Multimedia Security 1

Today, more than 80% of the data transmitted over networks and archived on our computers, tablets, cell phones or clouds is multimedia data - images, videos, audio, 3D data. The applications of this data range from video games to healthcare, and include computer-aided design, video surveillance and biometrics.

It is becoming increasingly urgent to secure this data, not only during transmission and archiving, but also during its retrieval and use. Indeed, in today’s "all-digital" world, it is becoming ever-easier to copy data, view it unrightfully, steal it or falsify it.

Multimedia Security 1 analyzes the issues of the authentication of multimedia data, code and the embedding of hidden data, both from the point of view of defense and attack. Regarding the embedding of hidden data, it also covers invisibility, color, tracing and 3D data, as well as the detection of hidden messages in an image by steganalysis. WILLIAM PUECH is Professor of Computer Science at Université de Montpellier, France. His research focuses on image processing and multimedia security in particular, from its theories to its applications.

Foreword by Gildas Avoine xi

Foreword by Cédric Richard xiii

Preface xv

illiam PUECH

CHAPTER 1 HOW TO RECONSTRUCT THE HISTORY OF A DIGITAL IMAGE, AND OF ITS ALTERATIONS 1

Quentin BAMMEY, Miguel COLOM, Thibaud EHRET, Marina GARDELLA, Rafael GROMPONE, Jean-Michel MOREL, Tina NIKOUKHAH and Denis PERRAUD

