CAPTCHA Using Strangeness in Machine Translation

Abstract

CAPTCHA is a technique that is used to prevent automatic programs from being able to acquire free e-mail or online service accounts. However, as many researchers have already reported, conventional CAPTCHA could be overcome by state-of-the-art malware since the capabilities of computers are approaching those of humans. Therefore, CAPTCHA should be based on even more advanced human-cognitive-processing abilities. We propose using the human ability of recognizing “strangeness” to achieve a new CAPTCHA. This paper focuses on strangeness in machine-translated sentences as an example, and proposes CAPTCHA using Strangeness in Sentences (SS-CAPTCHA), which detects malware by checking if users can distinguish natural sentences created by humans from machine-translated sentences. We discuss possible threats to SS-CAPTCHA and countermeasures against these threats. We also carried out basic experiments to confirm its usability by human users.