Abstract
CAPTCHAs are a widely deployed mechanism for ensuring that a web site user is a human, and not a software agent. They ought to be relatively easy for a human to solve, but hard for software to interpret. Most CAPTCHAs are visual, and this marginalises users with visual impairments. A variety of audible CAPTCHAs have been trialled but these have not been very successful, largely because they are easily interpreted by automated tools and, at the same time, tend to be too challenging for the very humans they are supposed to verify. In this paper an alternative audio CAPTCHA, jCAPTCHA (Jumbled Words CAPTCHA), is presented. We report on the evaluation of jCAPTCHA by 272 human users, of whom 169 used screen readers, both in terms of usability and resistance to software interpretation.