Multimedia Security, Volume 1. William Puech

8.2. Incompatible signature detection 8.3. Detection using statistical methods 8.4. Supervised learning detection 8.5. Detection by deep neural networks 8.6. Current avenues of research 8.7. Conclusion 8.8. References

      15  List of Authors

      16  Index

      17  End User License Agreement

      List of Tables

      1 Chapter 1Table 1.1 Description of the main sources of noise during the acquisition proces...

      2 Chapter 3Table 3.1 Galois field GF(24)Table 3.2 Summary table of different contributions in watermarking and correctio...Table 3.3. Table of acronyms in Table 3.2Table 3.4 Table of code correction parameters. For each row of the table, we hav...

      3 Chapter 4Table 4.1 Psychovisual experiments of marked image comparisons

      4 Chapter 7Table 7.1 Properties of data hiding methods, according to their embedding domainTable 7.2 Comparison with previous methods on the 3D Bunny objectTable 7.3 Evaluation of the quality of the approach studied (Itier and Puech 201...

      5 Chapter 8Table 8.1 The different possibilities of good and bad detection. For a color ver...

      List of Illustrations

      1 Chapter 1Figure 1.1 An example showing how an image has been modified several times in a ...Figure 1.2 Simplified processing pipeline of an image, from its acquisition by t...Figure 1.3 The Bayer matrix is by far the most used for sampling colors in camer...Figure 1.4 JPEG compression pipelineFigure 1.5 An example of the impact of quantization on a DCT block. Each DCT coe...Figure 1.6 Calibration model used for the construction of the temporal seriesFigure 1.7 Example of falsification: the vase in b) has been cut out and copied ...Figure 1.8 Percentage of points below the global noise curve and geometric mean ...Figure 1.9 Close-ups on an image before and after compression. The contrast has ...Figure 1.10 Derivative filter and vote map applied to the same image without com...Figure 1.11 Histogram of a DCT coefficient for an image before and after compres...Figure 1.12 In a), an area has been copied four times. The original image is sho...Figure 1.13 The image in a) represents two similar, but different objects, while...Figure 1.14 Example of detection of copy–paste type modification on the images i...Figure 1.15 Structure of the Mayer and Stamm (2019) network to compare the sourc...Figure 1.16 Example of modification detection with the Siamese network (Mayer an...

      2 Chapter 2Figure 2.1 The original image and adversarial images; the manipulations are almo...Figure 2.2 Illustration of the Deep Dreams process applied to the original image...Figure 2.3 Illustration of L-BFGS in 1DFigure 2.4 Illustration of C&W in 1D. The perturbation r is collinear to the gra...Figure 2.5. Illustration, in two dimensions, of adverse attacks on a binary clas...Figure 2.6 Illustration of DeepFool

      3 Chapter 3Figure 3.1 Classical diagram of watermark insertionFigure 3.2 Diagram of the transmission of an image in a channel affected by nois...Figure 3.3 Classical diagram of detection of a markFigure 3.4 Representation of the quantization space (or Euclidean network) in di...Figure 3.5 The different stages allowing the reliable transmission of a message ...Figure 3.6 Strategy by “concatenation codes”Figure 3.7 Decoding in the case of the code by “concatenation codes”Figure 3.8 Illustration of the creation of a mark by code concatenationFigure 3.9 Performance of RS codes against packet errorsFigure 3.10 Concatenation diagram (or hybrid coding) of two correction codes

1(...Figure 3.11 Comparison of the robustness of different codes against the JPEG att...Figure 3.12 Comparison for different images of the Kodack database of robustness...Figure 3.13 Image of a bearFigure 3.14 Image of a plantFigure 3.15 Comparison of the different encodings against an attack by adding Ga...Figure 3.16 Effects of changes to saturation on the Lena image. a) Minimal chang...Figure 3.17 Comparison of coding processes faced with saturation for an image of...Figure 3.18 Insertion strategy using rank metric code and image block decomposit...Figure 3.19 Cropped images with errors from the same rankFigure 3.20 Types of image cropping. The top two rows (a) represent type 1 error...Figure 3.21 Average error rate and rank as a function of the cropping percentage...Figure 3.22 Examples of attacked images with the poorest detection performance. ...

      4 Chapter 4Figure 4.1 Quantization in the RGB color space on an oriented line by a directio...Figure 4.2 Example of inserting a mark with different approaches and direction v...Figure 4.3 Relationship between physical space ϕ and perceptual space ψFigure 4.4 Nonlinear function of perceptionFigure 4.5 MacAdam ellipses in the luminance plane of the color space xyY, 1931....Figure 4.6 The ellipses obtained from the psychovisual model almost correspond t...Figure 4.7 Classical insertion diagram combined with color vector quantization (...Figure 4.8 Pairs of images (host image χ, associated scalar image

). Random ima...Figure 4.9 Classical mark detection diagram. As with insertion, we find the extr...Figure 4.10 Cropped color images (Lena and Kodak base of size 60 × 60) marked wi...Figure 4.11 Cropped color images (Lena and Kodak base of size 60 × 60) marked wi...Figure 4.12 Binary error for methods GA and AA depending on the parameter βFigure 4.13 Bit error rate for (a) hue; (b) saturation; and (c) value modificati...

      5 Chapter 5Figure 5.1 General operation of steganography and roles of different actors (Ali...Figure 5.2 Coding for steganography: the coding system generates several code wo...Figure 5.3 General principle of embedding, encoding and decoding in steganograph...Figure 5.4 Principles of LSB substitution and LSB correspondence, the dotted arr...Figure 5.5 a) Cover Image. b) Associated cost map for the HILL algorithm. Image ...Figure 5.6 Map of associated modification probabilities for the MiPod algorithm ...Figure 5.7 Principle of synchronized embedding using two or four latticesFigure 5.8 Visualization of modifications (− 1, + 1) made to the image without s...Figure 5.9 Principle of weighting by the quantization error: situation (a), wher...Figure 5.10 Natural steganography seeks to mimic the normal distribution of the ...Figure 5.11 The first two iterations of the strategy proposed by Bernard et al. ...Figure 5.12 Principle of steganography based on adversarial generators

      6 Chapter 6Figure 6.1 Sequential watermarking of a movie. A thumbnail image represents a vi...Figure 6.2 Π(p) :=

(Y = 1|p) for c = 3 a) and c = 5 b). All 1 attack: Π(p) = 1...Figure 6.3 The function p → E(

fp)| for

fn ≈ 1/2 for the six collusion strateg...Figure 6.5 Histograms of innocent people’s scores for the “Laarhoven score funct...

      7 Chapter 7Figure 7.1 Classification of data hiding methods, based on the work of Petitcola...Figure 7.2 Data hiding trade-offFigure 7.3 The Stanford Bunny in low-resolution, 1,889 vertices and 3,851 faces,...Figure 7.4 a) 3006 3D point cloud, and b) MST built on the point cloudFigure 7.5 The problem of sensitivity of EMSTsFigure 7.6 Structure at the i stage. The vertices and edges in red are already c...Figure 7.7 State of the sets at step i: vi the current vertexFigure 7.8 a) vi+1 is moved too close to the sub-path, and b) vi+1 is placed too...Figure 7.9 Embedding method of the algorithm of Cayre