Preuve existe f pas cca equivalant exists f BA pas randomguess

1 files changed, 89 insertions, 0 deletions

diff --git a/synthetic/bck/main.aux b/synthetic/bck/main.aux
new file mode 100644
index 0000000..845d201
--- /dev/null
+++ b/synthetic/bck/main.aux
@@ -0,0 +1,89 @@
+\relax 
+\citation{gan}
+\citation{shokri2017membership}
+\citation{hawkins2004problem}
+\citation{yeom}
+\citation{abadi2016deep}
+\citation{EO}
+\citation{song2020overlearning}
+\@writefile{toc}{\contentsline {section}{\numberline {I}Introduction}{1}{}\protected@file@percent }
+\@writefile{toc}{\contentsline {subsection}{\numberline {\mbox  {I-A}}Research questions}{1}{}\protected@file@percent }
+\newlabel{sec:question}{{\mbox  {I-A}}{1}{}{}{}}
+\@writefile{toc}{\contentsline {section}{\numberline {II}Background}{1}{}\protected@file@percent }
+\@writefile{toc}{\contentsline {subsection}{\numberline {\mbox  {II-A}}Machine learning and classification}{1}{}\protected@file@percent }
+\@writefile{toc}{\contentsline {subsection}{\numberline {\mbox  {II-B}}Synthetic datas}{1}{}\protected@file@percent }
+\@writefile{toc}{\contentsline {subsection}{\numberline {\mbox  {II-C}}Membership inference attack}{1}{}\protected@file@percent }
+\@writefile{toc}{\contentsline {subsection}{\numberline {\mbox  {II-D}}Attribut inference attack}{1}{}\protected@file@percent }
+\citation{ding2021retiring}
+\citation{zhifei2017cvpr}
+\citation{dcgan}
+\citation{ctgan}
+\citation{ctgan}
+\citation{dcgan}
+\citation{cnn}
+\citation{cgan}
+\citation{vgg16}
+\citation{yeom}
+\citation{stadler2020synthetic}
+\@writefile{toc}{\contentsline {section}{\numberline {III}Methodology}{2}{}\protected@file@percent }
+\@writefile{toc}{\contentsline {subsection}{\numberline {\mbox  {III-A}}Datasets}{2}{}\protected@file@percent }
+\@writefile{toc}{\contentsline {subsubsection}{\numberline {\mbox  {III-A}1}US census (Adult)}{2}{}\protected@file@percent }
+\@writefile{toc}{\contentsline {subsubsection}{\numberline {\mbox  {III-A}2}CelebA}{2}{}\protected@file@percent }
+\@writefile{toc}{\contentsline {subsection}{\numberline {\mbox  {III-B}}Generator training}{2}{}\protected@file@percent }
+\newlabel{sec:gen}{{\mbox  {III-B}}{2}{}{}{}}
+\@writefile{toc}{\contentsline {subsubsection}{\numberline {\mbox  {III-B}1}CTGAN}{2}{}\protected@file@percent }
+\@writefile{toc}{\contentsline {subsubsection}{\numberline {\mbox  {III-B}2}DCGAN}{2}{}\protected@file@percent }
+\@writefile{toc}{\contentsline {subsection}{\numberline {\mbox  {III-C}}Predictor training}{2}{}\protected@file@percent }
+\newlabel{sec:target}{{\mbox  {III-C}}{2}{}{}{}}
+\@writefile{toc}{\contentsline {subsection}{\numberline {\mbox  {III-D}}Attack training}{2}{}\protected@file@percent }
+\@writefile{toc}{\contentsline {subsubsection}{\numberline {\mbox  {III-D}1}Attribute inference attack (AIA)}{2}{}\protected@file@percent }
+\@writefile{toc}{\contentsline {subsubsection}{\numberline {\mbox  {III-D}2}Membership inference attack (MIA)}{2}{}\protected@file@percent }
+\@writefile{lof}{\contentsline {figure}{\numberline {1}{\ignorespaces In this figure we detail the OVR CTRL function. This function control the overfitting of the target model. It takes a dataset of size at least $N$ and output a dataset of size $M$. First, we sample $N$ rows denoted $r_0,\cdots  ,r_{N-1}$ from the input dataset. Second, we repeat the rows $\lfloor \frac  {M}{N}\rfloor $ times. Finaly we shuffle the repeated rows.}}{3}{}\protected@file@percent }
+\providecommand*\caption@xref[2]{\@setref\relax\@undefined{#1}}
+\newlabel{fig:ovr}{{1}{3}{}{}{}}
+\@writefile{toc}{\contentsline {subsection}{\numberline {\mbox  {III-E}}Overfitting control}{3}{}\protected@file@percent }
+\newlabel{sec:ovr}{{\mbox  {III-E}}{3}{}{}{}}
+\@writefile{toc}{\contentsline {subsection}{\numberline {\mbox  {III-F}}Data pipeline}{3}{}\protected@file@percent }
