GESTCHA: a gesture-based CAPTCHA design for smart devices using angular velocity

Abstract

The “Completely Automated Public Turing test to Tell Computers and Human Apart” (CAPTCHA) is a standard security protocol, wildly used to distinguish between human and malicious computer program, known as bot. This paper presents a gesture-based CAPTCHA system i.e. GESTCHA utilizing angular velocity data from embedded gyroscope sensors of handheld touch-sensitive devices. The angular velocity data collected within a short exposure of time was processed as gesture input. Several discriminatory feature sets were extracted from stable gyroscope reading by applying a newly proposed gesture feature extraction algorithm. We analyzed the performance of two different machine learning algorithms, i.e. Naive Bayes and Random Forest, over the training gesture pattern which led to the development of a robust gesture recognition model. Using the model, final prototype of GESTCHA was proposed. Based on findings from a comparative usability study with 850 participants, GESTCHA shows significant improvement in terms of solving rate and solving time compared to Google’s reCAPTCHA v3 i.e. NoCAPTCHA.