X-Git-Url: https://bilbo.iut-bm.univ-fcomte.fr/and/gitweb/canny.git/blobdiff_plain/e477428ed31d945b364bd9064476c1f0a34f86ef..af81455342fce2b4d7f96ecb4f1194635a5a13a4:/experiments.tex?ds=sidebyside diff --git a/experiments.tex b/experiments.tex index 9da4704..5ba43df 100644 --- a/experiments.tex +++ b/experiments.tex @@ -1,104 +1,141 @@ -For whole experiments, the whole set of 10000 images +First of all, the whole code of STABYLO can be downloaded +\footnote{\url{http://members.femto-st.fr/raphael-couturier/stabylo/}}. +For all the experiments, the whole set of 10,000 images of the BOSS contest~\cite{Boss10} database is taken. In this set, each cover is a $512\times 512$ grayscale digital image in a RAW format. We restrict experiments to this set of cover images since this paper is more focused on -the methodology than benchmarking. -Our approach is always compared to Hugo~\cite{DBLP:conf/ih/PevnyFB10} -and to EAISLSBMR~\cite{Luo:2010:EAI:1824719.1824720}. -The former is the less detectable information hidding tool in spatial domain -and the later is the work which is close to ours, as far as we know. - +the methodology than on benchmarks. + +We use the matrices $\hat{H}$ +generated by the integers given +in Table~\ref{table:matrices:H} +as introduced in~\cite{FillerJF11}, since these ones have experimentally +be proven to have the strongest modification efficiency. +For instance if the rate between the size of the message and the size of the +cover vector +is 1/4, each number in $\{81, 95, 107, 121\}$ is translated into a binary number +and each one constitutes thus a column of $\hat{H}$. + +\begin{table} +$$ +\begin{array}{|l|l|} +\hline +\textrm{Rate} & \textrm{Matrix generators} \\ +\hline +{1}/{2} & \{71,109\}\\ +\hline +{1}/{3} & \{95, 101, 121\}\\ +\hline +{1}/{4} & \{81, 95, 107, 121\}\\ +\hline +{1}/{5} & \{75, 95, 97, 105, 117\}\\ +\hline +{1}/{6} & \{73, 83, 95, 103, 109, 123\}\\ +\hline +{1}/{7} & \{69, 77, 93, 107, 111, 115, 121\}\\ +\hline +{1}/{8} & \{69, 79, 81, 89, 93, 99, 107, 119\}\\ +\hline +{1}/{9} & \{69, 79, 81, 89, 93, 99, 107, 119, 125\}\\ +\hline +\end{array} +$$ +\caption{Matrix Generator for $\hat{H}$ in STC}\label{table:matrices:H} +\end{table} -First of all, in our experiments and with the adaptive scheme, -the average size of the message that can be embedded is 16,445 bits. -Its corresponds to an average payload of 6.35\%. -The two other tools will then be compared with this payload. -The Sections~\ref{sub:quality} and~\ref{sub:steg} respectively present -the quality analysis and the security of our scheme. +Our approach is always compared to HUGO, to EAISLSBMR, to WOW and to UNIWARD +for the two strategies Fixed and Adaptive. +For the former one, the payload has been set to 10\%. +For the latter one, the Canny parameter $T$ has been set to 3. +When $b$ is 7, the average size of the message that can be embedded +is 16,445 bits, +that corresponds to an average payload of 6.35\%. +For each cover image the STABYLO's embedding rate with these two parameters +is memorized. +Next each steganographic scheme is executed to produce the stego content of +this cover with respect to this embedding rate. -\subsection{Image Quality}\label{sub:quality} +\subsection{Image quality}\label{sub:quality} The visual quality of the STABYLO scheme is evaluated in this section. For the sake of completeness, three metrics are computed in these experiments: the Peak Signal to Noise Ratio (PSNR), the PSNR-HVS-M family~\cite{psnrhvsm11}, -%the BIQI~\cite{MB10}, and the weighted PSNR (wPSNR)~\cite{DBLP:conf/ih/PereiraVMMP01}. The first one is widely used but does not take into account the Human Visual System (HVS). The other ones have been designed to tackle this problem. +If we apply them on the running example with the Adaptive and STC strategies, +the PSNR, PSNR-HVS-M, and wPSNR values are respectively equal to +68.39, 79.85, and 89.71 for the stego Lena when $b$ is equal to 7. +If $b$ is 6, these values are respectively equal to +65.43, 77.2, and 89.35. + \begin{table*} \begin{center} -\begin{tabular}{|c|c|c||c|c|c|c|c|} +\begin{small} +\setlength{\tabcolsep}{3pt} +\begin{tabular}{|c|c|c||c|c|c|c|c|c|c|c|c|c|} \hline -Schemes & \multicolumn{3}{|c|}{STABYLO} & \multicolumn{2}{|c|}{HUGO}& \multicolumn{2}{|c|}{EAISLSBMR} \\ +Schemes & \multicolumn{4}{|c|}{STABYLO} & \multicolumn{2}{|c|}{HUGO}& \multicolumn{2}{|c|}{EAISLSBMR} & \multicolumn{2}{|c|}{WOW} & \multicolumn{2}{|c|}{UNIWARD}\\ \hline -Embedding & Fixed & \multicolumn{2}{|c|}{Adaptive} & \multicolumn{2}{|c|}{Fixed}& \multicolumn{2}{|c|}{Fixed} \\ +Embedding & Fixed & \multicolumn{3}{|c|}{Adaptive (about 6.35\%)} & Fixed &Adaptive & Fixed &Adaptive & Fixed &Adaptive & Fixed &Adaptive \\ \hline -Rate & 10\% & + sample & + STC & 10\%&6.35\%& 10\%&6.35\%\\ +Rate & 10\% & + sample & +STC(7) & +STC(6) & 10\%&$\approx$6.35\%& 10\%&$\approx$6.35\%& 10\%&$\approx$6.35\%& 10\%&$\approx$6.35\%\\ \hline -PSNR & 61.86 & 63.48 & 66.55 (\textbf{-0.8\%}) & 64.65 & {67.08} & 60.8 & 62.9\\ +PSNR & 61.86 & 63.48 & 66.55 & 63.7 & 64.65 & {67.08} & 60.8 & 62.9&65.9 & 68.3 & 65.8 & 69.2\\ \hline -PSNR-HVS-M & 72.9 & 75.39 & 78.6 (\textbf{-0.8\%}) & 76.67 & {79.23} & 61.3 & 63.4\\ -%\hline -%BIQI & 28.3 & 28.28 & 28.4 & 28.28 & 28.28 & 28.2 & 28.2\\ +PSNR-HVS-M & 72.9 & 75.39 & 78.6 & 75.5 & 76.67 & {79.6} & 71.8 & 76.0 & +76.7 & 80.35 & 77.6 & 81.2 \\ \hline -wPSNR & 77.47 & 80.59 & 86.43(\textbf{-1.6\%}) & 83.03 & {87.8} & 76.7 & 80.6\\ +wPSNR & 77.47 & 80.59 & 86.43& 86.28 & 83.03 & {88.6} & 76.7 & 83& 83.8 & 90.4 & 85.2 & 91.9\\ \hline \end{tabular} - -\begin{footnotesize} -\vspace{2em} -Variances given in bold font express the quality differences between -HUGO and STABYLO with STC+adaptive parameters. -\end{footnotesize} - +\end{small} \end{center} -\caption{Quality Measures of Steganography Approaches\label{table:quality}} +\caption{Quality measures of steganography approaches\label{table:quality}} \end{table*} -Results are summarized into the Table~\ref{table:quality}. +Results are summarized in Table~\ref{table:quality}. +In this table, STC(7) stands for embedding data in the LSB whereas +in STC(6), data are hidden in the last two significant bits. + + Let us give an interpretation of these experiments. -First of all, the adaptive strategy produces images with lower distortion -than the one of images resulting from the 10\% fixed strategy. +First of all, the Adaptive strategy produces images with lower distortion +than the images resulting from the 10\% fixed strategy. Numerical results are indeed always greater for the former strategy than -for the latter. -These results are not surprising since the adaptive strategy aims at -embedding messages whose length is decided according to an higher threshold +for the latter one. +These results are not surprising since the Adaptive strategy aims at +embedding messages whose length is decided according to a higher threshold into the edge detection. -Let us focus on the quality of HUGO images: with a given fixed -embedding rate (10\%), -HUGO always produces images whose quality is higher than the STABYLO's one. -However our approach always outperforms EAISLSBMR since this one may modify -the two least significant bits whereas STABYLO only alter LSB. - -If we combine \emph{adaptive} and \emph{STC} strategies -(which leads to an average embedding rate equal to 6.35\%) -our approach provides equivalent metrics than HUGO. -The quality variance between HUGO and STABYLO for these parameters -is given in bold font. It is always close to 1\% which confirms -the objective presented in the motivations: -providing an efficient steganography approach with a lightweight manner. -Let us now compare the STABYLO approach with other edge based steganography -approaches, namely~\cite{DBLP:journals/eswa/ChenCL10,Chang20101286}. -These two schemes focus on increasing the -payload while the PSNR is acceptable, but do not -give quality metrics for fixed embedding rates from a large base of images. +If we combine Adaptive and STC strategies +the STABYLO scheme provides images whose quality is higher than +the EAISLSBMR's one but lower than the quality of high complexity +schemes. Notice that the quality of the less respectful scheme (EAILSBMR) +is lower than 6\% than the one of the most one. + + +% Let us now compare the STABYLO approach with other edge based steganography +% approaches, namely~\cite{DBLP:journals/eswa/ChenCL10,Chang20101286}. +% These two schemes focus on increasing the +% payload while the PSNR is acceptable, but do not +% give quality metrics for fixed embedding rates from a large base of images. @@ -107,53 +144,72 @@ give quality metrics for fixed embedding rates from a large base of images. -The steganalysis quality of our approach has been evaluated through the two -AUMP~\cite{Fillatre:2012:ASL:2333143.2333587} -and Ensemble Classifier~\cite{DBLP:journals/tifs/KodovskyFH12} based steganalysers. -Both aims at detecting hidden bits in grayscale natural images and are -considered as the state of the art