**1. Introduction**

Nowadays digital video is widely used to present information. This is because a video is richer in information compared to a single image, and generally a video has more pictures and sound. However, digital data such as video have advantages and disadvantages. Digital video could be edited, manipulated, or altered easily by using a video editor or other tools. For example, someone could change contrast of the video, resize, remove or add some frames, or add a new object to the video. Unfortunately, once a digital video is manipulated, its integrity is questionable. In some cases, we need to know authenticity of the video. For example, a court need to decide if a video as evidence is genuine or has been manipulated. If the video has been manipulated, how to prove it?

The integrity problem of digital video could be solved by using a fragile watermarking technique. In the fragile watermarking technique, we could embed one or more watermarks into frames of video. Once the watermarked video is manipulated, altered, or modified, the watermark inside will be fragile or damage. The damaged watermark is indication that the video has been manipulated. Therefore, fragile watermarking can be used to prove authentication of a video.

A watermarking algorithm consists of two processes: embedding and extraction of watermark. Watermark is an information that refers to the video owner. Usually the watermark is a binary image such as logo, random bits, or other information. The watermark is inserted into a host video, frame by frame, become to a watermarked video without affecting its perceptual (and audio) quality. Through an inverse process, the embedded watermark can be extracted again from the video. When the extracted watermark is compared to the original watermark, we could conclude if the video has been altered. The damage extracted watermark is indication that the video has been altered.

Digital watermarking schemes can operate on spatial domain or frequency domain. Assume the watermark is represented as string of bits. Watermarking schemes that operate in spatial domain embed watermark bits into pixel values of the video frame directly [1, 2]. Otherwise, on watermarking schemes that operate in transform domain, a host video has to be transformed first into a transform domain by using a specific transformation (DCT, DWT, DFT, etc.) [3]. Next embedding bits of watermark is performed by modifying the transform coefficients [2, 4].

Generally watermarking in transform domain is more robust than spatial domain through non-malicious attacks such as cropping, compression, scaling, rotation, etc. Therefore, it is used to the problems of copyright protection, proving ownership, illegal copying, and transaction tracking of video. Otherwise, watermarking in spatial domain is suitable to solve the problem of tamper detection of video content. The watermark in the video must be fragile when the video is manipulated. Robustness is not important for fragile watermarking.

The watermark could be originated from internal or external. The internal watermark means that the watermark is derived from host video directly, and then it is embedded into frames of the video. The second is external watermark which means that the watermark is an input from the user, and usually the watermark is a meaningful binary image such as logo or other image.

Much research on fragile watermarking for digital video has been done by many scientists. This means that the fragile watermarking is interesting research topics. Some related research is from Elgamal et al. [1], Zhi-yu et al. [4], and Rupali et al. [5]. Elgamal et al. [1] proposed a fragile video watermarking algorithm on transform. The original video is transformed from RGB model to YCbCr model and then Cr-component is partitioned into non-overlapping blocks of pixel. The watermark is a binary image from the video owner. Bits of the binary image are embedded for each block separately.

Like Elgamal et al. [1], Zhi-yu et al. [4] also proposed a fragile watermarking algorithm on transform domain. The original video is transformed from RGB model to YST model. The T-component is divided 4 × 4 blocks and then each block is transformed to frequency domain using DCT. The watermark is generated from the quantized DCT coefficient and then it is embedded into the last non-zero DCT coefficient.

Rupali et al. [5] also proposed a fragile video watermarking algorithm on transform domain. The original video is transformed from spatial domain to frequency domain using DCT. Rupali et al. [5] used two watermarks to embed. Both of these watermarks are internal, that means from the original video itself. The first watermark is used to detect tampering and the second watermark is used to localize tampered area.

All of the watermarking algorithms above operated on transform domain. A digital video contains more frames, generally hundreds to thousands frames. Transformation of each frame from the spatial domain to the transform domain consumes a lot of computing time. We need a simple fragile watermarking for the digital video but still meet the security aspect. A simple fragile watermarking on

**39**

value *x*0.

**2. Chaos map**

*Application of Chaos-Based Fragile Watermarking to Authenticate Digital Video*

spatial domain is by using LSB (least significant bit) modification method. This method is fast and it can detect tampering on video until pixel level. To fulfill the security aspect, we use the watermarking key(s) in embedding and extraction process, so that embedding and extraction of watermark are performed by an

The watermark itself is confidential, only the video owner knows about it. Therefore, the watermark needs to be encrypted using the secret key(s). The secret key(s) also serve to prevent the watermark from being extracted and used in the reassembling of videos by an authorized party, thus avoiding counterfeiting of

