Squiggly Words To Catch Comment Spam

Pretty sure Alan Turing wouldn’t care that the test that is his namesake and a staple of science fiction is now as common as Internet spam.

You’ve probably seen CAPTCHAs, those little tests that ask you to type in the word you see squiggled in a little box so you can leave your comment on a blog post or log into your bank’s Web site. (Hypest acronym ever: Completely Automated Public Turing Test To Tell Computers and Humans Apart.) They’re supposed to help sort real human users from bots, those crummy little programs that blast comment pages with links to porn and male enhancement shit sites, hack away at Web sites passwords, and generally make the Web a crappier place to be.

There are lots of ways to fight these bastard bots. I’ve already tried various forms of content filtering, which flags comments that have, say, 18 links to http://popka.007webs.com/421.html in their bodies, only 50 characters, or key words like ’levitra’ and ’lolita’ to pick two completely random L words. That alone has cut down spam on this site by more than 90%. But I bet this isn’t a completely generalizable solution, since Webmasters probably have to tailor their spam filters for their particular needs, which would be time consuming. And it still left a few hundred spammy comments popping up at a time when my site was essentially dormant. They were all over my Daily News pieces, especially the ones about strippers. And this godforsaken article once had about 1000 pieces of comment spam (now down to 132)—and coincidentally, no actual comments. Whack.

Since the only other viable alternative was prayer, figured I could come up with my own CAPTCHA system. The scripting was really easy to build into my comment system, and all I had to do was throw up some images of weirdly blotted phrases (made quickly and easily with Inkscape) that humans can read but stupid, stupid robots can’t. The script randomly picks one of a few dozen pre-made images and sees if the commenter can match the image to the magic word or if IT barely made it through Y2K. It’s been in operation for five hours now. All is quiet and serene.

There are some downsides. CAPTCHAs can be hard to read, even for someone with good (or well-assisted) eyesight and the ability to decipher graff writing. There are the usual problems with fonts other than the old standards, like characters looking alike or unconventional design. And then there are dylsecixs. I’m already good at driving away readers, so hopefully this doesn’t make me better at it.

And hell, even if he wouldn’t give two shits about how he revolutionized comment spam half a century before it existed, Turing would probably be happy about this. So, win-win.

Update | 09.10.12

I’ve been thinking about how Twitter might actually not be as completely and utterly useless as I thought.

Consider that one of the problems of comment spam is anonymity: if you knew who the spammers were, you could block them. Some Web sites, like Mother Jones, require users to log in to comment. That’s kept much of the spam, as well as the knuckle-dragging bullshit from disgruntled conservatives, at bay. But that solution sucks, because it involves remembering one more password and logging into one more system.

But there’s one solution: Use Twitter, or another microblogger, as your commenting system. That means Twitter users can comment without having to log into some other system. And Webmasters can block spammy users. Huffington Post already does this periodically, like in this article.

The biggest spam-related problem for Twitter seems to be in its trending topics feed. But I don’t think that will necessarily affect how legit, individual users can comment on articles. Huffpo seems to be going along fine without Twitter comment spam.

Wouldn’t it be nice if there were several—say, open source—microblogging projects to choose from, but each could be used for the same commenting system?

A little something for the robot fellas

A very effective Turing test would be to ask whether the respondent finds the model with the stick-like figure in this Ralph Lauren ad attractive. (Another good question: Does Photoshop have some sort of "Make pre-pubescent" plugin?) The correct, human answer would be: fuck no. Absolutely no one but automatons could possibly find Ann Coulter attractive.

So who exactly is Ralph Lauren’s target audience, anyways? End.