ABSTRACT
Completely Automated Public Turing Tests to Tell Computers and Humans Apart (CAPTCHAs) are now almost a routine security measure which protect against unwanted and malicious bot programmes on the Internet. There are several security threats on websites and it would be a major risk to the nation if defense websites or other classified material were exposed. Several algorithms are in place to solve CAPTCHAs automatically. The true definition of CAPTCHA is that it must be able to determine that humans, not computers, are attempting to get into a password-protected account. Our work provides an efficient and highly secure alternative to classic CAPTCHA. The objective is to essentially device a two-step “bot-proof” authentication process with camera access as a requirement. According to studies, implementing Multi-factor authentication makes a specific account 99.9% less likely to be penetrated, and similarly building a two-level CAPTCHA would surely improve the security of the user being attacked. The first level involves a Text-CAPTCHA and in the second level, a new CAPTCHA technique that recognizes user hand gestures in real time aids in preventing the possibility of algorithm breaking by attackers. Hand gesture detection was first implemented using Support Vector Machine (SVM) and improvised with Convolution Neural Network (CNN). Finally, Mediapipe assisted and produced faster and accurate results in real time. The authentication will be granted to the website if the human performs the given task successfully in the second level.
