Russian serfs paid $3 a day to break CAPTCHAs
Semi-automated attack or chain-mail gang
Why should miscreants bother to develop cutting edge programming techniques when they can pay $3 to somebody to set up spam-ready webmail accounts on their behalf? Evidence has emerged that people as well as malware are being used to defeat CAPTCHAs, challenge-response systems that are often used to stop the automatic creation of webmail accounts by spammers.
CAPTCHAs typically help ensure that online accounts can't be created until a user correctly identifies letters depicted in an image. The tactic is designed to frustrate the use of automated sign-up tools by spammers and other miscreants.
Over recent months security firms have reported that first the Windows Live CAPTCHA used by Hotmail, and later the equivalent system at Gmail, have been broken by automated attacks.
Obtaining a working Gmail account has a number of advantages for spammers. As well as gaining access to Google's services in general, spammers receive an address whose domain is highly unlikely to be blacklisted, helping them defeat one aspect of anti-spam defences. Gmail also has the benefit of being free to use.
An analysis of spam trends in February 2008 by net security firm MessageLabs revealed that 4.6 per cent of all spam originates from web mail-based services.
The proportion of spam from Gmail increased two-fold from 1.3 per cent in January to 2.6 per cent in February, most of which spamvertised skin-flick websites. Yahoo! Mail was the most abused web mail service, responsible for sending 88.7 per cent of all web mail-based spam.
The idea that automated tools have been used by spammers to set up these webmail accounts has become, if not the conventional wisdom, then at least a working hypothesis in security circles of late.
However a senior engineer at Google has stepped forward to cast doubt on these reports.
Brad Taylor, a Google software engineer, said internal evidence suggests that low-paid laborers in third-world countries (rather than compromised PCs) are being used to register accounts that are subsequently used to send spam.
"You can see it is clearly done by humans," Taylor told the New York Times . "There are patterns in the rate we find bogus accounts, like at night time and when people get off work" in particular locations around the world.
Taylor conceded that software might be used to partially automate the process - with bots signing-up for accounts before sending the puzzles to real people - but maintains that the CAPTCHA process remains effective.
Google's contention that low-wage workers are been paid to break watchers is supported by anecdotal evidence unearthed by Websense, which has been active in researching the issue over recent months. The firm found Russian language documents instructing modern day serfs on the art of CAPTCHA breaking.
"If you are unable to recognize a picture or she is not loaded (picture appears black, empty picture), just press Enter. In no case do not enter random characters! If there is delay in downloading images, exit from your account, refresh the page and go again," the documents, found on a website and translated into English, state.
The documents go on to say that CAPTCHA-busters are paid a minimum of $3 a day.
Even if miscreants need human involvement in breaking CAPTCHAs right now this might not always be the case. The solutions to solve puzzles might be fed back to make CAPTCHA-busting algorithms smarter, MessageLabs warns. It said that focusing on whether miscreants are an algorithm, a 'mechanical turk', or combination of the two to break CAPTCHAs misses the bigger point that the approach no longer provides a reliable security mechanism to protect email services from abuse. ®
A lot of email spam is sent via compromised machines on botnets with their own SMTP server. Check the headers of a couple and towards the bottom you'll probably see a received line with some kind of DSL host. There will probably be a couple of fake received lines as well. Personally I quite like the idea of ISPs keeping an eye out for botnet style behaviour or open relays and blocking traffic from potentially compromised hosts until the owner is made aware of what's happening and either agrees to do a virus check or explains why they need to send thousands of emails to random addresses. It would certainly be more useful than booting people off just for using Bit Torrent.
As for web mail accounts, a forum I admin gets about 40 - 50 spam registration attempts a day. I have various blocks in place so none are successful but I do get to see which domains they use for registering. A lot of these are free webmail services such as GMX or Gmail. I think a lot of the addresses are never actually read (although the XRumer forum spamming software does include something that can "process" verification emails) and are used because they're not likely to be blocked. The purpose of forum spam is linkspamming. A forum member list is just a collection of links. Get enough links pointing to the same site and it scores highly on Google. Search for one of the main spammy products and you'll probably find a memberlist.php or a vbulletin /members/ quite high in the results. Breaking the Google CAPTCHA means the spammers can also use Blogspot/Blogger for linkspamming.
As I mention above, I do think the only way to stop spam is through economic means, but I think this needs to take place at a higher level than just educating end users. Pump & dump spam might not be so popular if the shares on penny stocks were automatically suspended if certain "suspicious" activity was detected such as a sudden massive increase in the number of shares changing hands. OEM software spam would probably reduce if the software publishers found out who was selling it and took steps to stop it.
4.6%??? Then... SO WHAT?
OK we all seem to have missed the important statistic here. Only 4.6% of SPAM is sent by web-based mail accounts. So - the other 95.4 is being sent by other means; for instance, by hijacking a poorly protected SMTP server. I think a far better way to make a real impact on SPAM (well, the 95.4% anyway) would be to bring out a newer version of SMTP (SSMTP??) which incorporates certificate-based authentication between mailer and mail server, and between sending server and recipient server. The certificates for the "Mailer" would be tied to an individual and would therefore make hijacking a mail server totally impractical - because of course it would only send email from senders whose certificates it knows. Yes, the costs to provide email would ramp up... but lets face it who actually uses the "free" web-based interface for their email anyway? I always use Outlook to gather and send my web-based email - I only ever log into the web interface to check my spam folder for any legitimate messages before emptying the bucket. I don't use my ISP's mail service much because it has no filtering and because it gets blacklisted a lot (some of Virgin Media's servers do, anyway). The upshot being that if SSMTP came in and free web-based email disappeared, I would simply bring my email activities back to my ISP's offerings and live with the shortcomings.
your 2 pence
You mean like they (sort of) did with the invites?
Bring 'em back, I say!
Alright, world+dog had an account anyway, but having an address through an invite only service added a little... something, I think.