An Agent based Image Steganography using Information Theoretic ...

MASAUM Journal of Computing, Volume 1 Issue 2, September 2009 261
" As Agent is a computer system that is situated in
some environment and that is capable of autonomous
action in this environment in order to meet it design
objective".
Let us note that we are talking about (software agent)
whenever or any other researchers in the field say (agent),
we really mean software agent. The typical dictionary
definition of agent, as An entity having the authority to act
on behalf of another [14].
Some definitions were tied to specific implementation
technology such as being based on theorem proves , or
using internal data structure corresponding to the socalled
mentalist concepts , such as beliefs or knowledge ,
goal or desires , intention , and so on.
So a good working definition of agent is that [15]:
"It is a persistent computational entity that can
perceive, reason, Act, and communicate".
A. Software Agent Properties
The basic properties of software agent are that they are:
• Autonomous: being autonomous, mean that agents are
independent and make their own decision, this is one
property that distinguish agent from object.
• Situated ness : dose not constrain the notion of an agent
very much since virtually all software can be consider to
be situated in an environment [14].
• Flexibility : can be define to include the following
property :
I. Responsive: Refer to agent ability to perceive its
environment and respond in a timely fashion to change
that occur in it.
II. Pro-active : Agent are able to exhibit opportunistic ,
goal –driven behavior , take initiative where appropriate
III. Social: Agent should be able to interact, where
appropriate, with other agent or human in order to solve
their own problem and help other with their activities.
B. Agent Classification
The various definitions involve a host properties of an
agent. Having settled on a much less restrict definition of
an agent, this property may help us further classify agents
in useful ways.
Agent may be usefully classified according to the subset
of these properties that they enjoy. Every agent by our
definition, satisfies the first four property, adding other
property reduce potentially useful classes of agents. For
example: mobile, learning agent. Thus a hierarchical
classification based on set inclusion occurs naturally.
There are of course other possible classifying scheme ,
for example , we might classify software agent according
to the tasks they perform. For example, information
gathering agents or email filtering agent , or we might
classify them according to their control architecture. Then
would be fuzzy agent. Also agent could be classified by
the range and sensitivity of their senses , or by the much
internal state they posses [16].
IV. PROPOSED IMAGE STEGANORGRAPHY BASED ON
AGENT
This paper uses two methods of steganorgraphy, the
first one is the discrete cosine transform (DCT)
Steganography , and the second one is least significant bit
(LSB) steganorgraphy .
A. The Proposed Agent System in steganorgraphy
Our approach make general assumption about agent .
Also in this section we dedicate to present images
database which used by the agents. According to some
features of media (image) the agent will recommend the
user to select the best cover (image) and the suitable
image steganorgraphy methods that are assumed to be
available in this project for the selected secret image .
The Model of proposed steganorgraphy based agents
shown in Fig. 2, can be classified as: server agents, which
provides services to other. Since the Client- Server
connection is assumed to be used in this system, the
server (PC ) will provide the client with the STEGO.bmp
file ,which contains the secret image.
The basic role of agents in the system is to select image
from a database of images and make analyses of number
of image features , and according to these features, the
agent will choose the suitable steganography method and
best cover for the specific steganography method, and
give the user a recommendation for selected cover and
selected steganography system .
B. Agent system for Image feature calculation
In this system, the agent will place at random on
image (the environment). The choice of cover image is
important because it is significantly influences the result
obtained from the proposed system, and it will affect the
resulted security of the whole system.
The calculated image features by agent are:
• Histogram : is a plot of gray level values versus the
number of pixels. The histogram used as probability
distribution of gray level:
P (g) = N ( g)
(3)
M
Where M :is number of pixels in image .
N (g): is number of pixels at gray level g.
• Mean: is the average value, so it tells something
about general brightness of the image.
−
I( r,
c)
(4)
g = ∑∑
r c M
Where I(r,c) is the image pixel value.
• Standard deviation : which is known as square
root of variance. It tells something about contrast ,
and describes and expressed as follows:
L 1 −
=
σ ( g − g ) p(
g)
(5)
g
g∑ −
= 0
261