Digital Watermarking Techniques, Applications and Attacks Applied to Digital Media: A Survey

Digital Watermarking Techniques, Applications and Attacks Applied to Digital Media: A Survey

Ensaf Hussein

Senior Teaching Assistant, Computer Science Dept., Faculty of Computers & Information, Helwan Univ., Cairo, Egypt.

Mohamed A. Belal

Professor, Computer Science Dept.,

Faculty of Computers & Information, Helwan Univ., Cairo, Egypt.

Abstract

Digital watermarking techniques have been developed to protect the copyright of digital media. This paper aims to provide a detailed review and background about the watermarking definition, concept and the main contributions in this field. It begins with digital watermarking overview, general framework, attacks, application, and finally a comprehensive survey of existing and most recent watermarking techniques. We classify the techniques according various categories such as host signal, perceptivity, robustness, watermark type, necessary data for extraction, processing domain, and applications. In the survey our main concern is image only.

1. Introduction

   Nowadays, the rapid development of technologies has led to the significant increase of digital information, particularly multimedia such as image, audio and video content. Such technological advances have led to the ease in which it is possible to illegally share, distribute, and copy Intellectual Property (IP) [1-3].

   An obvious requirement, therefore, is the development of solutions for copyright protection and ownership identification for digital content.

   Digital watermarking is the process of embedding relevant ownership information (such as a logo, fingerprint and serial number), into a media in order to protect the ownership of different media formats [2-5]. This technique can be applied to different media types such as video, audio and image content. For the purpose of copyright protection and ownership identification, robust watermarking schemes are mainly used as they can tolerate a host of signal processing attacks that can be both unintentional and intentional.

   The paper begins with the fundamental concept, general framework, attacks and application of digital watermarking; follow by a more detailed survey of different transform based watermarking techniques.

   The paper is organized as follows:

   • In Section 2, we briefly describe the background of digital watermarking, then we discuss the concept of digital image watermarking, the effectiveness requirements, the embedding and extraction processes, and finally present the general framework.

   • Section 3 presents classification and analysis of digital watermarking techniques according different criteria such as host signal, perceptivity, robustness, watermark type, necessary data for extraction, processing domain, and applications.

   • Section 4 presents a survey of various robust image watermarking algorithms operating in the transform domain. We review algorithms applied in the spatial domain, and then extensively review transform-based robust watermarking methods.

   • Section 5 explores watermarking attacks.

   • Section 6 gives the concluding remarks.

2. Digital Watermarking Overview

   1. Watermarking Background

      Historically [1, 7], postage stamps and currency were commonly watermarked. Indeed, the currency watermark is still used today when printing banknotes. A digital watermark can be either visible or invisible. An example of digital visible watermark is the translucent logos that are often seen embedded at the corner of videos or images, in an attempt to prevent copyright infringement. However, these visible watermarks can be targeted and removed rather simply by cropping the media,

      or overwriting the logos. Subsequently, the field of digital watermarking is primarily focused on embedding invisible watermarks, which operate by tweaking the content of the media imperceptibly. As the watermark cannot be seen, there must exist a robustness property that ensures the watermark data survives if the image is altered. Typical applications of digital watermarking can include broadcast monitoring, owner identification, proof of ownership, transaction tracking, content authentication, copy control, device control legacy enhancement and content description

      Watermarking is the process of embedding a special data into media such as image, audio and video. This embedded information, known as a watermark, can be extracted from the multimedia contents later and used for supporting the ownership.

   2. General Watermarking Framework

      Mathematically [6], watermark embedding can be expressed like Eq.( 1). If an original image and a watermark are given, the watermarked image ' is represented as the following Eq(1).

      Figure 1 shows watermark embedding process. An original image , a watermark enter to the system, is a scaling factor and an optional public or secret key K may be used to guide the process. The output of the watermarking scheme is the watermarked image ' [5].

      Watermark

      Embedding

      '

      Original Image

      K

      Watermarked Image

      w

      Watermark

      Key

      Figure1: Watermark embedding

      The embedded watermark can be extracted later by two ways or more, either the original image is used to compare and find out the watermark (non-blind watermark) or a correlation measure is used to detect the watermark (blind watermarking).

