Designing New CAPTCHA Models Based on the Cognitive Abilities of Artificial Agents

A CAPTCHA is a program that allows or denies access to services by generating and grading test that humans can pass but current computer programs cannot. Captchas are generally used to protect free web services from automated programs. Also, they can provide an idea of which fields in artificial intelligence are the most explored and which could be researched in the near future. Most of the tests that are based on text recognition have been broken by optical character recognition (OCR) techniques while those based on images are vulnerable to machine learning attacks. Humans make use of cognitive abilities to carry out tasks in daily life, even if they are not conscious which ones or how they use them. The current state of technology is still not enough to reproduce some human cognitive abilities, and the identification of those abilities is the basis for the design of new CAPTCHA models. In this paper we introduce seven new models of CAPTCHA to test some cognitive abilities that are supposed to be beyond the capabilities of artificial agents. We use some APIs to proof that images used in the proposals are extremely difficult to be recognized by artificial agents.