Skip to Content

Visual Verification: Study Seeks to Expand the Usefulness of Audio CAPTCHA

September 1, 2007 • Darrell Shandrow Hilliker

Andy Schlaikjer, a Ph.D student at Carnegie Mellon University, has asked us to carry the following announcement:

I’m conducting a study to aid my research and development of a new form of audio CAPTCHA. If you’d like to participate, please visit the Audio reCAPTCHA study web site.

A CAPTCHA is a special kind of test which can be used to tell humans and computers apart. Many web sites use CAPTCHA’s to combat fraud and automated access to their services. Unfortunately, most CAPTCHA’s are based on a visual task, such as recognizing distorted letters in an image. Such a task can be quite difficult, or impossible, for visually impaired human users to perform.

In an attempt to alleviate this accessibility concern, audio-based CAPTCHA’s have been developed which require users to listen to and transcribe a short audio clip containing a series of random spoken digits. However, performance of state-of-the-art Automatic Speech Recognition technology suggests that this approach may not represent a very strong CAPTCHA in practice. Additionally, the data collected from such a test may only be used to determine the authenticity of the user, and is normally discarded once the test has been performed.

The goals of my research are (1) to develop a stronger form of audio CAPTCHA, (2) create a CAPTCHA which collects useful data, and (3) to strengthen support and adoption of audio-based CAPTCHA’s on the Web. To these ends, I am developing a new audio CAPTCHA based on a more complex task: Transcription of arbitrary speech. For more information, please contact me, or visit the study web site at the URL mentioned above.

Cheers,
Andy Schlaikjer

While we appreciate the new found consideration of accessibility by the people at Carnegie Mellon University, with respect to CAPTCHA, and recognize that audio CAPTCHA is the current state of the art, the considerable ongoing research in this area ought to bring all of us to one concern, which we must ultimately address. Audio CAPTCHA, like its visual cousin, inherently denies access to the deaf and hearing impaired population. This means that the presentation of both an audio and visual CAPTCHA continues to lock out those people whom happen to be both blind and deaf. It seems to us that greater focus ought to be placed, instead, on the development of a highly secure, non-sensory challenge response system that does not inherently discriminate against any legitimate human being, regardless of disability.

Since the reCAPTCHA team has taken considerable steps to improve the accessibility and usability of their current audio CAPTCHA scheme, let’s all help Andy with his study. At the same time, let us all remind CMU and others that, in the long run, audio and visual CAPTCHA does not afford equal access and full participation to all human beings. Instead, it is absolutely critical that a better method of authentication and authorization be devised.

Leave a Reply

Your email address will not be published.