Recognizing CAPTCHA using Neural Networks

Abstract

CAPTCHAs (Completely Automated Public Turing test to tell Computers and Humans Apart) are distortions of texts or images so that they are still recognizable to most humans but can become difficult for a computer to recognize. Additionally, many CAPTCHAs also inject noise or additional structures such as blobs or lines into the image to further make it difficult to recognize. They are primarily employed as security measures on websites to prevent bots from accessing or performing transactions on the site. On the other hand, machine learning (ML) algorithms, in particular deep learning (DL) neural networks have been trained with significant success on similar problems such as handwritten digit recognition. This motivates us to build an ML-based CAPTCHA breaker that maps CAPTCHAs to their solutions.