1 The flowcharts given in Fig.~\ref{fig:sch}
2 summarize our steganography scheme denoted by
3 STABYLO, which stands for STeganography with cAnny, Bbs, binarY embedding at LOw cost.
4 What follows are successively details of the inner steps and flows inside
5 both the embedding stage (Fig.~\ref{fig:sch:emb})
6 and the extraction one (Fig.~\ref{fig:sch:ext}).
7 Let us first focus on the data embedding.
11 \subfloat[Data Embedding.]{
12 \begin{minipage}{0.49\textwidth}
14 %\includegraphics[width=5cm]{emb.pdf}
15 \includegraphics[scale=0.5]{emb.ps}
20 \subfloat[Data Extraction.]{
21 \begin{minipage}{0.49\textwidth}
23 %\includegraphics[width=5cm]{rec.pdf}
24 \includegraphics[scale=0.5]{rec.ps}
30 \caption{The STABYLO Scheme.}
41 \subsection{Security Considerations}
42 Among methods of message encryption/decryption
43 (see~\cite{DBLP:journals/ejisec/FontaineG07} for a survey)
44 we implement the Blum-Goldwasser cryptosystem~\cite{Blum:1985:EPP:19478.19501}
45 that is based on the Blum Blum Shub~\cite{DBLP:conf/crypto/ShubBB82}
46 pseudorandom number generator (PRNG) and the
48 It has been indeed proven~\cite{DBLP:conf/crypto/ShubBB82} that this PRNG
49 has the property of cryptographical security, \textit{i.e.},
50 for any sequence of $L$ output bits $x_i$, $x_{i+1}$, \ldots, $x_{i+L-1}$,
51 there is no algorithm, whose time complexity is polynomial in $L$, and
52 which allows to find $x_{i-1}$ and $x_{i+L}$ with a probability greater
54 Equivalent formulations of such a property can
55 be found. They all lead to the fact that,
56 even if the encrypted message is extracted,
57 it is impossible to retrieve the original one in
60 Starting thus with a key $k$ and the message \textit{mess} to hide,
61 this step computes a message $m$, which is the encrypted version of \textit{mess}.
64 \subsection{Edge-Based Image Steganography}
69 already presented \cite{Luo:2010:EAI:1824719.1824720,DBLP:journals/eswa/ChenCL10} differ
70 in how carefully they select edge pixels, and
73 %Image Quality: Edge Image Steganography
74 %\JFC{Raphael, les fuzzy edge detection sont souvent utilisés.
75 % il faudrait comparer les approches en terme de nombre de bits retournés,
76 % en terme de complexité. } \cite{KF11}
77 %\RC{Ben, à voir car on peut choisir le nombre de pixel avec Canny. Supposons que les fuzzy edge soient retourne un peu plus de points, on sera probablement plus détectable... Finalement on devrait surement vendre notre truc en : on a choisi cet algo car il est performant en vitesse/qualité. Mais on peut aussi en utilisé d'autres :-)}
79 Many techniques have been proposed in the literature to detect
80 edges in images (whose noise has been initially reduced).
81 They can be separated in two categories: first and second order detection
82 methods on the one hand, and fuzzy detectors on the other hand~\cite{KF11}.
83 In first order methods like Sobel, Canny~\cite{Canny:1986:CAE:11274.11275}, \ldots,
84 a first-order derivative (gradient magnitude, etc.) is computed
85 to search for local maxima, whereas in second order ones, zero crossings in a second-order derivative, like the Laplacian computed from the image,
86 are searched in order to find edges.
87 As for as fuzzy edge methods are concerned, they are obviously based on fuzzy logic to highlight edges.
89 Canny filters, on their parts, are an old family of algorithms still remaining a state-of-the-art edge detector. They can be well approximated by first-order derivatives of Gaussians.
90 As the Canny algorithm is well known and studied, fast, and implementable
91 on many kinds of architectures like FPGAs, smartphones, desktop machines, and
92 GPUs, we have chosen this edge detector for illustrative purpose.
94 This edge detection is applied on a filtered version of the image given
96 More precisely, only $b$ most
97 significant bits are concerned by this step, where
98 the parameter $b$ is practically set with $6$ or $7$.
99 If set with the same value $b$, the edge detection returns thus the same
100 set of pixels for both the cover and the stego image.
101 In our flowcharts, this is represented by ``edgeDetection(b bits)''.
102 Then only the 2 LSBs of pixels in the set of edges are returned if $b$ is 6,
103 and the LSB of pixels if $b$ is 7.
104 Let $x$ be the sequence of these bits.
106 \JFC{il faudrait comparer les complexites des algo fuzy and canny}
109 % First of all, let us discuss about compexity of edge detetction methods.
110 % Let then $M$ and $N$ be the dimension of the original image.
111 % According to~\cite{Hu:2007:HPE:1282866.1282944},
112 % even if the fuzzy logic based edge detection methods~\cite{Tyan1993}
113 % have promising results, its complexity is in $C_3 \times O(M \times N)$
114 % whereas the complexity on the Canny method~\cite{Canny:1986:CAE:11274.11275}
115 % is in $C_1 \times O(M \times N)$ where $C_1 < C_3$.
116 % \JFC{Verifier ceci...}
117 % In experiments detailled in this article, the Canny method has been retained
118 % but the whole approach can be updated to consider
119 % the fuzzy logic edge detector.
126 As argue in the introduction section, we do not adapt the parameters of the
127 the edge detection as in~\cite{Luo:2010:EAI:1824719.1824720} but we modify
128 the size of the embedding message. Practically, the lenght of $x$
129 has to be at least twice as large
130 as the size of the embedded encrypted message.
131 Otherwise, a new image is used to hide the remaning part of the message.
133 \subsection{Minimizing Distortion with Syndrome-Treillis Codes}
138 % Edge Based Image Steganography schemes
139 % already studied~\cite{Luo:2010:EAI:1824719.1824720,DBLP:journals/eswa/ChenCL10,DBLP:conf/ih/PevnyFB10} differ
140 % how they select edge pixels, and
141 % how they modify these ones.
143 % First of all, let us discuss about compexity of edge detetction methods.
144 % Let then $M$ and $N$ be the dimension of the original image.
145 % According to~\cite{Hu:2007:HPE:1282866.1282944},
146 % even if the fuzzy logic based edge detection methods~\cite{Tyan1993}
147 % have promising results, its complexity is in $C_3 \times O(M \times N)$
148 % whereas the complexity on the Canny method~\cite{Canny:1986:CAE:11274.11275}
149 % is in $C_1 \times O(M \times N)$ where $C_1 < C_3$.
150 % \JFC{Verifier ceci...}
151 % In experiments detailled in this article, the Canny method has been retained
152 % but the whole approach can be updated to consider
153 % the fuzzy logic edge detector.
155 % Next, following~\cite{Luo:2010:EAI:1824719.1824720}, our scheme automatically
156 % modifies Canny parameters to get a sufficiently large set of edge bits: this
157 % one is practically enlarged untill its size is at least twice as many larger
158 % than the size of embedded message.
162 %%RAPH: paragraphe en double :-)
167 \subsection{Data Extraction}
168 The message extraction summarized in Fig.~\ref{fig:sch:ext} follows data embedding
169 since there exists a reverse function for all its steps.
170 First of all, the same edge detection is applied (on the 7 first bits) to
172 which is sufficiently large with respect to the message size given as a key.
173 Then the STC reverse algorithm is applied to retrieve the encrypted message.
174 Finally, the Blum-Goldwasser decryption function is executed and the original
175 message is extracted.