The chaos system can be used to get a secure fragile watermarking scheme. In recent years, chaos theory has attracted the attention of scientists, especially in the information security field. Chaos has been used to increase security [6]. The reason is the chaotic systems that have sensitivity on initial conditions. It means if we perform a little bit change to the initial conditions of the chaos system, after some iterations, the system will result values that differ significantly. In the field of cryptography and watermarking, generally a chaos map is used to generate pseudo-

Munir et al. [9] used a chaos map, i.e. Arnold Cat's map, to encrypt the watermark before embedding it into video frames. The original watermark is encrypted by XOR-ing it with a random image. The random image is generated from the replicated watermark by using an Arnold Cat Map. The encrypted watermark is

Unfortunately, the random image generated using Arnold Cat's map still shows patterns of the replicated watermark, so it is not completely random. The embedding algorithm is also redundant, because the random image is XOR-ed with the replicated watermark. Therefore, the watermarking algorithm has redundancy.

In this paper, we modified the previous algorithm by using another chaos map so that the random image is generated from the chaos map directly. We used a random bit generator based on two Skew Tent Maps. The generator is abbreviated as CCCBG

The paper is organized into six sections. The first section is this introduction. The second section will review some supported theories. The algorithm to embed and extract the watermark will be explained in the third section. The fourth and fifth section will present the experiment results and discussion. Finally, the last

> *x xx i ii* <sup>+</sup><sup>1</sup> = m

where μ is a parameter of map and 0 < *μ* ≤ 4. According to [7], the map is in chaotic state when 3.57 < μ ≤ 4. In this state, the resulting values appear random. Because of its random behavior, a logistic map can be used as a pseudo-random generator. Hence, initial value of the chaos map, *x*0, and constant μ serve as secret keys. When we iterate Eq. (1) from an initial value (*x*0), we get a random sequences between 0 and 1. The random values generated from the chaos Logistic Map are sensitive to small changes of the initial values. By changing *x*0, the random values generated different significantly from the previous chaotic values with initial

(1 ) (1)

embedded into each frame of the video using LSB modification method.

One of the popular chaotic maps is a Logistic Map, described by

(Cross-Coupled Chaotic random Bit Generator) [10].

section will resume the conclusion and future work.

*DOI: http://dx.doi.org/10.5772/intechopen.93151*

authorized party who has the secret key(s).

the videos.

random numbers [7, 8].

### *Application of Chaos-Based Fragile Watermarking to Authenticate Digital Video DOI: http://dx.doi.org/10.5772/intechopen.93151*

spatial domain is by using LSB (least significant bit) modification method. This method is fast and it can detect tampering on video until pixel level. To fulfill the security aspect, we use the watermarking key(s) in embedding and extraction process, so that embedding and extraction of watermark are performed by an authorized party who has the secret key(s).

The watermark itself is confidential, only the video owner knows about it. Therefore, the watermark needs to be encrypted using the secret key(s). The secret key(s) also serve to prevent the watermark from being extracted and used in the reassembling of videos by an authorized party, thus avoiding counterfeiting of the videos.

The chaos system can be used to get a secure fragile watermarking scheme. In recent years, chaos theory has attracted the attention of scientists, especially in the information security field. Chaos has been used to increase security [6]. The reason is the chaotic systems that have sensitivity on initial conditions. It means if we perform a little bit change to the initial conditions of the chaos system, after some iterations, the system will result values that differ significantly. In the field of cryptography and watermarking, generally a chaos map is used to generate pseudorandom numbers [7, 8].

Munir et al. [9] used a chaos map, i.e. Arnold Cat's map, to encrypt the watermark before embedding it into video frames. The original watermark is encrypted by XOR-ing it with a random image. The random image is generated from the replicated watermark by using an Arnold Cat Map. The encrypted watermark is embedded into each frame of the video using LSB modification method.

Unfortunately, the random image generated using Arnold Cat's map still shows patterns of the replicated watermark, so it is not completely random. The embedding algorithm is also redundant, because the random image is XOR-ed with the replicated watermark. Therefore, the watermarking algorithm has redundancy.

In this paper, we modified the previous algorithm by using another chaos map so that the random image is generated from the chaos map directly. We used a random bit generator based on two Skew Tent Maps. The generator is abbreviated as CCCBG (Cross-Coupled Chaotic random Bit Generator) [10].

The paper is organized into six sections. The first section is this introduction. The second section will review some supported theories. The algorithm to embed and extract the watermark will be explained in the third section. The fourth and fifth section will present the experiment results and discussion. Finally, the last section will resume the conclusion and future work.

### **2. Chaos map**

*Digital Forensic Science*

tion that the video has been altered.

