Skip to content. Skip to navigation
You are here: Home » Forum » Support » Post-mortem on site downtime of 20100614
Document Actions

Post-mortem on site downtime of 20100614

Up to Support

Post-mortem on site downtime of 20100614

Posted by David Hostetler (Admin) at June 15. 2010

This is just an FYI on what happened with the server yesterday...

I'm fairly certain that a brute force ssh attack succeeded in breaching a shell account on the system.  The sshd logs show a successful login after multiple failed attempts.  The account had no home directory, so once inside the attacker would only have had /tmp available to them with write permissions.  I believe (but can't confirm) that they sftp'd one or more tools into /tmp and proceeded to run them in an attempt to escalate the breech.  In particular, when I detected the attack there were dozens of active processes named 'slowdssh'.  There is nothing on the system that provides 'slowdssh' and in fact (amazingly) google knows *nothing* about any such process.

Sadly, in my haste to recover from the attack, I didn't think to archive what was in /tmp, and when I rebooted the system, anything in there was wiped.

As I was telling AJ over chat when it happened, I had temporarily lost my ability to su to root, but that occasionally happens 'normally' when nscd crashes (which it has been prone to do).  After rebooting, I didn't detect anything amiss about the system, including all of the account and credential information (I did a full dump and diff of my ldap db, and there were no changes dating back over a week).

So my hope is that the attack succeeded only in gaining a temporary beachhead, but that the attacker was subsequently unable to do anything permanent.

My evening was then spent setting up fail2ban, which I stupidly had never done.

fail2ban for the win.  It's now happily swinging a heavy ban cudgel on any IP that even looks like it hasn't showered and shaved in the morning.

I also buttoned up my sshd config even tighter than it already was (namely restricting ssh access to a short manual list of usernames).  And I deleted some old shell accounts that I know are no longer used.  The way the ssh brute force attacks work is by trying common usernames and then poking away at any that are discovered to exist on a given system in a (slow) attempt to get lucky with a dictionary password attack.

Re: Post-mortem on site downtime of 20100614

Posted by sammy thiels at June 15. 2010

 AHH now i understand thats exactly what i thought had happened. I was going to repair the server from my cell phone myself but I too lost my ability to su to root and on second thought ,,better let Hoss handel this since I have no idea what so ever what the hell he just said. Hoss I continue to be amaized thanks for everything you do for us..sammy

Re: Post-mortem on site downtime of 20100614

Posted by Jason Weber at June 17. 2010

I have my external ssh going to a VM that just hosts my simple web server.  Hopefully, there is little obvious incentive to make a second jump.  Of the hundreds of ssh requests a day I get, they are almost exclusively for root or without name.  I haven't seen a username sweep in a long time.  I block root and also filter on known places to ssh from.  Also, by using openDNS, I effectively get blocks from their known list of bad places:

 

host name/name mismatch: 189-72-11-242.sance300.ipd.brasiltelecom.net.br != hit-nxdomain.opendns.com

Re: Post-mortem on site downtime of 20100614

Posted by David Hostetler (Admin) at June 18. 2010

Slow or trickle brute force attacks have seen a resurgence the last year or so.  I'm surprised if you're not getting hit.

 

This guy's been chronicaling it pretty well (note the links at the bottom to the previous episodes in attack activity):

http://bsdly.blogspot.com/2009/10/third-time-uncharmed.html

Powered by Ploneboard