Information Technology Reference
In-Depth Information
Simulation-Based Exploration of SVD-Based Technique
for Hidden Communication
by Image Steganography Channel
Vladimir Gorodetsky and Vladimir Samoilov
St. Petersburg Institute for Informatics and Automation
39, 14-th Liniya, St.Petersburg, 199178, Russia
{gor,samovl}@mail.iias.spb.su
Abstract.
The paper presents an empirical study of the
properties
of a new im-
age-based information hiding technique that are critical for hidden communica-
tion. The technique is based on the Singular Value Decomposition (SVD) trans-
form of a digital image and uses embedding a bit of data through slight
modifications of a linear combination of singular values of a small block of the
segmented cover image. The primary objective of the study is to establish the
dependence between the capacity rate of the invisibly embedded data and ro-
bustness of the steganographic channel distorted by JPEG compression while
varying embedding procedure attributes. The second objective of the empirical
study is to establish practical recommendations concerning adjusting of the con-
trollable attributes of the SVD-based data embedding procedure given the mes-
sage size and the threshold for Bit Error Rate. The results of the study provide
evidence that the developed information hiding approach is promising for hid-
den communication.
1 Introduction
Hiding information in digital images is a rapidly developing area providing an effec-
tive way for
information assurance
in covert communications, for
watermarking
of
digital objects and for other applications. Unlike cryptography that aims at hiding the
message content, the goal of information hiding is the concealment of the very fact
that a message exists.
The common requirement to information hiding techniques is providing invisibility
of a message. Given invisibility, quality of a hiding technique is determined by
rate
of
the hidden data and
robustness
to common and specific types of intentional distor-
tions. These properties are always contradictive and particular applications favor one
of them [1].
To date a number of techniques for hiding information in digital images have been
developed. Most of them are of value because theory and practice proved that there is
no one superior technique. Each technique has its own advantages and flaws and the
overall evaluation of a technique is application dependent. The overviews of hiding
information techniques are given in [7], [8], [12], [13], [14], [21], [27], etc. Embed-
ding techniques that insert data into the spatial image domain by the use of a selected