A watermarking algorithm consists of two processes: embedding and extraction of watermark. Watermark is an information that refers to the video owner. Usually the watermark is a binary image such as logo, random bits, or other information. The watermark is inserted into a host video, frame by frame, become to a watermarked video without affecting its perceptual (and audio) quality. Through an inverse process, the embedded watermark can be extracted again from the video. When the extracted watermark is compared to the original watermark, we could conclude if the video has been altered. The damage extracted watermark is indica-

Digital watermarking schemes can operate on spatial domain or frequency domain. Assume the watermark is represented as string of bits. Watermarking schemes that operate in spatial domain embed watermark bits into pixel values of the video frame directly [1, 2]. Otherwise, on watermarking schemes that operate in transform domain, a host video has to be transformed first into a transform domain by using a specific transformation (DCT, DWT, DFT, etc.) [3]. Next embedding bits

of watermark is performed by modifying the transform coefficients [2, 4]. Generally watermarking in transform domain is more robust than spatial domain through non-malicious attacks such as cropping, compression, scaling, rotation, etc. Therefore, it is used to the problems of copyright protection, proving ownership, illegal copying, and transaction tracking of video. Otherwise, watermarking in spatial domain is suitable to solve the problem of tamper detection of video content. The watermark in the video must be fragile when the video is

manipulated. Robustness is not important for fragile watermarking.

meaningful binary image such as logo or other image.

each block separately.

coefficient.

tampered area.

The watermark could be originated from internal or external. The internal watermark means that the watermark is derived from host video directly, and then it is embedded into frames of the video. The second is external watermark which means that the watermark is an input from the user, and usually the watermark is a

Much research on fragile watermarking for digital video has been done by many scientists. This means that the fragile watermarking is interesting research topics. Some related research is from Elgamal et al. [1], Zhi-yu et al. [4], and Rupali et al. [5]. Elgamal et al. [1] proposed a fragile video watermarking algorithm on transform. The original video is transformed from RGB model to YCbCr model and then Cr-component is partitioned into non-overlapping blocks of pixel. The watermark is a binary image from the video owner. Bits of the binary image are embedded for

Like Elgamal et al. [1], Zhi-yu et al. [4] also proposed a fragile watermarking algorithm on transform domain. The original video is transformed from RGB model to YST model. The T-component is divided 4 × 4 blocks and then each block is transformed to frequency domain using DCT. The watermark is generated from the quantized DCT coefficient and then it is embedded into the last non-zero DCT

Rupali et al. [5] also proposed a fragile video watermarking algorithm on transform domain. The original video is transformed from spatial domain to

frequency domain using DCT. Rupali et al. [5] used two watermarks to embed. Both of these watermarks are internal, that means from the original video itself. The first watermark is used to detect tampering and the second watermark is used to localize

All of the watermarking algorithms above operated on transform domain. A digital video contains more frames, generally hundreds to thousands frames. Transformation of each frame from the spatial domain to the transform domain consumes a lot of computing time. We need a simple fragile watermarking for the digital video but still meet the security aspect. A simple fragile watermarking on

**38**

One of the popular chaotic maps is a Logistic Map, described by

$$\boldsymbol{\mathfrak{x}}\_{i\ast\mathbf{z}} = \mu \boldsymbol{\mathfrak{x}}\_{i} \left(\mathbf{z} - \boldsymbol{\mathfrak{x}}\_{i}\right) \tag{1}$$

where μ is a parameter of map and 0 < *μ* ≤ 4. According to [7], the map is in chaotic state when 3.57 < μ ≤ 4. In this state, the resulting values appear random. Because of its random behavior, a logistic map can be used as a pseudo-random generator. Hence, initial value of the chaos map, *x*0, and constant μ serve as secret keys. When we iterate Eq. (1) from an initial value (*x*0), we get a random sequences between 0 and 1. The random values generated from the chaos Logistic Map are sensitive to small changes of the initial values. By changing *x*0, the random values generated different significantly from the previous chaotic values with initial value *x*0.

Another chaos map is Tent Map. It iterates a point *x*0 and gives a sequence *xi* in [0, 1]:

$$\begin{aligned} \left(\boldsymbol{\omega}\_{i\ast 1} = \boldsymbol{f}\_{\boldsymbol{\mu}}\left(\boldsymbol{\omega}\_{i}\right)\right) = \begin{cases} \boldsymbol{\mu}\mathbf{x}\_{i} & \text{, } \boldsymbol{\omega}\_{i} < \frac{\mathbf{1}}{2} \\\\ \boldsymbol{\mu}\left(\mathbf{1} - \boldsymbol{\omega}\_{i}\right) & \text{, } \boldsymbol{\omega}\_{i} \ge \frac{\mathbf{1}}{2} \end{cases} \end{aligned} \tag{2}$$

where μ is a positive real constant. Varian of Tent Map is Skew Tent Map [8] which is defined as:

$$\mathbf{x}\_{i\leftrightarrow 1} = f\_{\mu} \left( \mathbf{x}\_{i} \right) = \begin{cases} \frac{\mathbf{x}\_{i}}{\mu} & , \mathbf{o} \le \mathbf{x}\_{i} < \mu \\\\ \frac{\mathbf{1} - \mathbf{x}\_{i}}{\mathbf{1} - \mu} & , \mu < \mathbf{x} \le \mathbf{1} \end{cases} \tag{3}$$

where *μ* is system parameter and *x*0 is initial condition of map. When a Skew Tent Map is iterated from *x*0 value, it produces a sequence in the interval [0, 1] and distributed uniformly.

