Improving Captcha on account creation page using Machine Learning
I am an Outreachy Round 15 intern and I am working on AICaptcha
project. This project is aimed at creating a better captcha system (like
the Google invisible captcha) which can prevent/reduce the incidence of
bots creating user accounts and spamming Wikipedia. My mentors on this
project are Gergő Tisza and Adam Roses Wight.
The key aspects of this project are:
1. Data capture for training a machine learning classifier which is
elaborated in Phabricator task. The data can be captured from the
registration page using the WikiMediaEvents extension.
2. Feature selection, dealing with selecting the most appropriate features
which can improve the classification model, explained in Phabricator task
3. Finding appropriate machine learning classifier to create the model .
Kindly provide suggestions/ideas on Phabricator, so that any idea missed by
oversight can be discussed. Also if there are any possible issues which I
have not thought about yet, please comment on the tasks so that I can take
care of them sooner rather than later.