After a recent mailing that I sent out to a subset of my proxy mailing list, I got back 18 auto-replies from Hotmail users, all substantially similar to this:

Dear friend:
We are an electronic products wholesale .Our products are of high quality and low price. If you want to do business , we can offer you the most reasonable discount to make you get more profits. We are expecting for your business.

Please visit our website: www.wedosale.com

Email: wedosale@vip.188.com .
MSN: wedosale@hotmail.com .

Looking forward to your contact and long cooperation with us!

Our mainly products such the phones, PSP, display TV, notebook, video, computers, Mp4, GPS, xbox 360, digital cameras and so on.

Welcome to visit our website!

Some of the spam auto-replies advertised different websites, and the wording varied between the different auto-responses, but they were all similar advertisements for Chinese electronics "retailers." (And so, I assume, the websites are all fronts for the same company -- if multiple spammers had independently hacked Hotmail users' accounts to set their auto-replies, it would be vanishingly unlikely that those spammers would all happen to be electronics hawkers.) This was from a mailing that I sent to a set of subscribers that included about 26,000 users with "hotmail.com" e-mail addresses. If 18 out of 26,000 users in my sample have had their accounts hacked to send spam auto-replies, then this must be happening to a large number of Hotmail users -- not a large proportion (only one in 1,500, in my sample), but with about 300 million Hotmail users, that would still be a large absolute number.

The same spammers have apparently been spamming through Hotmail auto-replies for at least 11 months, according to this post in the Windows Live Help community forum from January 2009. At first, some pundits seemed to have assumed that spammers had created these accounts themselves and subscribed the accounts to people's lists, in order to spam the list owners (and, if it's a list that accepts subscriber posts, broadcast the spam to the other list readers). However, looking at the addresses in my proxy mailing list that were sending the spam auto-replies, I noticed that (1) our records show that the auto-reply-spamming subscribers joined the mailing list by various means, signing up through different Circumventor websites, not indicative of how a spammer would have joined the list by automated means, and (2) many of their email addresses are associated with legitimate-looking Myspace and Facebook accounts. Thus it looks as if these were real users who joined the list legitimately, and then got their accounts hacked by the spammers, who set those users' accounts to send the spam as an auto-response.

(If you happened to look at the spammers' www.wedosale.com website, at this point you might be thinking: I don't want to give money to spammers, but can I really get a Blackberry for only $295? Couldn't I just order from the website, and then if the goods don't show up or they're not as advertised, I can dispute the charge on my credit card? Well, I signed up for a dummy account on the www.wedosale.com page and got as far as the order page, and the only payment types that they accept are wire transfer, Western Union, and Moneygram -- precisely those types where you cannot get the money back or dispute fraudulent charges. If you've already gone and ordered a Blackberry, don't hold your breath.)

If my 26,000 users were a representative sample of the 300 million current Hotmail users, then with 1 out of 1,500 users in my sample being "infected," I could estimate that about 200,000 Hotmail users (1/1500 times 300 million) are currently set to send spam auto-replies. Hotmail claims to process 3 billion non-spam e-mails per day, for an average of about 10 non-spam e-mails per Hotmail user. That's the average for all users; what's the average for the infected users? Some factors would tend to lead to a lower average for infected users -- if they have lots of friends sending them mail, it's more likely that one of their friends would have told them about the auto-reply spam and told them to turn it off, so perhaps the users still sending the spams are the ones who don't receive a lot of messages from their friends. On the other hand, some of the infected accounts may be receiving more (non-spam) e-mail than average; one reason people sometimes abandon webmail accounts is that they're getting too much mail, even from newsletters like the Circumventor list that they had legitimately subscribed to. So, figuring that factors in both directions roughly cancel out, if each infected user is receiving the average number of 10 emails per day and sending 10 auto-reply spams in response, that's still a total of 2 million outgoing spams per day shilling for nonexistent Chinese iPhones.

These are just back-of-the-envelope calculations, but even I'm overestimating by a whole order of magnitude, that's still 0.2 million auto-reply spams per day, or about 70 million spams that will be sent by this one company through Hotmail's servers in the coming year, if Hotmail doesn't stop it. (And closer to a billion spams in the coming year if I'm not overestimating.)