+\@writefile{lof}{\contentsline {figure}{\numberline {2}{\ignorespaces Methodology for finding an amount of repetition that both achieve satisfying utility and a high sensitivity to MIA. We use a total number of 100000 points. In this experiment the only generator used is CTGAN. The results presented are for the Adult dataset but we apply the same process for CelebA using DCGAN. }}{3}{}\protected@file@percent }
+\newlabel{fig:tune_ovr}{{2}{3}{}{}{}}
+\newlabel{sec:data}{{\mbox  {III-F}}{3}{}{}{}}
+\@writefile{toc}{\contentsline {section}{\numberline {IV}Comparisons between synthetic and real data}{3}{}\protected@file@percent }
+\citation{bellovin2019privacy}
+\citation{stadler2020synthetic}
+\citation{bellovin2019privacy}
+\citation{ping2017datasynthesize}
+\@writefile{lof}{\contentsline {figure}{\numberline {3}{\ignorespaces This figure presents the data splits and subsets used to compute results. It is a representation of the whole methodology described in this sections. The reader may start on the top left corner, with the real data. The rectangle boxes represents functions where the inputs are incoming arrows and the outputs outcomings arrows. In the case of trainable functions such as machine learning models, we indicate that an input is the training data with the label "training". We use a similar notation for evaluation. }}{4}{}\protected@file@percent }
+\newlabel{fig:split}{{3}{4}{}{}{}}
+\@writefile{toc}{\contentsline {section}{\numberline {V}Results}{4}{}\protected@file@percent }
+\newlabel{sec:res}{{V}{4}{}{}{}}
+\@writefile{toc}{\contentsline {subsection}{\numberline {\mbox  {V-A}}Utility}{4}{}\protected@file@percent }
+\newlabel{sec:uti}{{\mbox  {V-A}}{4}{}{}{}}
+\@writefile{toc}{\contentsline {subsection}{\numberline {\mbox  {V-B}}Membership inference attack}{4}{}\protected@file@percent }
+\@writefile{toc}{\contentsline {subsection}{\numberline {\mbox  {V-C}}Attribute inference attack}{4}{}\protected@file@percent }
+\@writefile{lof}{\contentsline {figure}{\numberline {4}{\ignorespaces Utility of the target model in terms of balanced accuracy evaluated on unseen data. The "Real" label refers to a generator equal to identity, hence the synthetic data used to train the target model is the real data. THe "Synthetic" label refers to a CGAN generator for Adult and CTGAN for CelebA, hence the synthetic data are generated sampled according to a distribution learned by the genrator model. In this case the target model is not trained on real date.}}{4}{}\protected@file@percent }
+\@writefile{lof}{\contentsline {figure}{\numberline {5}{\ignorespaces Success of the MIA in terms of balanced accuracy evaluated on the Train part of MIA dataset.}}{4}{}\protected@file@percent }
+\@writefile{toc}{\contentsline {section}{\numberline {VI}Related work}{4}{}\protected@file@percent }
+\@writefile{lof}{\contentsline {figure}{\numberline {6}{\ignorespaces Success of the AIA in terms of balanced accuracy evaluated on the Train part of AIA dataset.}}{4}{}\protected@file@percent }
+\citation{jordon2021hide}
+\bibstyle{plain}
+\bibdata{biblio}
+\bibcite{abadi2016deep}{1}
+\bibcite{bellovin2019privacy}{2}
+\bibcite{ding2021retiring}{3}
+\bibcite{gan}{4}
+\bibcite{EO}{5}
+\bibcite{hawkins2004problem}{6}
+\bibcite{jordon2021hide}{7}
+\bibcite{cgan}{8}
+\bibcite{dcgan}{9}
+\bibcite{cnn}{10}
+\bibcite{shokri2017membership}{11}
+\bibcite{vgg16}{12}
+\bibcite{song2020overlearning}{13}
+\bibcite{stadler2020synthetic}{14}
+\bibcite{ctgan}{15}
+\bibcite{yeom}{16}
+\bibcite{zhifei2017cvpr}{17}
+\@writefile{toc}{\contentsline {section}{\numberline {VII}Conclusion}{5}{}\protected@file@percent }
+\@writefile{toc}{\contentsline {section}{References}{5}{}\protected@file@percent }
+\gdef \@abspage@last{5}