No Bot Anticipates The Deep Captcha Presenting Disposed Illustrations With Applications to Captcha Generation

The Completely Automated Public Turing Test to Tell Computers and Humans Apart (CAPTCHA) is a security technique that has been widely used to combat spam, dangerous bot programmes, and automated registrations. An effective CAPTCHA usually presents tests that are easy for humans to accomplish but difficult for computers, with humans achieving over a 90% success rate and computers less than 1%. Thus, a good CAPTCHA is considered both robust and user- friendly. Our research introduces Deep CAPTCHA, an innovative CAPTCHA generation approach designed to deceive deep learning classification algorithms using specially crafted adversarial noise, include other machine learning technologies since hostile samples are portable. To maintain human readability, the noise intensity is kept low, while still defending against removal attacks. The proof-of-concept system developed in this study demonstrates improved usability and security compared to traditional CAPTCHAs. Initially, we will explore well-established CAPTCHA generation research before delving into deep learning and focusing on techniques for creating adversarial instances. By the end of the project, we will assess the resilience of different adversarial example types against preprocessing attacks. The implementation of a proof-of- concept system will also be included, and its evaluation will indicate that our proposed approach offers superior security and significant usability benefits when compared to existing CAPTCHAs.