overview.tex.bak

\section{System Design Goals and Overview}
    We start this section by defining the design goals of our authentication system. We then gives an system overview which leverages the dynamic deformation features of facial expressions to authenticate identity.
    \begin{figure*}[!ht]
        \centering
        \includegraphics[width=\textwidth]{fig/overview.pdf}
        \caption{Overview of the system.}
        \label{fig:overview}
    \end{figure*}
    \subsection{Design Goals}
        Biometrics which refers to either the static physiological traits or dynamic behavioral modalities are all non-revocable. This is one of the major reasons why biometrics is possible to be stolen or replayed by an expert adversary. Unlike traditional biometrics, facial behavioral traits is unique and can be changed with the changes of facial expressions. This is the biggest difference comparing with other biometric-based authentication systems. Therefore, facial behavior-based modalities are more secure than other biometrics in some extent.  Furthermore, it is possible to decrease the authentication time as much as possible with the development of facial detection technologies as well as not requiring the visual stimuli. This making facial biometrics are more user-friendly. In conclusion, a secure biometric-based authentication system should be secure, fast and user-friendly. The following we summary the design goals of our system:
        \begin{itemize}
            \item \emph{Simple:} This system should be simple as much as possible. Specifically, user should be minimal coordinate the system during authenticating process.
            \item \emph{Fast:} A single authentication duration should be as short as possible.
            \item \emph{Secure:} The system is able to immune to the replay and impersonation attacks.
            \item \emph{Revocable:} The biometrics can be changed once it is stolen or leakage.
        \end{itemize}

    \subsection{System Overview}
        The system authenticates the user's identity by analyzing the dynamic changes of facial expressions. It records the entire change of facial expressions using in-built front camera of smartphone. The deformation features of facial expressions are extracted by existing facial detection algorithm and they are used to recognize the user's identity.
        Figure ~\ref{fig:overview} depicts the steps of this system:

        \noindent \circled{1} \textbf{Filming and Landmarks Tracking:} The authentication process begins from capturing the facial expressions. The video footage is filmed by in-built front camera of smartphone. During recording, the facial landmarks are simultaneously tracked automatically by facial detection algorithm.
        
        \noindet \circled{2} \textbf{}

        \noindent \circled{3} \textbf{Feature Extraction:} Once the facial landmarks are located, two types of feature extraction algorithms will be separately applied to extract two kinds of facial expression features. One is the Gabor features and the other is facial deformation features.

        \noindent \circled{4} \textbf{Identity Authentication:} In this step, the system first verify the reality of the facial expressions by the part of facial deformation features.  That is to say that whether or not the current facial expression is recorded by attacker in advance. Then it will check the user's identity combining the deformation features and the Gabor features. Finally, the system outputs the authentication results comparing to the owner's template trained when enrollment.