Spammers' Upend DNS

Saint Aardvark writes "eWeek reports on the latest trick of spammers: getting around DNS-based lookups. By registering a domain *after* the spam goes out advertising it, they can get around blacklists. However, that causes all sorts of problems for ISPs and anti-spam services. Paul Judge, CTO at Ciphertrust, says "Even in large enterprises, it's becoming very common to see a large spam load cripple the DNS infrastructure.""

Comments

That's not the sky falling...

[winkydink]

The article goes on to say that some anti-spam applications do as many as 30 dns lookups. This is a design problem with the apps, not with DNS. Do less lookups, minimize the problem. I'd venture that after checking with a few of the major blacklists, you've pretty much hit the point of diminishing return in distinguishing spam/ham.

Re:That's not the sky falling...

[Zocalo]

No, it's a problem with spammers making references to multiple domains in their email, each of which might need to be checked against several SURBLs. Personally, I'm not fretting this one at all; while it's an ingenious work around from the spammers to get around the SURBLs, there's a trivial fix.

At the moment, each domain referenced in the body of a spam is checked against one or more SURBLs to see if it has been spamvertised - hence the 30 lookups figure. Instead of immediately checking the SUBLS, we can just make a single check to see if the domain exists at all, if it doesn't then skip the SURBL checks and bias the score towards being spam. If it does exist, then we can proceed to check the SURBLs as normal and still nail any spams using known spamvertised domains. If the domain does exist, then it's a single extra DNS lookup which is possibly going to be cached, so a root server query may be avoided. If it doesn't exist, then we skip the SURBL checks and save our 30 DNS queries.

Yup, it's the old spam arms race again. Give it a month or so and we'll all be moaning about some completely new spammer tactic brought in to replace this one.

Re:Dammit

[punkass]

yeah! and make drugs illegal too! that'll teach 'um.

spam protocol hogging

[Doc Ruby]

DNS could play a role in beating spam. DNS servers suffering from "spam overload" can see that they're handling a lot of the same lookups, that are overloading them. They could flag their responses back to the isolated SMTP servers that are processing the spams, which can tell that they're all the same message. So the distributed network can identify spams, and at least require the senders to share some of the processing load (through another extension to the SMTP and DNS protocols). A more severe response that might affect mere mass-mailers (different from "spam" because content is either noncommercial, or was solicited by the recipient) would be to report such spam-suspects to blacklist servers, which in turn inform users spam filters.

Having had several mass-mailed (big Cc: lists) urgent messages filtered out by corporate spam filters in the past couple of months, I know we need a much better system. Spam is taking down DNS, blocking SMTP, and, even worse, censoring legitimate message needles in the spam haystack. We need network protocols to get smarter, taking advantage of the distributed intelligence that can kill spam. Can the IETF overcome its interest in perpetuating the spam that pays for so much of the Internet, in leading us out of the spam trap?

Negative Caching

[whoever57]

BIND, at least, does negative caching. Surely this means the load on DNS servers due to looking up the non-existent spam domains is minimal.

Also, once the mail server has decided that a bounce reply is undeliverable (because of no DNS records), surely it is going to dump the email immediately, rather than continuning to attempt to deliver it?

So is this a case of SOME brain dead implementaions of DNS and mail servers, or a real problem for all?