of steganalysers in spatial domain~\cite{FK12}. -The former approach is based on a simplified parametric model of natural images. -Parameters are firstly estimated and an adaptive Asymptotically Uniformly Most Powerful -(AUMP) test is designed (theoretically and practically), to check whether -an image has stego content or not. -This approach is dedicated to verify whether LSB has been modified or not. -In the latter, the authors show that the -machine learning step, which is often -implemented as support vector machine, -can be favorably executed thanks to an ensemble classifier. +The steganalysis quality of our approach has been evaluated through the % two +% AUMP~\cite{Fillatre:2012:ASL:2333143.2333587} +% and +Ensemble Classifier~\cite{DBLP:journals/tifs/KodovskyFH12} based steganalyser. +Its particularization to spatial domain is +considered as state of the art steganalysers. +Firstly, a space +of 686 co-occurrence and Markov features is extracted from the +set of cover images and the set of training images. Next a small +set of weak classifiers is randomly built, +each one working on a subspace of all the features. +The final classifier is constructed by a majority voting +between the decisions of these individual classifiers. + + +%The former approach is based on a simplified parametric model of natural images. +% Parameters are firstly estimated and an adaptive Asymptotically Uniformly Most Powerful +% (AUMP) test is designed (theoretically and practically), to check whether +% an image has stego content or not. +% This approach is dedicated to verify whether LSB has been modified or not. +% , the authors show that the +% machine learning step, which is often +% implemented as a support vector machine, +% can be favorably executed thanks to an ensemble classifier. \begin{table*} \begin{center} -%\begin{small} -\begin{tabular}{|c|c|c|c|c|c|c|c|} +\begin{small} +\setlength{\tabcolsep}{3pt} +\begin{tabular}{|c|c|c|c|c|c|c|c|c|c|c|c|c|} \hline -Schemes & \multicolumn{3}{|c|}{STABYLO} & \multicolumn{2}{|c|}{HUGO}& \multicolumn{2}{|c|}{EAISLSBMR}\\ +Schemes & \multicolumn{4}{|c|}{STABYLO} & \multicolumn{2}{|c|}{HUGO}& \multicolumn{2}{|c|}{EAISLSBMR} & \multicolumn{2}{|c|}{WOW} & \multicolumn{2}{|c|}{UNIWARD}\\ \hline -Embedding & Fixed & \multicolumn{2}{|c|}{Adaptive} & \multicolumn{2}{|c|}{Fixed}& \multicolumn{2}{|c|}{Fixed} \\ +Embedding & Fixed & \multicolumn{3}{|c|}{Adaptive (about 6.35\%)} & Fixed & Adapt. & Fixed & Adapt. & Fixed & Adapt. & Fixed & Adapt. \\ \hline -Rate & 10\% & + sample & + STC & 10\%& 6.35\%& 10\%& 6.35\%\\ +Rate & 10\% & + sample & +STC(7) & +STC(6) & 10\%& $\approx$6.35\%& 10\%& $\approx$6.35\% & 10\%& $\approx$6.35\%& 10\%& $\approx$6.35\%\\ \hline -AUMP & 0.22 & 0.33 & 0.39 & 0.50 & 0.50 & 0.49 & 0.50 \\ -\hline -Ensemble Classifier & 0.35 & 0.44 & 0.47 & 0.48 & 0.49 & 0.43 & 0.46 \\ +%AUMP & 0.22 & 0.33 & 0.39 & 0.45 & 0.50 & 0.50 & 0.49 & 0.50 \\ +%\hline +Ensemble Classifier & 0.35 & 0.44 & 0.47 & 0.47 & 0.48 & 0.49 & 0.43 & 0.47 & 0.48 & 0.49 & 0.46 & 0.49 \\ \hline \end{tabular} -%\end{small} +\end{small} \end{center} \caption{Steganalysing STABYLO\label{table:steganalyse}} \end{table*} -Results are summarized in Table~\ref{table:steganalyse}. -First of all, STC outperforms the sample strategy for the two steganalysers, as +Results of average testing errors +are summarized in Table~\ref{table:steganalyse}. +First of all, STC outperforms the sample strategy %for % the two steganalysers + as already noticed in the quality analysis presented in the previous section. -Next, our approach is more easily detectable than HUGO, which -is the most secure steganographic tool, as far as we know. -However by combining \emph{adaptive} and \emph{STC} strategies -our approach obtains similar results than HUGO ones. +Next, our approach is more easily detectable than HUGO, +WOW and UNIWARD which are the most secure steganographic tool, +as far as we know. +However by combining Adaptive and STC strategies +our approach obtains similar results than the ones of these schemes. + +Compared to EAILSBMR, we obtain better results when the strategy is +Adaptive. However due to its -huge number of features integration, it is not lightweight, which justifies -in the authors' opinion the consideration of the proposed method. +huge number of integration features, it is not lightweight. +All these numerical experiments confirm +the objective presented in the motivations: +providing an efficient steganography approach in a lightweight manner.