blur text, context and conclusion

This commit is contained in:
Aravinth Manivannan 2021-09-20 19:31:02 +05:30
parent c4952942f1
commit aa42a34764
Signed by: realaravinth
GPG key ID: AD9F0F08E855ED88
6 changed files with 90 additions and 6 deletions

31
analysis/blur-text.tex Normal file
View file

@ -0,0 +1,31 @@
\subsection{Noisy Text}
\subsubsection{Privacy}
Excellent\\
The method doesn't employ any tracking elements and works when used in anonymous
networks like TOR\@.
\subsubsection{Effectiveness}
Bad\\
The method relies on Optical Character Recognition (OCR) capabilities of human
users. OCR technology is becoming increasingly sophisticated which would render
this method ineffective in the future.
Without OCR, this method can be bypassed using human-powered CAPTCHA farms with
% TODO cite CAPTCHA farm cost analysis paper
ease.
\subsubsection{Accessibility}
Bad\\
Since the method relies on OCR, it is inaccessible to users with visual
disabilities and the added noise poses challenges to users with cognitive
disabilities. The noise sometimes makes the challenge unsolvable which results
in failures for even users without any disabilities,
\subsubsection{Accuracy}
Good\\
Success and failure are absolute states in this method. A solution that doesn't
match the challenge text results in failure while a match is considered a
success. But when using employing CAPTCHA farms for circumvention, the method
fails absolutely.
The method uses only text matches in its decision process. No other
external factors are involved.

36
analysis/context.tex Normal file
View file

@ -0,0 +1,36 @@
\subsection{Context-based}
\subsubsection{Privacy}
Excellent\\
The method doesn't employ any tracking elements and works when used in anonymous
networks like TOR\@.
\subsubsection{Effectiveness}
Bad\\
Domain context is limited in nature. An attacker could either become familiar
with the service and then mount an attack, or they could mount brute force attack
to aggregate all possible challenges that the service presents.
Both of these attacks are practical.
% TODO cite CAPTCHA farm cost analysis paper
ease.
\subsubsection{Accessibility}
Bad\\
This method poses challenges to users with cognitive disabilities as it requires
them to retain information and recollect in a timely manner. Also, this method
makes poses challenges to new visitors to service. The familiarization period to
solve CAPTCHAs in a timely fashion might be too high for someone new.
\subsubsection{Accuracy}
Bad\\
Success and failure are absolute states in this method. A solution that doesn't
match the challenge text results in failure while a match is considered a
success. But when above-mentioned circumvention methods are used, there will be
results will be completely inaccurate and as the system lacks any adaptational
capabilities, the failure will be long-lasting.
The method uses only challenge answers in its decision process. No other
external factors are involved.

View file

@ -1,2 +1,4 @@
\section{Analysis} \section{Analysis}
\input{analysis/align-obj.tex} \input{analysis/align-obj.tex}
\input{analysis/blur-text.tex}
\input{analysis/context.tex}

13
conclusion.tex Normal file
View file

@ -0,0 +1,13 @@
\section{Conclusion}
In today's web, CAPTCHA systems are the deciding factor in weather a user should
be able to access a service or not. Every false positive makes the web less
accessible and even a single challenge posed to persons with disabilities makes
the web inaccessible entire to groups of users. Also, the
ubiquitous presence of privacy-invasive systems like reCAPTCHA post serious
threats to the internet and its users as right to access to information is
now determined by single entity using an opaque process.
So it is imperative that future system designers take these factors into
consideration. We believe mCaptcha is an attempt at making the web more
accessible while securing it with a decentralised command structure.

View file

@ -6,14 +6,14 @@ parameters. These are popular methods are currently in deployment.
\subsubsection{Align object} \subsubsection{Align object}
Objects in various degrees of misalignments are displayed to the user and are Objects in various degrees of misalignments are displayed to the user and are
asked to choose the one that is perfectly aligned. asked to choose the one that is perfectly aligned.
% Example GitHub/Kik inverted Hipop % Example GitHub/Kik inverted Hippo
\subsubsection{Blurred Text} \subsubsection{Noisy Text}
A sequence of randomly generated letters and digits are A sequence of randomly generated letters and digits are presented to the user
presented to the user with added noise, scattered distribution and with added noise, scattered distribution and rotations. Sometimes, they are also
rotations. Sometimes, they are also presented in 3D form. presented in 3D form.
\subsubsection{Context based} \subsubsection{Context-based}
This method is personalised to the platforms they are displayed on. They usually This method is personalised to the platforms they are displayed on. They usually
pose challenges which can only be solved if the user is familiar with the pose challenges which can only be solved if the user is familiar with the
platforms. Some examples are: platforms. Some examples are:

View file

@ -13,4 +13,6 @@
\input{pre-req/main.tex} \input{pre-req/main.tex}
\input{analysis/main.tex} \input{analysis/main.tex}
\input{conclusion.tex}
\end{document} \end{document}