Narendra et al. in [10] proposed a random bit generator based on two Skew Tent Maps. The generator is abbreviated as CCCBG (Cross-Coupled Chaotic random Bit Generator). In the CCCBG, random bit stream is generated by comparing outputs of the couple maps. If *fμ*(*xi*) and *gμ*(*yi*) are two Skew Tent Maps and are given as:

$$\mathcal{X}\_{i\ast\mathbf{1}} = f\_{\mu}(\mathcal{X}\_{i}) \tag{4}$$

$$\mathcal{Y}\_{i\ast\ast} = \mathcal{g}\_{\mu}(\mathcal{Y}\_{i}) \tag{5}$$

**41**

**Figure 2.**

*Application of Chaos-Based Fragile Watermarking to Authenticate Digital Video*

In digital watermarking, the random bits play important role to increase security. We will use CCCBG Map to generate random bits. The random bits will be-XOR-ed to the original watermark to yield the encrypted watermark. For example, by iterating (6) 240,000 times (i.e., 480 × 854) and using parameter *μ* = 0.48999, initial conditions *x*0 = 0.500684, and *y*0 = 0.538167586, we get a sequence of random

This section will explain the proposed fragile video watermarking on spatial domain based on the chaotic map. The watermark is a binary image. To detect manipulation in the video frames until pixel level, the watermark must have the same size as the video frame size. Therefore, if watermark size is less than video frame size, the watermark need to be replicated by duplicating it a number of times in order to produce a new watermark that has the same size with the host video frame size. **Figure 2** shows example of replication. The original watermark "ASEAN logo" has a size of 200 × 194 pixels, whereas the video frames have a size of 480 × 854 pixels. This original watermark must be duplicated a number of times so

Next, to increase security, before embedding, the replicated watermark is encrypted by XOR-ing it with a random image. A random image is generated by iterating CCCBG a number of *mn* times where *m* and *n* are frame sizes (**Figure 1**). The replicated watermark is encrypted with the random image by using XOR operation to produce an encrypted watermark (**Figure 3**). Next, we embed the encrypted

We design a simple, but secure, fragile video watermarking based on chaos. The fragile video watermarking algorithm consists of two processes: embedding algo-

There are two scenarios for embedding the watermark into a digital video. The first scenario is embedding each frame of the video with the same watermark. The second scenario is embedding each frame of the video with the different watermark. The second scenario is not practical because the video owner have to provide the watermark of as many frames. Actually, the watermarks can be generated from the video itself (i.e., internal watermarks), we generate the watermark for each video frame that

*DOI: http://dx.doi.org/10.5772/intechopen.93151*

bits figured as a binary bit image (**Figure 1**).

that produce a replicated watermark that has size 480 × 854.

rithm and extraction algorithm, each will be described below.

**3. Proposed algorithm**

watermark into the host video.

**3.1 Watermark embedding algorithm**

*Left: the original watermark; right: the replicated watermark.*

where *μ* is system parameter and is same for both maps. The CCCBG generated a sequence of random bits by comparing the outputs of the maps in the following way:

$$h(\mathbf{x}\_{i\leftrightarrow 1}, \mathbf{y}\_{i\leftrightarrow 1}) = \begin{cases} \mathbf{o} & \text{, } \mathbf{x}\_{i\leftrightarrow 1} < \mathbf{y}\_{i\leftrightarrow 1} \\ \mathbf{1} & \text{, } \mathbf{x}\_{i\leftrightarrow 1} \ge \mathbf{y}\_{i\leftrightarrow 1} \end{cases} \tag{6}$$

Based on several tests performed by Narendra et al., the CCCBG successfully passes all the randomness tests [10], therefore the random binary sequences can be used for encryption.

**Figure 1.** *A random image generated by using the CCCBG.*

*Application of Chaos-Based Fragile Watermarking to Authenticate Digital Video DOI: http://dx.doi.org/10.5772/intechopen.93151*

In digital watermarking, the random bits play important role to increase security. We will use CCCBG Map to generate random bits. The random bits will be-XOR-ed to the original watermark to yield the encrypted watermark. For example, by iterating (6) 240,000 times (i.e., 480 × 854) and using parameter *μ* = 0.48999, initial conditions *x*0 = 0.500684, and *y*0 = 0.538167586, we get a sequence of random bits figured as a binary bit image (**Figure 1**).
