PDA

View Full Version : relax please, Botnets are easy to detect and block for SMALL domains


CamFraser
11-22-2006, 07:14 PM
There seems to be some needless worrying going on all over the net about recent reports of the surge in botnets.

Don't Panic! (in large Friendly letters) :P

Relax please. All those reports focus on problems restricted to large scale providers.

From a technical point of view, the problem was identified, analyzed, and solved in principle a year ago. The botnet spam volume has been insufficient for most to bother implementing those solutions.

The caveat is that many of those solutions work best if you have your own small scale domain.
Hmmmm... anybody here to whom that description applies? :wink: ( :cough: everybody? :cough: )

As I mentioned in this anti-spam thread (http://www.aota.net/forums/showthread.php?t=22121) one solution is being called the "Jayne Cobb commemorative Grenades test" (a tribute to the superb science fiction TV series Firefly (http://www.news-journalonline.com/column/247/03SceneTWEN072005.htm)). :yeah:

It uses the fact that small domain owners know which accounts are in use, and the botnet spammers are demonstrably stupider than average in that they have huge quantities of invalid addresses in their DBs. Basically, all you need to do is turn your catchall on, and use a filter that tracks stuff that is near simultaneously sent to at least one invalid account (or whatever cutoff number you're comfortable with).
If an IP only sends to valid accounts, the pin stays in the grenade, otherwise, kaboom! :P

I regularly run a "grenade simulation tool", so I know this will work. Every other solution my anti-spam provider has come up with has also worked wonderfully, so I have a rational basis to trust that track record.

If you're like me and often generate unique one-off account names for site regs, not a problem. Set your "grenade dispersal" to require at least two account hits. A legit sender is almost never going to send two or more simultaneous messages from the same IP to two separate non existent accounts (or one real and one unreal). Main legit case would be if you forgot about a rarely used legit account name, or two people on the same domain gave out ad hoc account names to the same (legit) site. Even in those rare cases, that sort of email is low priority, and in my view not a loss. Besides, the anti-spam suite (http://Puffin.net/software/spam/index.htm) I use (as mentioned in that other thread), maintains a "corpse" pile, and has tools designed to easily spot and salvage FPs.

That's just the simplest level of detection complexity. Next level up is an RBL powered by catchalls! Think about how trivial it would be to redirect your catchall to one server, and how easy it would be to chomp thru that data in real-time, statistically ranking each sending IP. In other words, turn the tables on the spammers! That's kiddie level to do. It just needs enough people who are serious enough to kick in their time and expertise, and a larger group to kick in the money to fund a dedicated server and bandwidth. I've already made a modest pledge of time (1 hour/month) and money ($20/year), and I'm optimistic there's enough others interested to make that possible.

Why haven't those approaches been used so far?
Because anyone using a sophisticated anti-spam tool has barely noticed the spam events of the past month.

Bottom line: relax!
The moment the botnets really become a serious problem, there's well tested solutions waiting for deployment.

Most technical problems can be solved. It just takes enough serious people to get the job done. In the words of Tolkien: "If we all got angry together something might be done." (Of course, he meant "productively" angry - decent people don't waste energy getting angry if that anger isn't matched with action.)

CamFraser
12-11-2006, 08:55 PM
Was kinda rushed the other night, so just wanted to expand a bit...

My post above was (almost) verbatim what I posted in another forum, except for the "local content" (links and cheerleader bottom part). Like I said, there's been chatter about this phantom menance for weeks now, not just here.

If a few of you banded together, you could implement the simple "grenades" filter I described, or you could hire someone at one of the fixed price bid sites. Feel free to quote from the algorithm I described (that's my own summary of someone else's WhitePaper detailing it, with some obvious quotes from the author). The people who've reported that they have implemented grenades, say it takes a half day or less to code and test. For a good local programmer, that'd be in the $200 to $1000 range. If that's too much, how much does it cost to rent 4 hours of a good "offshore" programmer's time? $20? $40? $100?

It's easy to simulate what sort of results you would get from applying either the "grenades" or RBL concept:

turn on your catchall for a week
if you can find one, install a global simple filter that captures the originating IP, and SMTP recipient
(look around here for existing scripts - Sheila and DB's excellent EFM (http://diamond-back.com/software/efm/) does logging and extracts both IP and recipient, so it should be practical to modify it)
analyze the data using a spreadsheet or other tool

Older, heavily compromised domains should get good results from a standalone filter. Younger, less compromised domains would get much better results from an RBL implementation, due to the collaborative nature of the data feed.

Anyone who's interested in that anti-botnet RBL project I mentioned, and who has a decent track record, is free to email me if you'd like in on it (warning: their cutoff date for Alpha testers is soon). This is the same group of former AT&T Labs engineers I mentioned a few months ago. The #1 need is donations of catchalls, ideally on domains at least 2 years old with a static IP. One of the reasons I brought it up at FQ is that it, unlike many webhosts, does provide a static IP.

If there seemed to be interest, I was planning to post a link to the project recruiting page, but the only feedback I've received has been strongly negative. I will respect those expressed wishes, and cease discussing this project. :rolleyes:

For the rest of you, best of luck to anyone who chooses to take on this challenge! :yeah:

Jeff
12-13-2006, 04:30 AM
The #1 need is donations of catchalls, ideally on domains at least 2 years old with a static IP. One of the reasons I brought it up at FQ is that it, unlike many webhosts, does provide a static IP.
Note that FutureQuest does not provide static IP's to the older accounts that were setup on shared IP's years ago when namebased hosting was promoted and requested by ICANN as more "internet friendly" to conserve available IP addresses.

(although I think all hosts host on "static" meaning non-dynamic/non-changing IP's -- just some IP's are used for many sites in namebased hosting vs. one IP per site in IP based hosting.)