      In non-blind watermarking, the extraction of the embedded watermark can simply achieved by applying Eq. 2 , in which simply subtract the original signal from the watermarked one then divide it by the gain factor.

      In blind watermarking, the extraction of the watermark can be achieved by using similarity measures as shown in Figure 2.

      '

      Watermarked Image

      Watermark Extraction

      'w

      Similarity Measure

      Watermark Or Confidence measure

      Original Image

      Key

      Embedded Watermark

      Figure2: Watermark extraction and detection.

      There are some ways which can evaluate the similarity between the original and extracted watermarks. However, mostly used similarity measures are the correlation-based method. The following Eq.(3) is widely used as a watermark similarity measure. Generally, the extracted watermark will not be identical to the original

      watermark w. Eq. (3) computes the similarity between and as the follows;

      To decide whether and match, one may determine, , ) >T, where T is some threshold [5, 6].

   3. Requirements of Watermarking

      There are three main requirements of digital watermarking. They are transparency, robustness, and capacity [8].

      Capacity

      Trade-off

      The host or cover signal could be:

      • Text watermarking: it inserts a watermark in the font shape and the space between characters and line spaces.

        • Image watermarking: this embeds special information to an image and detects or extracts it later for ownership confirmation.

        • Video watermarking: it is an extension of image watermarking. This method requires real time extraction and robustness for compression.

        • Audio watermarking: this application area becomes a hot issue because of the internet music, MP3.

          3.2 According to perceptivity

          Digital watermarking is divided into two main

          Imperceptibility

          Robustness

          categories: visible and invisible.

          Figure 3: Trade-off among the imperceptibility, robustness and capacity

        Transparency

        Transparency or fidelity is perceptual similarity between the original and the watermarked versions of the cover work.

        The digital watermark should not affect the quality of the original image after it is watermarked.

        • Robustness

          Rrobustness is the ability to detect the watermark after common signal processing operations.

          Watermarks should be robust against variety of geometrical and non-geometrical attacks.

        • Capacity

          Capacity or data payload is the number of bits a watermark encodes within a unit of time or work. This property describes how much data should be embedded as a watermark to successfully detect during extraction.

3. Watermarking Classification

   In this section digital watermarks and their techniques is classified and segmented into various categories; Figure 4 shows Classification of watermarking techniques, for example, they can be classified according to [1-7]:

   1. According host signal

      • Visible watermark: it is equivalent to stamping a watermark on paper, (ex.) television channels, like BBC, whose logo is visibly superimposed on the corner of the TV picture.

      • Invisible watermarking: It is used to identify copyright data, like author, distributor, and so forth but it's more complex.

      1. According to robustness

         Watermarks need robustness to protect the ownership from various attacks. The followings show the classification dependent on the robustness of a watermark.

         • Robust: resist various attacks, geometrical or non-geometrical without affecting embedded watermark.

         • Semi-fragile: is capable of tolerating some degree of the change to a watermarked image, such as the addition of quantization noise from lossy compression.

         • Fragile: fragile watermark is designed to be easily destroyed if a watermarked image is manipulated in the slightest manner. This watermarking method be used for the protection and the verification of original contents.

      2. According to watermark type

         Watermark types can be classified into two types: noise type and image type.

         • Noise types have pseudo noise, Gaussian random and chaotic sequences.

         • Image types, there are binary image, stamp, logo and label.

         Semi-fragile

         Semi-blind

         Invisible

         Image

         Image Format

         Fragile

         Non-blind

         Text

         Noise

         Robust

         Visible

         Watermarking Classification

         Host Signal

         Perceptivity

         Robustness

         Watermark type

         Necessary Data

         for Extraction

         Audio

         Blind

         Video

         Domain

         Application

         Spatial

         Frequency

         Feature

         Fingerprinting

         LSB

         Cosine Transform DCT

         Spatial Domain

         SVD

         Broadcast Monitoring

         Contourlet Transform CT

         Wavelet Transform DWT

         Spread Spectrum

         Transform Domain

         Copy protection

         Data Authentication

         Copyright Protection

         Figure 4: Classification of watermarking techniques

      3. According to necessary data for extraction

         In order to detect the watermark information, blind, semi-blind and non-blind techniques are used.

