Combining Movement Model with Finger-Stroke Level Model Towards Designing a Security Enhancing Mobile Friendly CAPTCHA.

Although many solutions have been proposed for peripheral device controlled desktop applications to separate Human-Bot using CAPTCHAs, no significant work has addressed the same issue for handheld touch-sensitive devices. In this work, we propose a novel CAPTCHA system based on flicking or dragging segmented sub-images to a specific output box to regenerate a sample image. The successful matching of sample and output images can be considered to be submitted by humans only if the movement data of dragging object shows a human-like pattern and threshold crossing mismatch with the pattern for dragging performed by the bot. We designed the prototype of our proposed CAPTCHA for both PC and Mobile platform to underline the subtle, yet inevitable difference between key-stroke and finger-stroke. After comparing and analyzing evaluation factors like solving time, accuracy and reat-tempt requirements from 31 users with 3 different physical issues, we concluded our paper with two important discussions. First, the inclusion of extreme noise may become a handful in first-order bot identification, but an alarming drop of accuracy for specific users with vision and motor disabilities forced us to keep noise level in a moderated state. And finally, the task completion time of required action factors associated with unit tasks to solve the puzzle for both mobile and desktop versions indicates that our proposed finger-stroke based CAPTCHA performs faster than a key-stroke based traditional model.