Digital Audio Watermarking Techniques and Technologies

Digital Audio Watermarking Techniques and Technologies: Appl icat ions and Benchmarks

Nedeljko Cvejic University of Bristol, UK

Tapio Seppänen University of Oulu, Finland

Information Science

REFERENCE INFORMATION SCIENCE REFERENCE

Hershey • New York

Detailed Table of Contents

Preface x

Acknowledgment xv

Chapter I Introduction to Digital Audio Watermarking I Nedeljko Cvejic and Tapio Seppänen 1

This chapter gives a general overview of the audio watermarking fundamental definitions. Audio wa­termarking algorithms are characterized by five essential properties, namely: perceptual transparency, watermark bit rate, robustness, blind/informed watermark detection, and security. The chapter also reviews the most common signal processing manipulations that are frequently applied to the watermarked audio in order to prevent detection of the embedded watermark. Finally, several application areas for digital audio watermarking are presented and advantages of digital watermarking over standard technologies are examined.

Chapter II Spread Spectrum for Digital Audio Watermarking I Xing He and Michael Scordilis 11

This chapter is a comprehensive presentation of spread spectrum-based digital audio watermarking methods. The problem is viewed as the realization of a basic communications system, where the host signal presents the available channel and the watermark presents the transmitted information that needs to survive and be recovered in conditions that include noise distortion, signal transformation, standard compression, and deliberate attacks. Basic spread spectrum theory as it relates to audio watermarking is introduced followed by state-of-the-art improvements. The important synchronization problem is analyzed in detail, existing techniques are presented, and a novel, precise synchronization method is included. Finally, the role of psychoacoustics in effective watermarking is emphasized and an enhanced psychoacoustic model based on the discrete wavelet packet transform (DWPT), which ensures efficiency and transparency is included.

Chapter III Audio Watermarking Through Parametric Synthesis Models / Yi-Wen Liu 50

This chapter promotes the use of parametric synthesis models in digital audio watermarking. It argues that, because human auditory perception is not a linear process, the optimal hiding of binary data in digital audio signals should consider parametric transforms that are generally nonlinear. To support this argument, an audio watermarking algorithm based on aligning frequencies of spectral peaks to grid points is presented as a case study; its robustness is evaluated and benefits are discussed. Toward the end, research directions are suggested, including watermark-aided sound source segregation, cocktail watermarking, and counter-measure against arithmetic collusive attacks.

Chapter IV Robust Zero-Bit and Multi-Bit Audio Watermarking Using Correlation Detection and Chaotic / Nikos Nikolaidis and Alexia Giannoula 82

Digital rights management of audio signals through robust watermarking has received significant atten­tion during the last years. Two approaches for blind robust watermarking of audio signals are presented in this chapter. Both approaches use chaotic maps for the generation of the watermark signals and a correlation-based watermark detection procedure. The first method is a zero-bit method that embeds high-frequency chaotic watermarks in the low frequencies of the discrete Fourier transform (DFT) domain, thus achieving reliable detection and robustness to a number of attacks. The second method operates on the temporal domain and robustly embeds multiple bits of information in the host signal. Experimental results are provided to demonstrate the performance of the two methods.

Chapter V Three Techniques of Digital Audio Watermarking / Say Wei Foo 104

Based on the requirement of watermark recovery, watermarking techniques may be classified under one of three schemes: nonblind watermarking scheme, blind watermarking schemes with and without synchronization information. For the nonblind watermarking scheme, the original signal is required for extracting the watermark and hence only the owner of the original signal will be able to perform the task. For the blind watermarking schemes, the embedded watermark can be extracted even if the original signal is not readily available. Thus, the owner does not have to keep a copy of the original signal. In this chapter, three audio watermarking techniques are described to illustrate the three different schemes. The time-frequency technique belongs to the nonblind watermarking scheme; the multiple-echo hiding technique and the peak-point extraction technique fall under the blind watermarking schemes with and without synchronization information respectively.

Chapter VI Advanced Audio Watermarking Based on Echo Hiding: Time-Spread Echo Hiding / Ryouichi Nishimura, Yöiti Suzuki, and Byeong-Seob Ко 123

This chapter introduces time-spread echo hiding as an advanced audio watermarking method based on echo hiding. After reviewing some relevant researches, theoretical derivation of echo hiding and

time-spread echo hiding is presented. As for embedding process, differences in the structure of echo kernels between the ordinary echo hiding and the time-spread echo hiding are schematically depicted to explain advantages of the new method. Several watermark-extracting methods are introduced for the decoding process to raise the detection rate for various conditions. Performance of the method in ro­bustness against several attacks is evaluated in terms of d' because of its statistical preciseness. Results of computer simulations tell that the time-spread echo hiding show fairly better performance than the ordinary echo hiding. —^

Chapter VII Analysis-by-Synthesis Echo Watermarking/ Wen-Chih Wu and О seal Chen 152

In this chapter, detailed explanations would be given on the role of echo hiding playing in audio wa­termarking in terms of background, functions, and applications. Additionally, a section is dedicated to discuss the various approaches proposed in the past to solve the flaws of echo hiding. Lastly, the proposed analysis-by-synthesis echo watermarking scheme based on interlaced kernels is introduced. Comparisons in audio quality and robustness performance are also looked at the proposed and conven­tional echo watermarking schemes.