         • Non-blind: this requires at least an original media. It extracts a watermark from the possibly distorted image and the original media.

         • Semi-blind: it does not require an original media for detection.

         • Blind: it requires neither an original media nor the embedded watermark. It is also referred to as public watermarking.

      4. According to Processing Domain

         Image watermarking can be applied to spatial domain and transform domain [1-3].

         In spatial domain the watermark is inserted into the intensity values. The best widely known algorithm is LSB techniques; it is based on modifying the least significant bit layer of images. This technique based on the fact that the changes in least significant bits in an image would not have any effect on an image.

         Another approach for embedding in spatial domain is correlation based techniques. The noise that statistically resembles common processing distortion is introduced to pixels in random walk. The noise is produced by a pseudo random generator using a shared key.

         In Transform Domain image is represented in terms of frequencies , to transfer image to its frequency representation we can use several reversible transform like Discrete Cosine Transform DCT, Discrete Wavelet Transform DWT, Discrete Fourier Transform DFT, Contourlet Transform CT, or Singular Value Decomposition SVD. Each of the transforms has its own characteristics and represents the image indifferent ways.

         Table 1: Comparison between watermarking techniques

         Characteristics	Spatial Domain	Transform Domain
         Computation Cost	Low	High
         Robustness	Fragile	Robust
         Perceptual quality	High Control	Low control

         Capacity	High(depends on image size)	Low
         Application	Authentication	Copyright

         Year	Author Name	Processing Domain	Features
         2004	Fangjun Huang et al. [9]	DCT SVD	– achieve the highest possible robustness without losing the transparency.
         2006	Enping Li et al.[10]	DWT	– robust to both common image processing operations and geometric attacks such as rotation, scaling, cropping.
         2006	Jayalakshmi M. et al. [11]	CLT	– contourlet based algorithm outperforms wavelet and DCT based methods
         2007	R. A. Ghazy et al.[12]	SVD	– Robust
         2008	Shu Zhibiao et al.[13]	DWT
         2008	Hanane Mirza et al.[14]	DWT
         2008	B.Chandra Mohan et al. [15]	CLT	– Robust
         2008	Azadeh Mansouri et al. [16]	DCT SVD
         2009	Tang Wenliang et al. [17]	Features based
         2009	Shereen Ghannam et al. [18]	CLT	– robust
         2010	Zaho et al.[7]	CLT	invisibility
         2010	Akhaee et al. [19]	CLT
         2010	Swanirbhar Majumder et al. [20]	SVD Neural Network	– Increase the robustness against malicious attacks.
         2010	Hongbo BI et al. [21]	SVD CT	– very robust against scaling, JPEG compression, cropping
         2011	R. H. Laskar et al.[22]	DCT DWT

         • Blind

         • Robust

         • invisible

         • secure

         • High robustness

         • Reassuring rate of image recovery.

         • robust against large amount of attacks

         • the best quality for watermarked image

         • robust

         • resist geometrical attacks

         • Robust

         • Excellent perceptual

         • Robust

         • good transparency

         • Good imperceptibility

         • Higher robustness

         Table 2: Image Watermarking Survey

      5. According to Applications

         Digital watermarking can be applied to [1]:

         • Copyright

           Protection and Authentication

         • Fingerprinting and Digital "Signatures"

         • Copy Protection and Device Control

         • Broadcast Monitoring

         • Data Authentication

4. Image Watermarking Survey

   In this paper we would focus on robust image watermarking algorithms in transform domain like DCT, DWT, CT and SVD.

   Table 2 presents briefly the most recent algorithms presented in robust image watermarking.

   The classic and still most popular domain for image processing is that of the Discrete-Cosine- Transform, or DCT. Although changing the DCT coefficients will cause unnoticeable visual artifices, they do cause detectable statistical changes.