1.1 Introduction 2

1.1.1 General context 2

1.1.2 Criminal background 3

1.1.3 Issues for law enforcement 4

1.1.4 Current methods and tools of law enforcement 5

1.1.5 Outline of this chapter 5

1.2 Describing the image processing chain 8

1.2.1 Raw image acquisition 8

1.2.2 Demosaicing 8

1.2.3 Color correction 10

1.2.4 JPEG compression 11

1.3 Traces left on noise by image manipulation 11

1.3.1 Non-parametric estimation of noise in images 11

1.3.2 Transformation of noise in the processing chain 13

1.3.3 Forgery detection through noise analysis 15

1.4 Demosaicing and its traces 18

1.4.1 Forgery detection through demosaicing analysis 19

1.4.2 Detecting the position of the Bayer matrix 20

1.4.3 Limits of detection demosaicing 23

1.5 JPEG compression, its traces and the detection of its alterations 23

1.5.1 The JPEG compression algorithm 23

1.5.2 Grid detection 25

1.5.3 Detecting the quantization matrix 27

1.5.4 Beyond indicators, making decisions with a statistical model 28

1.6 Internal similarities and manipulations 31

1.7 Direct detection of image manipulation 33

1.8 Conclusion 34

1.9 References 35

CHAPTER 2 DEEP NEURAL NETWORK ATTACKS AND DEFENSE: THE CASE OF IMAGE CLASSIFICATION 41

Hanwei ZHANG, Teddy FURON, Laurent AMSALEG and Yannis AVRITHIS

2.1 Introduction 41

2.1.1 A bit of history and vocabulary 42

2.1.2 Machine learning 44

2.1.3 The classification of images by deep neural networks 46

2.1.4 Deep Dreams 48

2.2 Adversarial images: definition 49

2.3 Attacks: making adversarial images 51

2.3.1 About white box 52

2.3.2 Black or gray box 62

2.4 Defenses 64

2.4.1 Reactive defenses 64

2.4.2 Proactive defenses 66

2.4.3 Obfuscation technique 67

2.4.4 Defenses: conclusion 68

2.5 Conclusion 68

2.6 References 69

CHAPTER 3 CODES AND WATERMARKS 77

Pascal LEFEVRE, Philippe CARRE and Philippe GABORIT

3.1 Introduction 77

3.2 Study framework: robust watermarking 78

3.3 Index modulation 81

3.3.1 LQIM: insertion 81

3.3.2 LQIM: detection 82

3.4 Error-correcting codes approach 82

3.4.1 Generalities 84

3.4.2 Codes by concatenation 86

3.4.3 Hamming codes 88

3.4.4 BCH codes 90

3.4.5 RS codes 93

3.5 Contradictory objectives of watermarking: the impact of codes 96

3.6 Latest developments in the use of correction codes for watermarking 98

3.7 Illustration of the influence of the type of code, according to the attacks 102

3.7.1 JPEG compression 103

3.7.2 Additive Gaussian noise 106

3.7.3 Saturation 106

3.8 Using the rank metric 108

3.8.1 Rank metric correcting codes 109

3.8.2 Code by rank metric: a robust watermarking method for image cropping 113

3.9 Conclusion 121

3.10 References 121

CHAPTER 4 INVISIBILITY 129

Pascal LEFEVRE, Philippe CARRE and David ALLEYSSON

4.1 Introduction 129

4.2 Color watermarking: an approach history? 131

4.2.1 Vector quantization in the RGB space 132

4.2.2 Choosing a color direction 133

4.3 Quaternionic context for watermarking color images 135

4.3.1 Quaternions and color images 135

4.3.2 Quaternionic Fourier transforms 137

4.4 Psychovisual approach to color watermarking 139

4.4.1 Neurogeometry and perception 139

4.4.2 Photoreceptor model and trichromatic vision 141

4.4.3 Model approximation 144

4.4.4 Parameters of the model 145

4.4.5 Application to watermarking color images 146

4.4.6 Conversions 147

4.4.7 Psychovisual algorithm for color images 148

4.4.8 Experimental validation of the psychovisual approach for color watermarking 151

4.5 Conclusion 155

4.6 References 157

CHAPTER 5 STEGANOGRAPHY: EMBEDDING DATA INTO MULTIMEDIA CONTENT 161

Patrick BAS, Remi COGRANNE and Marc CHAUMONT

5.1 Introduction and theoretical foundations 162

5.2 Fundamental principles 163

5.2.1 Maximization of the size of the embedded message 163

5.2.2 Message encoding 165

5.2.3 Detectability minimization 166

5.3 Digital image steganography: basic methods 168

5.3.1 LSB substitution and matching 168

5.3.2 Adaptive embedding methods 169

5.4 Advanced principles in steganography 172

5.4.1 Synchronization of modifications 173

5.4.2 Batch steganography 175

5.4.3 Steganography of color images 177

5.4.4 Use of side information 178

5.4.5 Steganography mimicking a statistical model 180

5.4.6 Adversarial steganography 182

5.5 Conclusion 186

5.6 References 186

CHAPTER 6 TRAITOR TRACING 189

Teddy FURON

6.1 Introduction 189

6.1.1 The contribution of the cryptography community 190

6.1.2 Multimedia content 191

6.1.3 Error probabilities 192

6.1.4 Collusion strategy 192

6.2 The original Tardos code 194

6.2.1 Constructing the code 195

6.2.2 The collusion strategy and its impact on the pirated series 195

6.2.3 Accusation with a simple decoder 197

6.2.4 Study of the Tardos code-Škori´c original 199

6.2.5 Advantages 202

6.2.6 The problems 204

6.3 Tardos and his successors 205

6.3.1 Length of the code 205

6.3.2 Other criteria 205

6.3.3 Extensions 207

6.4 Research of better score functions 208

6.4.1 The optimal score function 208

6.4.2 The theory of the compound communication channel 209

6.4.3 Adaptive score functions 211

6.4.4 Comparison 213

6.5 How to find a better threshold 213

6.6 Conclusion 215

6.7 References 216

CHAPTER 7 3D WATERMARKING 219

Sebastien BEUGNON, Vincent ITIER and William PUECH

7.1 Introduction 220

7.2 Preliminaries 221

7.2.1 Digital watermarking 221

7.2.2 3D objects 222

7.3 Synchronization 224

7.3.1 Traversal scheduling 224

7.3.2 Patch scheduling 224

7.3.3 Scheduling based on graphs 225

7.4 3D data hiding 230

7.4.1 Transformed domains 231

7.4.2 Spatial domain 231

7.4.3 Other domains 232

7.5 Presentation of a high-capacity data hiding method 233

7.5.1 Embedding of the message 234

7.5.2 Causality issue 235

7.6 Improvements 236

7.6.1 Error-correcting codes 236

7.6.2 Statistical arithmetic coding 236

7.6.3 Partitioning and acceleration structures 237

7.7 Experimental results 238

7.8 Trends in high-capacity 3D data hiding 240

7.8.1 Steganalysis 240

7.8.2 Security analysis 241

7.8.3 3D printing 242

7.9 Conclusion 242

7.10 References 243

CHAPTER 8 STEGANALYSIS: DETECTION OF HIDDEN DATA IN MULTIMEDIA CONTENT 247

Remi COGRANNE, Marc CHAUMONT and Patrick BAS

8.1 Introduction, challenges and constraints 247

8.1.1 The different aims of steganalysis 248

8.1.2 Different methods to carry out steganalysis 249

8.2 Incompatible signature detection 250

8.3 Detection using statistical methods 252

8.3.1 Statistical test of χ2 252

8.3.2 Likelihood-ratio test 256

8.3.3 LSB match detection 261

8.4 Supervised learning detection 263

8.4.1 Extraction of characteristics in the spatial domain 264

8.4.2 Learning how to detect with features 269

8.5 Detection by deep neural networks 270

8.5.1 Foundation of a deep neural network 271

8.5.2 The preprocessing module 272

8.6 Current avenues of research 279

8.6.1 The problem of Cover-Source mismatch 279

8.6.2 The problem with steganalysis in real life 279

8.6.3 Reliable steganalysis 280

8.6.4 Steganalysis of color images 280

8.6.5 Taking into account the adaptivity of steganography 281

8.6.6 Grouped steganalysis (batch steganalysis) 281

8.6.7 Universal steganalysis 282

8.7 Conclusion 283

8.8 References 283

List of Authors 289

Index 293