08-01-2013, 10:18 AM
A Statistical Watermark Detection Technique Without Using Original Images for Resolving Rightful Ownerships of Digital Images
A Statistical Watermark.pdf (Size: 458.36 KB / Downloads: 52)
Abstract
Digital watermarking has been recently proposed
as the mean for copyright protection of multimedia data. Many
of existing watermarking schemes focused on the robust means
to mark an image invisibly without really addressing the ends
of these schemes. This paper first discusses some scenarios in
which many current watermarking schemes fail to resolve the
rightful ownership of an image. The key problems are then
identified, and some crucial requirements for a valid invisible
watermark detection are discussed. In particular, we show that,
for the particular application of resolving rightful ownership
using invisible watermarks, it might be crucial to require that the
original image not be directly involved in the watermark detection
process. A general framework for validly detecting the invisible
watermarks is then proposed. Some requirements on the claimed
signature/watermarks to be used for detection are discussed to
prevent the existence of any counterfeit scheme. The optimal
detection strategy within the framework is derived. We show the
effectiveness of this technique based on some visual-model-based
watermark encoding schemes.
INTRODUCTION
THE RAPID growth of digital imagery, the increasingly
easy access to digital media, and the increasingly powerful
tools available for manipulating digital media have made
media security a very important issue. Digital watermarks have
been proposed recently as the means for intellectual property
right protection of multimedia data. Digital watermarking is a
process of embedding information (or signature) directly into
the media data by making small modifications to them. With
the detection/extraction of the signature from the watermarked
media data, it has been claimed that digital watermarks can be
used to identify the rightful owner, the intended recipients, as
well as the authenticity of a media data [1], [2].
INVALIDATING CLAIMS OF OWNERSHIPS
We first describe a generalized formulation of a class of
current invisible watermarking schemes [2], [6], [7], [9]–[11],
[4]. Fig. 1, which is similar to what have been described in [4]
and [10], illustrates both the watermark encoding and decoding
processes.
Given the original image , one generates the signature ,
and then embeds into to create a watermarked image
which should be visually close to . In general, the embedding
process could be an arbitrary function of the original image
and the signature . However, in many watermarking schemes
[2], [6], [7], [9]–[11], the embedding operator is simply the
addition operator.
In watermark detection, given the test image assumed to
have been marked and whose ownership is to be determined,
one first extracts the signature from , usually making
use of the original image . Then is compared to the
original signature . The similarity index is then compared to
a threshold to determine if the test image is a watermarked
version of the original image.
It is a common view that invisible watermarking schemes
may be used to protect the rights of image owners. This is
because the signature extracted from the watermarked images
will match the signature of the owner. However, it has been
shown in [4] that a counterfeit watermarking scheme can allow
multiple claims of ownerships. In other words, “the true owner
of an image can no longer argue his claim based only on the
digital watermarks that he invisibly embedded in the image,
as others can engineer an equal amount of evidence that they
too own the image” [4].
Detecting Watermarks with a Quantitative Measure
Although the signature in a watermarked image is perceptually
invisible, it should be “statistically visible.” In other
words, it should be detectable using some statistical techniques
without looking at other images. Fig. 2 shows a general
architecture of our proposed watermarking system that we
believe is effective for resolving the rightful ownership. To
be more concrete, we focus on the so-called feature-based [4]
watermarking schemes [2], [9]–[11] in which an independent
identically distributed (i.i.d.) pseudorandom sequence is
embedded into a set of features derived from the original
image . For example, the feature set could be a subset
of all the discrete cosine transform (DCT) coefficients of 8
8 blocks in the original image [10]. The signature is
modulated by , where controls the strength of the
watermarks, and could potentially be a function of the value
of . For example, in perceptual model based watermarking
systems [10], [11], is derived from the characteristics of
the human visual systems.
Optimal Choice of and Its Validity
The choice of has a great impact on the detection validity
and the detection capability. For a valid claim of ownership,
should satisfy some constraints. First, the creation process
of should be independent of any image. This is to guarantee
that is not obtained by analyzing the test image, but
instead created by the true owner when the watermarks were
inserted. We will propose a way to guarantee that is created
independent of the test image in Section III-C. Second, to
determine the threshold for a given false alarm detection
probability without access to the original image, it is necessary
that the mean value of is zero. With these two constraints,
under Hypothesis where the test image is not watermarked,
the expected output value of is alway zero. We call a
random sequence that satisfies the above two constraints a
valid correlating signature. Of course, for the benefit of the
owner, (5) suggests that should be chosen to be closely
correlated with .
CONCLUSION
This paper first identifies the key problems existing in many
current watermarking schemes regarding resolving the rightful
ownership of an image. We show that, for the particular
application of resolving rightful ownership using invisible
watermarks, the original image should not be directly involved
in the watermark detection process. Some other requirements
on the claimed watermark sequence are also identified and
discussed. We believe that, for a valid claim of ownership, it
is necessary to require that the claimed watermark sequence
be generated from some meaningful signature or registered
owner ID through a potentially certified one-way deterministic
function.