And it's actually worse than that, because these spams are less likely than average to be filtered, since they're coming from Hotmail's servers. Normally you'd think that the content-based module of a spam filter would have no problem catching a message like the one at the top of this article, especially if millions of similar messages have been spewed out over the past year. However, messages from Hotmail's servers, regardless of content, are less likely to be blocked, since their network has a good reputation for sending little spam overall (due to measures such as requiring users to fill out a CAPTCHA when signing up, blocking each account from sending more than 500 messages per day, etc.). When I sent messages to the infected Hotmail users from my Gmail account, to see if the auto-responses would get through Gmail's spam filter, Gmail's blocked only half of the replies. When I mailed all the users again from my Hotmail account, the results were strange -- most of the users' accounts sent back no auto-reply at all, not even a reply that got routed to my junk folder. (Why would Hotmail accounts not send an auto-reply in response to a message from a Hotmail user? Please post if you have any idea what's going on there.) However, of the infected Hotmail accounts that did send a spam auto-reply, 100% of those auto-reply spams were delivered to my inbox. (Apparently, Hotmail's spam filter usually assumes that messages from other Hotmail users can't possibly be spam.) Only Yahoo Mail's spam filter, when I sent a test message to the infected users from my Yahoo Mail account, blocked all of the auto-replies as junk mail.

For the infected users on my mailing list, I sent them a link to a set of instructions I'd written about how to set and un-set their Hotmail auto-reply and how to change their Hotmail password, with the hopes that they'd eventually see the message and follow the steps. 18 users rescued, 200,000 to go.

So this is basically what's happening, but it still leaves some unanswered questions, such as: Why Hotmail accounts, but not Yahoo Mail, GMail, or AOL accounts? I've never noticed any auto-reply spam sent from any accounts at any of those other services. Whatever the spammers did to gain control of so many Hotmail accounts, if it was profitable for them, why didn't they do the same thing for Yahoo Mail? And, why did only one spammer do this? If they're sending between 1 and 10 million spams per day for free, they're probably making money at it. Whatever they did to hack those accounts, why wouldn't other spammers figure out the same method and copy them?

Presumably the Chinese spammers stole large numbers of passwords from Hotmail users either via a huge phishing attack, or through a security hole in Hotmail or some other part of the Windows Live service. If it was done via a security hole in Hotmail that the spammers discovered, then that would explain why the spammer's methods only worked for Hotmail accounts, and also why no other spammers have copied their techniques. (A phishing attack, on the other hand, would be easy to modify for other webmail services, and would also be easy for other spammers to emulate, so that's not consistent with the observed evidence so far.) I also found this post from blogger Stuart Shelton describing how his account was hacked by Chinese spammers -- and from the blog post, it's clear that he's very tech-savvy and would have been unlikely to fall for a run-of-the-mill password phish. If the attack happened even to people who know what they're doing, that seems to make the security hole explanation more likely.

Perhaps others can come up with some theories about what happened. It's easy to come up with guesses, but the hard part is to reconcile them with the fact that it has only affected Hotmail users so far, and no other spammer seems to have figured out how to copy the same technique yet.

But there's a much simpler question too: Why doesn't Microsoft just turn off the auto-replies for these users' accounts? They can query to see exactly which users have these messages in their auto-replies, and then un-set the auto-reply automatically. Yes, I know that even for a simple database operation like that, there's always more to it when you're managing hundreds of millions of accounts across multiple servers -- but if it will stop this one sender from sending between 50 million and 500 million spams (that in many cases will bypass people's spam filters) from Hotmail's servers in the coming year, isn't it probably worth it?

And even if it wasn't a phishing attack this time, sooner or later some other spammer will probably capture tens or hundreds of thousands of Hotmail accounts using a phish or some other method, and try spamming through auto-replies as well. So if Hotmail "fixes" this batch of auto-reply spam for practice, then the next time it happens, they'll know exactly what to do to take care of it.

I've written some columns where I strongly believed every word but expected a lot of opposition, some where I wasn't sure if I was right and just wanted to see what people thought, and . But I rarely argue something that I think is a no-brainer. Hotmail should un-set the auto-replies for those users whose accounts are spamming for nonexistent Chinese electronics knockoffs, before those accounts send another several hundred million spams in the coming year. Am I smoking crack?

Then again, maybe expectations for Hotmail shouldn't be set too high. I use SpeakEasy for my mail provider, and on about November 19th I found that all messages sent to hotmail.com addresses from SpeakEasy's servers were being bounced with an error message rejecting them for "spam-like characteristics."I called SpeakEasy and they confirmed that they knew Hotmail was blocking all mail from their users (although for "security reasons," SpeakEasy couldn't tell me what they were trying to do about it). The block wasn't lifted until about November 28th, when my messages started getting through again.

If SpeakEasy, which has been in business for 15 years, has annual revenues of $60 million, and was bought in 2007 by Best Buy, can't even get through to Microsoft in less than 10 days to tell them to stop blocking all mail from their servers, then Microsoft should first fix their postmaster trouble ticket system, so that people are not blocked from writing to their friends and family members at Hotmail for a week and a half. Then get to work on the spam auto-responders.