minimizes the detectability of the hidden data and can
approach the theoretically upper bound of embedding efficiency.
10.6 Conclusions
In this paper, we propose a novel message embedding method that uses the
concept of DM allocation method. We limit ourselves to so-called ±1 embed-
ding changes, which requires that the sender modifies each pixel by at most
one, the smallest possible modification. We also give a theoretically achiev-
able bound on embedding efficiency for a given embedding rate. In the pro-
posed method, log
2
3
q
bits of messages can be embedded in (3
q
- 1)/2 pixels by
performing at most one embedding change. Therefore, our communication
setup provides better embedding efficiency compared with traditional ±1
message embedding methods.
Moreover, the DM allocation method-based message embedding enables
communication without the need to share the placement of the embedding