Chapter VIII Robustness Analysis of Patchwork Watermarking Schemes / Hyoung-Joong Kim, Shijun Xiang, In-Kwon Yeo, and Subhamoy Maitra 172

The patchwork watermarking scheme is investigated in this chapter. The performance of this algorithm in terms of imperceptibility, robustness, and security has been shown to be satisfactory. Robustness of the patchwork algorithm to the curve-fitting attack and blind multiple-embedding attack is presented also in this chapter. Robustness against jitter attack, which is a natural enemy of this watermarking algorithm, is also studied.

Chapter IX Time-Frequency Analysis of Digital Audio Watermarking / Sridhar Krishnan, Behnez Ghoraani, and Serhat Erkucuk 187

In this chapter, we present an overview of our time frequency (TF) based audio watermarking methods. First, a motivation on the necessity of data authentication, and an introduction in digital rights manage­ment (DRM) to protect digital multimedia contents is presented. TF techniques provide flexible means to analyze nonstationary audio signals. We have explained the joint TF domain for watermark repre­sentation, and have employed pattern recognition schemes for watermark detection. In this chapter, we introduce two watermarking methods; embedding nonlinear and linear TF signatures as watermarking signatures. Robustness of the proposed methods against common signal manipulations is also studied in this chapter.

Chapter X Digital Audio Watermarking Techniques for MP3 Audio Files / Dimitrios Koukopoulos and Yannis C. Stamatiou 205

In this chapter, we will present a brief overview of audio watermarking with the focus on a novel audio watermarking scheme which is based on watermarks with a 'semantic' meaning and offers a method for MPEG Audio Layer 3 files which does not need the original watermark for proving copyright ownership of an audio file. This scheme operates directly in the compressed data domain, while manipulating the time and subband/channel domains and was implemented with the mind on efficiency, both in time and space. The main feature of this scheme is that it offers users the enhanced capability of proving their ownership of the audio file not by simply detecting the bit pattern that comprises the watermark itself, but by showing that the legal owner knows a hard to compute property of the watermark bit sequence ('semantic' meaning). Our discussion on the scheme is accompanied by experimental results.

Chapter XI Digital Watermarking of Speech Signals I Aparna Gurijala 229

The main objective of the chapter is to provide an overview of existing speech watermarking technol­ogy and to demonstrate the importance of speech processing concepts for the design and evaluation of watermarking algorithms. This chapter describes the factors to be considered while designing speech watermarking algorithms, including the choice of the domain and speech features for watermarking, watermarked signal fidelity, watermark robustness, data payload, security, and watermarking applica­tions. The chapter presents several state-of-the-art robust and fragile speech watermarking algorithms and discusses their advantages and disadvantages.

Chapter XII Robustness Against DA/AD Conversion: Concepts, Challenges, and Examples / Martin Steinebach 248

This chapter discusses the robustness of digital audio watermarking algorithms against digital-to-ana­logue (D/A) and analogue-to-digital (A/D) conversions. This is an important challenge in many audio watermarking applications. We provide an overview on distortions caused by converting the signal in various scenarios. This includes environmental influences like background noise or room acoustics when taking microphone recordings, as well as sound sampling effects like quantisation noise. The aim is to show the complexity of influences that need to be taken into account. Additionally, we show test results of our own audio watermarking algorithm with respect to analogue recordings using a microphone which proves that a high robustness in this area can be achieved. To improve even the robustness of algorithms, we briefly introduce strategies against downmixing and playback speed changes of the audio signal.

Chapter XIII Subjective and Objective Quality Evaluation of Watermarked Audio / Michael Arnold, Peter Baum, and Walter Voeßing 260

Methods for evaluating the quality of watermarked objects are detailed in this chapter. It will provide an overview of subjective and objective methods usable in order to judge the influence of watermark embedding on the quality of audio tracks. The problem associated with the quality evaluation of water­marked audio data will be presented. This is followed by a presentation of subjective evaluation standards used in testing the transparency of marked audio tracks as well as the evaluation of marked items with intermediate quality. Since subjective listening tests are expensive and dependent on many not easily controllable parameters, objective quality measurement methods are discussed in the section, Objective Evaluation Standards. The section Implementation of a Quality Evaluation presents the whole process of testing the quality taking into account the methods discussed in this chapter. Special emphasis is devoted to a detailed description of the test setup, item selection, and the practical limitations. The last section summarizes the chapter-Chapter XIV Watermarking Security / Teddy Furon, Francois Cayre, and Caroline Fontaine 278

Digital watermarking studies have always been driven by the improvement of robustness. Most of articles of this field deal with this criterion, presenting more and more impressive experimental assess­ments. Some key events in this quest are the use of spread spectrum, the invention of resynchronisation schemes, the discovery of side information channel, and the formulation of the embedding and attacking strategies as a game. On the contrary, security received little attention in the watermarking community. This chapter presents a comprehensive overview of this recent concept. We list the typical applications which require a secure watermarking technique. For each context, a threat analysis is purposed. This presentation allows us to illustrate all the certainties the community has on the subject, browsing all key papers. The end of the chapter is devoted to what remains not clear, intuitions, and future trends.

Compilation of References 300

About the Contributors 318

Index 324