Random TypeCon2007 thought -- spam sucks!

by Michael S. Kaplan, published on 2007/08/03 21:01 -04:00, original URI: http://blogs.msdn.com/b/michkap/archive/2007/08/03/4214516.aspx

On the first talk this morning, Kevin Larson did a great talk entitled Microsoft's Ugly, Distorted Text that covered (among other things) several studies related to enhancements to HIP/CAPTCHA type solutions an the types of variables that seem (under current level of spammer technological progress) to be more effective at foiling automated schemes while at the same time not questioning the humanity of people.

And then later in the day Bill Hill did a talk entitled The Digital Renaissance where he mentioned (among other things) about how having text that available immediately in accessible formats for the blind and sight-impaired.

In my mind I was struck by the fact that both of these people are widely connected to ClearType and other advanced reading technologies, and that even though Kevin's talk really was covering work that was far afield from his usual stomping grounds (making text more readable, not less!), there seemed to be more than a little bit of diametric opposition between the two goals of a more effective HIP/CAPTCHA type solution for spam and services for those with accessibility needs being immediately available.

Now this is kind of something you would expect to be contradictory -- anything that can be automated for use by a screen reader can be automated for use by a spam technologist (or whatever title they use).

But it is depressing to think that one of the net effects of the nine billion pieces of spam that get sent out a day (Kevin's figure, listed here citing him!) might well be technological advances that set back accessibility efforts.

As someone who used to be able to run who has trouble walking these days, I tend to think of most of my abilities (such as sight!) as being temporary accidents that can be lost at any time (whether over the course of many years or the course of just a few TV show episodes like the fictionalized account of Mary Ingalls).

We need a better solution here than one whose goal is to hurt the efforts of screen readers.

ADDENDUM 6:27pm: Wikipedia has some text related to Accessibility within CAPTCHA that are relevant here.

Why do you not like spam? It's delicious.

Mmm... Fresh Spam Sandwich. BOING! (http://xn--sailor-183m.com/japan/boing.jpg)

I love a good Freshness Burger! :-)

We've had a lot of success, actually, with those question-based CAPTCHAs ("what colour is the sky?" etc). The thing is, you don't just have one question, you have a lot of them and choose one at random. If one of the questions is cracked (or even all of them), simply change the question -- that's much simpler than changing your CAPTCHA image generator...

I'm finding so much of the CAPTCHA stuff harder to read now, and never seem to get 0 vs O, or 1 vs l correctly the first time.

On a related note, I think we'll all need screen readers soon. One of my biggest problems with Vista is the low contrast of so much text from its background. Too much cyan on gray or similar,so that I really have to wear glasses much more whenever I use Vista.

I remember these issues being bugged a bit on Vista but usually returned "Not Repro" by someone with presumably 20-20 vision and large hi definition screen.

There seems to be an easy way to provide CAPTCHA to people who use screenreaders. Simply provide the same number in mp3 format. It would be great if web browsers could standartize on an IMG attribute similar to ALT which would channel digital sound information directly to screenreaders.

Well, it's true that this still leaves many impaired people outside; it's true that (this blog being a very appropriate place to mention it) if the numbers are recorded with British accent, people in Texas will never be able to understand it, it's true that SPAM SUCKS no matter how hard we all try fight it... But still, there are ways to fight!

The MP3 option would probably require some musical or other "noise" in the background, otherwise spammers' speech reco software would decode it faster than the images.

Now that gives me pause to think. Some days I could hardly pass a Turing test myself, yet it is becomming a matter of survival for spammers to do so. Anti-spam measures could very well be the selective pressure that leads to machine consciousness.

The best way to handle captcha I've seen to date is where you are asked a question. Sometimes multiple choices:

Which is not a color:

1) red

2) blue

3) Microsoft Internet Explorer

and you have a box to copy the correct answer into.

Or simple one like:

Write "four" as a number into the text field.

Of course you need to have a database which gets updated for it to work.

Please consider a donation to keep this archive running, maintained and free of advertising.
Donate €20 or more to receive an offline copy of the whole archive including all images.