System for beating CAPTCHA using artificial intelligence
Rishabh Dalal · UNI ScholarWorks (University of Northern Iowa) · 2019
A CAPTCHA (completely automated public Turing test to tell computers and humans apart) is a type of challenge-response test used in websites to determine whether the user is a human or not. CAPTCHAs were invented to prevent computer bots from spamming websites. With the advent of online businesses, CAPTCHAs became more and more important and are now employed by almost all websites. For example, Paypal uses CAPTCHA before any money transfer to make sure that the user owns the profile, Facebook limits the creation of fraudulent profiles which are used to cheat people, and Reddit prevents bots from flooding the webpage with directed messages. Although CAPTCHAs play a crucial role in the web, as computer hardware and artificial intelligence techniques evolved, artificial intelligence gained the ability to beat CAPTCHAs. If we know what kind of algorithms can beat it, only then can we prepare next-generation CAPTCHAs that are more resilient. This project focuses on devising and implementing various artificial intelligence, image processing and procedural programming techniques to create a system that solves CAPTCHAs with significant accuracy. It tackles the most widely used CAPTCHA 1 scheme which uses a combination of distorted characters that humans can recognize but that may be difficult for automated scripts.