   Another transform domain which is being exploited is Singular Value Decomposition (SVD) due to its simplicity in implementation and attractive mathematical features. It is one of the most powerful numerical analysis technique and used in various applications. Some SVD based algorithms are purely SVD based in a sense that only SVD domain is used to embed watermark into image.

   Another domain exploited for embedding the watermark is the wavelet domain. The DWT (Discrete Wavelet Transform) separates an image into a lower resolution approximation image (LL) as well as horizontal (HL), vertical (LH) and diagonal (HH) detail components. The process can then be repeated to computes multiple scale wavelet decomposition.

   One of the many advantages over the wavelet transform is that that it is believed to more accurately model aspects of the HVS as compared to the FFT or DCT. This allows us to use higher energy watermarks in regions that the HVS is known to be less sensitive to, such as the high resolution detail bands {LH, HL, HH). Embedding watermarks in these regions allow us to increase the robustness of our watermark, at little to no additional impact on image quality.

   2011	Jianhua Song et al.[23]	DWT
   2011	LI Xiuguang et al.[24]	SVD	scaling attacks
   2011	Mingli Zhang et al. [25]	SVD DWT	– robust to a wide range of attacks, especially geometrical attacks.
   2012	Chen Li et al.[26]	DWT	– Conclude that biorthogonal wavelet better than others
   2012	Subramanyam et al.[27]	DWT	– A robust watermarking algorithm to watermark JPEG2000 compressed and encrypted images.
   2012	Wang, Chuntao et al. [28]	DWT GA
   2012	Xiong Shunqing et al. [29]	NSCT SVD

   • Invisible

   • Robust

   • Bears very good security.

   • Blind

   • Robust

   • Resist rotation and

   • High robustness

   • Greatly reduced arithmetic complexity.

   • Good impercbtbilty

   • Good robustness

5. Watermarking Arracks

   Watermark attacks is classified into four distinct categories namely removal attacks, geometric attacks, cryptographic attacks and Removal attacks [8]

   • Removal attacks This type of watermark attack does not attempt to find out the encryption techniques used or how the watermark has been embedded. Included in this category noising, histogram equalization, blur and sharpen attacks.

     Geometry attacks this type of attack intends to distort the watermark signal. It is however still theoretically possible for the detector to recover the original watermark if the detail of the geometry attack can be established and a countermeasure applied. The process of correcting this type of attack is often referred to as synchronization. However, the complexity of the required synchronization process might be too prohibitively expensive and slow. Included in this category of watermark attacks are image rotation, scaling, translation and skewing.

   • The aim of cryptographic attacks is to crack the security methods in watermarking schemes and thus find a way to remove the embedded watermark information or to embed misleading watermarks. One of the techniques in this category is the brute-force search method.

   • Another technique is called the Oracle attack, which is used to create a non-watermarked signal when a watermark detector device is available.

   • Protocol attacks it add the attackers own watermark signals onto the data in question. This results in ambiguities on the true owners question. Protocol attacks target the entire concept of using watermarking techniques as a solution to copyright protection. Another protocol attack is the copy attack: instead of destroying the watermark, the copy attack estimates a watermark from watermarked data and copies.

   The previous classification of watermark attacks can be summarized in Figure 5. [8]

   Watermark Attacks

   Removal

   Cryptographic

   Geometry

   Protocol

   Sharpen

   Brute Force

   Translation

   Invertible

   Blur

   Oracle

   Rotation

   Copy Attacks

   Median Filter

   Scaling

   Noise

   Figure 5 Classification of watermark attacks

6. Conclusion

   In this paper we presented digital watermarking overview, application, attacks and techniques. We tried to classify and analyse many recent watermarking techniques to help new researchers in related areas.

   We classified the previous researches from various point of views such as: host signal, perceptivity, robustness, watermark type, necessary data for extraction, processing domain, and applications.

   Due to space limitation we presented image watermarking survey only, but audio and video techniques are also required to study intensively.

7. References