I thought you may like to know (unless I'm the last one to cotton on!) that a scenario we've theorised about on the ASRG mailing list for years is finally here, there's now at least one commercial service that will translate CAPTCHA's for cash.
I found the link (shown below) to a service that will decode captcha's for you for $2 per 1000 successes. (Ironically it was in an ad served to me by gmail.)
So to anyone who ever proposed a spam "solution" that relied on differentiating between people and machines, and doubted us when we told them that the commercial imperative would be its downfall, read this:
We told you so, and it only costs two bucks for a thousand!I predict much more of this in time, with costs falling as more competition enters the market, and I honestly think it presages the end of the useful life of the captcha.
That link: http://www.decaptcher.com/client/
-- correction: Kevin H. politely pointed out, in a comment, that I hadn't read it properly
That isn't $2 per success - it's $2 per 1000 successes. Minimum purchase is $10 = 5,000 spam, er, advertising opportunitiesThanks Kevin, I've updated the post to reflect this.
-- updates: According to ASRG folks
a) This site has been known about since Oct, '09, so yes I am last to the party as ever!
b) Chris Lewis kindly point this out:
... found sweatshops in India quoting some small number of rupees per thousand, claiming to be able to supply up to 250,000 per day. At _least_ three years ago.c) More intriguingly Steve Atkins said:
...
Incidentally, some of the spam filter companies, as part of their technology for trying to figure out whether the jpeg is naughty or not, _claim_ they can solve >90% of all captchas on the fly.
There's even a neural net implemented in javascript as a greasemonkey plugin that'll solve some simple captchas in the browser.I had a Googlearound and found this, that may interest you http://ejohn.org/blog/ocr-and-neural-nets-in-javascript/
I found that particularly fascinating, because last year Nikki was telling me all about cognitive psychology and various theories of word and character recognition in humans. I still have all that to look forward to, in my OU degree, but I'm beginning to get an idea of the kind of things I could do for my project if I ever get that far!
Comments: