Commit.tex

\documentclass[10pt, a4paper]{article}
\usepackage{xcolor}
\pagestyle{headings}
\usepackage{setspace}
\setstretch{0.9}
%----
\begin{document}
\renewcommand{\thepage}{\arabic{page}}% Arabic page numbers
\pagecolor{yellow!20}
% Page 1
% Here are our name.
\begingroup  
  \centering
  \large \underline{\textbf{Malicious PDF File Detection - Commitment Document}}\\
  \bigbreak  
  \large \textit{Shir Bentabou \qquad Alexey Titov}\\ 
  \bigbreak 
  \large \textit{Supervisors: Dr. Amit Dvir and Dr. Ran Dubin}\\
\endgroup
\section{\large Accomplished Tasks:}
\begin{itemize}
% 1
\item Hand in project abstract.
% 2
\item Hand in project proposal to supervisors.
% 3
\item Hand in project poster.
% 4
\item Deeply know PDF file structure, features and fields.
% 5
\item Study about phishing, URLs, and JavaScript uses in PDF files.
% 6
\item Research methods from previous researches.
% 7
\item Build the plan and schedule for our project: phases, tasks in each phase, deadlines for each task.
% 8
\item Research existing tools for our usage in the project.
% 9
\item \underline{First phase} – Researching and creating our work tools:
	\begin{itemize}
	% 9.1
	\item Extracting telemetry.
	% 9.2
	\item Extracting text from picture.
	% 9.3
	\item Extracting text from pdf file (using PDFMiner).
	% 9.4
	\item Extracting URLs (using pyPDF).
	% 9.5
	\item Extracting URLs from JS in the file tags (using peePDF).
	% 9.6
	\item Extracting preview of a PDF file (using PIL + pdf2image).
	\end{itemize}
% 10
\item \underline{Second phase} – Creating an image-based classification machine:
	\begin{itemize}
	% 10.1
	\item Research vector features.
	% 10.2
	\item Building the feature vector.
	% 10.3
	\item Applying machine learning algorithms on the feature vector.
	\end{itemize}
% 11
\item \underline{Third phase} – Creating a text-based classification machine:
	\begin{itemize}
	% 11.1
	\item Research vector building methods.
	% 11.2
	\item Applying the vector methods on our samples.
	% 11.3
	\item Applying machine learning algorithms on the text vector.
	% 11.4
	\item Applying a deep learning method on the text vector.
	\end{itemize}
% 12
\item \underline{Fourth phase} – Creating a classification machine based on PDF tags, JS, URLs, objects and streams:
	\begin{itemize}
	% 12.1
	\item Researching the features that will build the vector for this machine in each one of the four parts: PDF tags, JS, URLs, objects and streams.
	% 12.2
	\item Research existing tools for the extraction of the features chosen (JAST, Analyze PDF, peePDF).
	% 12.3
	\item Extraction of the features from samples.
	% 12.4
	\item Building the feature vector.
	% 12.5
	\item Applying machine learning algorithms on the feature vector.
	% 12.6
	\item Applying a deep learning method on the feature vector.
	\end{itemize}
\end{itemize}


% Page 2
\newpage
\section{\large To Be Accomplished:}
\begin{itemize}
% 1
\item Prepare project day presentation.
% 2
\item Writing project book.
% 3
\item \underline{Fifth phase} – Creating an ensemble machine:
	\begin{itemize}
	% 3.1
	\item Combining the three machines into an ensemble machine.
	% 3.2
	\item  Determining the overall classification method for the ensemble machine.
	% 3.3
	\item Applying machine learning algorithms to ensemble machine:
		\begin{itemize}
		% 3.3.1
		\item Random Forest, AdaBoost (Adaptive Boosting), Gradient Tree Boosting, XGBoost.
		\end{itemize}
	\end{itemize}
% 4
\item \underline{Improvement phase} - Deciding improvement phase aim, and numeric success rate for each classifier, and ensemble machine as well.
% 5 Phases
\item Improvements for each phase:
	\begin{itemize}
	% 5.1 	
	\item Second phase:
		\begin{itemize}
		% 5.1.1
		\item Try to improve picture classification in two ways:
			\begin{itemize}
			% 5.1.1.1
			\item Applying additional vector building methods (such as near similar image matching).
			% 5.1.1.2
			\item Applying additional machine learning algorithms on the vectors.
			\end{itemize}
		\end{itemize}
		% 5.1.2 		
		\item Third phase:
		\begin{itemize}
		% 5.1.2.1
		\item Try to improve text classification in the following way:
			\begin{itemize}
			% 5.1.2.1.1
			\item Applying additional machine learning algorithms on the different vector building methods (word2vec, TF-IDF) to achieve better results.
			\end{itemize}
		\end{itemize}
		% 5.1.2.2
		\item Fourth phase:
		\begin{itemize}
		% 5.1.2.2.1
		\item Improve feature selection in the following ways:
			\begin{itemize}
			% 5.1.2.2.1.1
			\item Random choice method.
			% 5.1.2.2.1.2
			\item Summing features method.
			% 5.1.2.2.1.3
			\item Combining features as new features in vector.
			\end{itemize}
		% 5.1.2.2.2
		\item Applying additional machine learning algorithms on the vectors.
		\end{itemize}
	\end{itemize}
% 6
\item Overall improvements:
	\begin{itemize}
	% 6.1
	\item Applying iterative retraining methods on the machines.
	\end{itemize}
\end{itemize}

\end{document}