In this section, we will simply introduce public key and private key techniques in 2.3.1.
In section 2.3.2, we will describe symmetric and asymmetric watermarking scheme.
2.3.1 Public Key and Private Key Digital Watermarking System
In order to making digital watermarking systems safer now, some systems already introduce cryptography techniques among them. On the basis of using the different cryptography techniques, we divided them into public key digital watermarking system and private key digital watermarking system respectively.
In the private key digital watermarking system, sender and receiver own a same secrete key together and use a traditional private key generator, which generates a very long series of pseudo random number sequence, that can choose which bit positions to embed or extract the digital watermark.
For the public key digital watermarking system, if Tom and John have no chance to own a same private key together, but John have a public key of Tom, so John can use Tom’s public key to encrypt secret messages that he wanted to transfer, then embedded these cipher text into original system according to embedding digital watermarking methods. Only Tom owns a decrypt key, so he can extract these secret messages.
Watermarking algorithms also can be classified into fragile watermarking and robust watermarking [32] [33]. The public-key watermarking algorithm can be more effective to face the attacks. Also it has more potential value of applications. Some public-key fragile
watermarking algorithms are proposed. The Wong's algorithm [31], which based on public key cryptography, is a typical one. However, there are few public-key robust watermarking algorithms proposed since they are requested to survive severe tampering. In Wong's algorithm, any subtle change in watermarked media will make the decrypt failure. Thus Wong's method can not be applied simply to obtain a robust watermarking algorithm.
2.3.2 Symmetric and Asymmetric Watermarking Scheme
Symmetric watermarking scheme means the key used for watermark embedding must be available at the watermark detector, which uses the same sequence forembedding and
detecting, has the security problem that with only the detector [4], an attacker can easily estimate and remove the embedded watermark [29] [30]. That’s to say, symmetric means that the detection process makes use of the parameters used by the embedding process. The knowledge of these parameters allows pirates to forge illegal contents by modifying or
removing watermark. This set of parameters is called the secret key and must be stored safely.
This is not possible in consumer electronics. Tamper proof device is too expensive.
This is the reason why the cryptography domain has been recently studied [34] [35] [36].
They should be robust symmetric techniques with a detector needing a set of parameters called the public key different from the embedding’s secret key. Knowing the public key, it should be neither possible to deduce the private key nor possible to remove the watermark.
One particular problem with state-of-the-art watermarking schemes is that they are symmetric [3]. The keys necessary for watermark embedding and detection are identical.
Thus, the watermark detector knows all critical parameter of the watermarking scheme that also allows efficient removal of the embedded watermark.Using watermark technology for copy protection, the watermark detector needs to implement in many cheap consumer devices all over the world. A symmetric watermarking scheme presents a security risk, since the detector has to know the required private key. However, cheap tamper-proof devices are
hardly producible, and thus, pirates can obtain the private key from such devices and use them to outwit the copy protection mechanism. For this reason, we would like to develop a
watermarking scheme where detection of the watermark is possible with a public key that does not give enough information to impair the embedded watermark. Such a scheme is called asymmetric.
However, in general, most researchers focus on symmetric watermarking, i.e., it is different or takes lots of efforts to embed, change, remove, and extract a watermark. This will make image authentication more laborious and limit the applications of digital watermarking.
Thus, asymmetric watermarking is addressed. It demands that a watermark is easy to extract, but only authorized people can embed, change, or remove this watermark.
First, we will explain our notation and describe a general point of view on watermarking schemes below. For a better understanding of the differences between symmetric and
asymmetric schemes, we will describe both of them [3].
We view digital watermarking as a communications problem, where the watermark informationb∈β, withβdenoting the finite set of all possible watermark messages, is transmitted over a hostile channel. The host signalxserves as the carrier for the watermark information. In this paper, we adopt vector notation for signals that is
[ ] [ ] [ ] [
x x x N]
Tx= 0, 1,..., −1 withx
[ ]
i being the th signal sample. We do not focus on a specific data type. The signal x can denote audio, image or video data, or any transform domain representation of such multimedia data. In practice, watermarking schemes have to be optimized for the specific features of different host signals. Here, our intention is to compare basic concepts without considering details that are strongly dependent on the specificmultimedia data.
i
Any modification of the host signalxdoes affect its quality, thus an assessment of watermarking schemes is not possible without defining a quality measurement. Good quality
measurements are again strongly dependent on the data at hand. However, as a rough approximation, the mean squared error (MSE) between the original host signals and any modified signal can be used as a quality measurement.
Fig.2-3 depicts a general blind symmetric watermarking scheme [3]. The term “blind”
indicates that the host signalxis not known at the watermark detector. The watermark information bis embedded into the host signalxdependent on a private key. All modifications introduced by the embedding process are denoted by the watermark signalw, so that the public signal s can be expressed ass=x+w. The distortion introduced by the embedding of the watermark is given by DE = E｛(s−x)2｝= E｛ ｝. Here, E｛‧｝denotes expectation.
w2
The public signalsis subject to a variety of different attacks. We use the term attack for any signal processing that, intentionally or not, reduces the reliability of watermark detection.
The modifications introduced by the attack( ) can be summarized by the additive, but not necessarily independent, signal
s
v. Of course, an attack is useless if the attacked signal v
s
r = + has such poor quality that its value is lost. Thus, the quality of the attacked signal must be sufficiently good. Many watermarking schemes can be successfully attacked by desynchronizing the embedded watermark relative to the watermark signal the detector is looking for. We do not consider desynchronized attacks formally, but point out where synchronization is a particularly difficult problem. Assuming synchronization, the quality of the attacked signal ris measured relative to the original host signalx. We measure the
Embed
Fig.2-3 General blind symmetric watermarking scheme
distortion of an attacked signal by DA = E
{ (
r−x)
2}
.Finally, the detector computes an estimate of the transmitted watermark information according to the private key and the received signal
bˆ
b r. The probabilityPr
( )
bˆ≠b of falsedetection should be as small as possible.
The constraints on the qualities DE and DA are strongly dependent on the given data and the application in mind. However, it is reasonable to assume that the allowable DA is at least at the order of DE, and in many cases even much larger. We use the ratio DA,min ⁄ DE as a robustness criteria, with DA,min being the minimal distortion for a successful attack. Chen and Wornell [37] introduced the term “distortion penalty” for DA,min ⁄ DE.
Fig.2-4 depicts a general asymmetric watermarking scheme. With aid of a private and a public key, the watermark is embedded into the host signalx. The significant difference to the symmetric scheme depicted in Fig.2-3 is that all entities, embedding, attack and detection, have access to the public key necessary for watermark detection. Obviously, an attacker can try to use the knowledge of the public key to destroy the embedded watermark information.