1 To make this article self-contained, this section recalls
2 the basis of the Syndrome Treillis Codes (STC).
4 $x=(x_1,\ldots,x_n)$ be the $n$-bits cover vector of the image $X$,
5 $m$ be the message to embed, and
6 $y=(y_1,\ldots,y_n)$ be the $n$-bits stego vector.
7 The usual additive embedding impact of replacing $x$ by $y$ in $X$
8 is given by a distortion function
9 $D_X(x,y)= \Sigma_{i=1}^n \rho_X(i,x,y)$, where the function $\rho_X$
10 expresses the cost of replacing $x_i$ by $y_i$ in $X$.
11 Let us consider that $x$ is fixed:
12 this is for instance the LSBs of the image edge bits.
13 The objective is thus to find $y$ that minimizes $D_X(x,y)$.
15 Hamming embedding proposes a solution to this problem.
16 This is why some steganographic
17 schemes~\cite{DBLP:conf/ih/Westfeld01,DBLP:conf/ih/KimDR06,DBLP:conf/mmsec/FridrichPK07} are based on this binary embedding.
18 Furthermore this code provides a vector $y$ s.t. $Hy$ is equal to
19 $m$ for a given binary matrix $H$.
21 Let us explain this embedding on a small illustrative example where
22 $\rho_X(i,x,y)$ is equal to 1,
23 whereas $m$ and $x$ are respectively a 3 bits column
24 vector and a 7 bits column vector.
25 Let then $H$ be the binary Hamming matrix
28 \begin{array}{lllllll}
29 0 & 0 & 0 & 1 & 1 & 1 & 1 \\
30 0 & 1 & 1 & 0 & 0 & 1 & 1 \\
31 1 & 0 & 1 & 0 & 1 & 0 & 1
35 The objective is to modify $x$ to get $y$ s.t. $m = Hy$.
36 In this algebra, the sum and the product respectively correspond to
37 the exclusive \emph{or} and to the \emph{and} Boolean operators.
38 If $Hx$ is already equal to $m$, nothing has to be changed and $x$ can be sent.
39 Otherwise we consider the difference $\delta = d(m,Hx)$, which is expressed
42 \delta = \left( \begin{array}{l}
48 \textrm{ where $\delta_i$ is 0 if $m_i = Hx_i$ and 1 otherwise.}
50 Let us thus consider the $j-$th column of $H$, which is equal to $\delta$.
51 We denote by $\overline{x}^j$ the vector obtained by
52 switching the $j-$th component of $x$,
53 that is, $\overline{x}^j = (x_1 , \ldots, \overline{x_j},\ldots, x_n )$.
54 It is not hard to see that if $y$ is $\overline{x}^j$, then
56 It is then possible to embed 3 bits in only 7 LSBs of pixels by modifying
58 In the general case, communicating $n$ message bits in
59 $2^n-1$ pixels needs $1-1/2^n$ average changes.
63 Unfortunately, for any given $H$, finding $y$ that solves $Hy=m$ and
64 that minimizes $D_X(x,y)$, has an exponential complexity with respect to $n$.
65 The Syndrome-Trellis Codes
66 presented by Filler \emph{et al.} in~\cite{DBLP:conf/mediaforensics/FillerJF10}
67 is a practical solution to this complexity. Thanks to this contribution,
68 the solving algorithm has a linear complexity with respect to $n$.
70 First of all, Filler \emph{et al.} compute the matrix $H$
71 by placing a small sub-matrix $\hat{H}$ of size $h × w$ next
72 to each other and by shifting down by one row.
73 Thanks to this special form of $H$, one can represent
74 every solution of $m=Hy$ as a path through a trellis.
76 Next, the process of finding $y$ consists in two stages: a forward and a backward part.
78 \item Forward construction of the trellis that depends on $\hat{H}$, on $x$, on $m$, and on $\rho$.
79 \item Backward determination of $y$ that minimizes $D$, starting with
80 the complete path having the minimal weight.