Web security through improvised image based CAPTCHA

Abstract

CAPTCHA is a Completely Automated Public Turing Test to tell Computers and Humans Apart [1]. Whenever a website is hosted onto a server it is not human that always tries to access the website. Sometimes a human generated computer bot also known as Zombie may try to access the website. In such situations bots need to be filtered from legitimate users and this can be accomplished by using a simple method called CAPTCHA. Initially CAPTCHA was developed on text-based platform later it was evolved to audio, image etc. In this paper a new technique to differentiate bots from humans is introduced. It is a question-based image-oriented technique where a question comprising of keywords is posed along with four different images out of which only one image has relation with all the keywords. The user needs to spot the image to pass the test. A flag will be counting the number of times the user failed the test and depending on the number of failure attempts the user is judged.