WE have kept silent about it in order not to encourage the attacker/s, but it's true. We have been under heavy DDOS attacks since Thursday night. What has happened since then? Well, a lot. Our previous host is no more as far as we are concerned. After struggling with the botnet for like 10 hours (filtering to no avail) our Web site got isolated. It did not serve any pages for almost 2 days. A reader of Boycott Novell was kind enough to lend us room on his server (more or less dedicated), on which he fought the botnets for over a day. The attackers kept changing tactics. Some other readers offered filtering advice and we are grateful to all of them. Ultimately, the attacks halted yesterday afternoon.

“Ultimately, the attacks halted yesterday afternoon.”The migration from the old server was not simple because the site was disabled abruptly following the early attacks. But now we have ensured that all data has been migrated. The only 'good' thing which came out of this attack is that, as oiaohm put it, the ordeal sort of made us more robust to future attacks.

Now that we have a new host in place, we also have more features. Data on the site (comments, posts, etc.) was not lost in the migration, just heaps of time and effort affecting several people. We have moved to a bigger, more robust environment that will hopefully facilitate the needs of the Web site as it continues to grow (we served about 200GB of data last month). We apologise for the downtime, which is unprecedented.

The plan is to carry on exposing Comes exhibits next month and also organise the Wiki. There is enough for years of work.

Again: we would like to thank all those who helped during the downtime and especially our generous reader ( Copilotco) who offered to host the Web site, taking us away from shared hosting in the process. Dedicated servers on normal Web hosts are just far too expensive for us to afford and I swear that I never made a single dime from this Web site. The ads merely covered the hosting fees which Shane has been paying since 2006.

One last clarification for lunatics who are now suggesting that we DDOSed ourselves, where to even begin refuting such nonsense (coupled with personal abuse)?

The attacks came from many addresses, for example 88.198.60.8 which is "tor-proxy.va6.de". Multiple such IPs hit us constantly and relentlessly (all tor exit nodes at first). At one stage it seemed like the front page alone received 3 page requests per second. But the IPs were also doing a HEAD on the Web site as many times as possible, bringing the server down to its knees (both the old server and new server, the former running Red Hat and the latter CentOS). ⬆

Update: Here is a report from the administrator.

I took over hosting of boycottnovell.com for Roy in the middle of the DDOS attack. I am looking at the squid log for boycottnovell.com during the DDOS. I have squid caching/proxying/url-rewriting for apache for various reasons.

The attack initially (or at least, at the time the DNS was re-pointed to my server) consisted of lots of HEAD requests. Then I wrote up a script to tail the log finding anyone doing lots of HEAD requests and putting the offending IP into the iptables packet filter while I cooked up a more permanent solution. Eventually they figured this out and switched to a full on GET of the root of the site and then I think they started getting random pages from the site as fast as they could although I'm not sure about that.

The interesting part starts around timestamp 1242543590.804 which is apparently when most of the world's DNS cut over to me including that of the machines in whatever bot net was employed in the attack.

If we run this command on the logfile with the logfile being /tmp/bn.log:

grep " HEAD http://boycottnovell.com/ " /tmp/bn.log | awk
'{print $3}' | sort | uniq -c |sort -n | tail -10

we get:

   2716 81.175.61.4
   2960 212.24.147.228
   3056 204.209.56.56
   5637 87.236.199.73
   6645 145.100.100.190
   7261 212.42.236.140
   8487 88.198.14.120
   9640 62.141.58.13
  11008 87.118.104.203
  11269 88.198.60.8

and if we do:

grep " GET http://boycottnovell.com/ " /tmp/bn.log |
awk '{print $3}' | sort | uniq -c |sort -n|tail -10

we get:

   5801 94.136.16.242
   5854 85.25.152.185
   5865 212.24.147.228
   6367 66.35.1.170
   6682 205.209.142.210
   6977 87.118.104.203
   8102 83.140.125.188
   8300 85.25.145.98
   8441 212.42.236.140
  20065 66.230.230.230

So one IP did a get of the root of the site 20k times before I really effectively got everything blocked off and another did a HEAD around 11k times. You can get a feel for how the attack progressed using:

egrep ' GET http://boycottnovell.com/ | HEAD
http://boycottnovell.com/ ' /tmp/bn.log | less

Assuming that everyone who did a GET or a HEAD more than 100 times (a conservative estimate) is involved in the attack:

egrep ' GET http://boycottnovell.com/ | HEAD http://boycottnovell.com/ '
/tmp/bn.log | awk '{print $3}' | sort | uniq -c| sort -n > /tmp/attackers

and then counting only the lines with greater than 100 hits we can see that there were 281 unique IP addresses involved in the attack.

However, it looks like they switched to targeting various different parts of the site later on or maybe just random pages because if we look at all of the accesses to the site which made more than 100 requests we get 863 IPs involved the top 19 being the following:

   6193 62.141.53.224
   7153 85.25.151.22
   7764 145.100.100.190
   8524 66.35.1.170
   8757 94.136.16.242
   9256 85.25.152.185
  10369 83.140.125.188
  10464 212.24.147.228
  10874 205.209.142.210
  10935 87.236.199.73
  11441 88.198.14.120
  12094 62.141.58.13
  12208 88.198.60.8
  12994 66.249.70.134
  13940 85.25.145.98
  19119 212.42.236.140
  19867 87.118.104.203
  26480 216.105.40.113
  29854 66.230.230.230

So 66.230.230.230 made 29k requests to the site in total.

Putting some iptables rules in place (which I document here):

http://www.kernel-panic.org/pipermail/kplug-list/2009-May/108075.html

nicely cut the problem down to size and now the effect of the DOS is unnoticeable.

11M of gzipped log are used for this sample.