Re: any audio CAPTCHA use in Voice Applications accessed by phone?

Hey Al,

Most information that is available by phone is already available via the
web.  Considering the complexity of automated phone usability and stress
testing (aka the hammer) compared to the ease of automated web testing
(canoo, junit) it is significantly easier even with a captcha for
information to be gleaned via a visual interface than an audio one.  Any
information captured via the phone would then have to be transcribed into a
usable form, since audio isn't the most appropriate way to use the ill
gotten information.

The most important step to insulate yourself from automated harvesting would
be to ensure that your vxml web application is only available via a phone
based interface and not accessible via http to the internet at large.  If it
must be accessible, then the type of captcha to use might be embedded into
the vxml via javascript, but would be seemless to the phone caller and only
necessary to figure out how to fetch the next page. (the captcha would
prevent automated web based harvesting of your application just like a html
interface, but designed in such a way that anyone using a phone based
browser wouldn't notice it.)

Regards,
Shane Smith

On 7/14/07, Al Gilman <Alfred.S.Gilman@ieee.org> wrote:
>
>
>
> An ongoing accessibility problem has to do with robot harvesting of
> free services vs.
> disability access to those services.
>
> http://www.w3.org/TR/turingtest
>
> One of the current strategies is to make optional CAPTCHA
> transactions available
> that depend on different human functions: one you need to see, another you
> need to hear, one you need to work logic puzzles, etc.
>
> It would be helpful for the 'universal' interaction design if there
> were any audio-only
> audio-CAPTCHA practice in the wild to benchmark.
>
> Is anyone here aware of voice-enabled or other IVR applications that
> use a muddy
> prompt to weed out robot callers?  Or is this unnecessary because the
> phone
> call to access the IVR in the first place is enough of a barrier
> compared with a
> TCP/IP connection?  Or it's useless because of the prevalence of good
> speech
> recognition in this domain?
>
> Al
>
> PS:  Compare with the discussions of the under-development reCAPTCHA audio
> CAPTCHA.
>
> http://tinyurl.com/ywyqs2
>
>