FP-Captcha: An Improved Captcha Design Scheme Based on Face Points

Abstract

Captcha’s are random significant challenges, used to differentiate between humans and bots in order to prevent unauthorized access to web services. Over a decade lots of research have been carried out for generating numerous types of Captcha’s; specially using image and face detection and recognition techniques. Rather than machine, humans are outstandingly good at several things, including pattern recognition, linguistic abilities, and innovative thinking. Although, computers are rapidly improving the same, but most programs still don’t recognize as well as a child. Humans are inherently capable of detecting faces in a variety of contexts; even parts of face portions are not visible yet. Modern advancement in Artificial Intelligence, Deep Learning and advanced Image Processing techniques have made the Captcha based security system unsecure and vulnerable. However, as the time progresses Captcha’s are equipped with high usability, robustness and producing new unique challenges. This paper proposes a novel face point based Captcha, which employs various face points detection as its test, where user will ask to click on correct face points of all human faces presented in the Captcha challenge; which comprises of real and fake face images, with balanced noise and distortions, embedded in a composite background. Over 100 unique and random FP-Captcha’s are generated and solved by 115 UG-students, with accuracy of 97.39% for correct responses. Consequently, the probability of passing a FP-Captcha test by random guessing attack was 0.00000245%. Therefore, we conclude that, FP-Captcha is secure and robust against malicious attacks and offer better human accuracy.