Re: ipfilter traffic blocking and tcpdump snort etc

2003-12-05 Thread horio shoichi
On Fri, 5 Dec 2003 00:24:12 +
Jez Hancock [EMAIL PROTECTED] wrote:
 Hi,
 
 I've blocked a dozen or so addresses using ipfilter:
 
 block in quick on fxp0 from 208.186.60.116 to any
 block in quick on fxp0 from 216.230.149.11 to any
 
 etc
 
 but I still see a lot of traffic those hosts in trafshow, snort and
 other packet capturing utils.  Why is this?
 
 Is there any alternative method of blocking access from certain hosts
 so that this traffic is not 'seen' by higher level /userland apps?
 
 As background, the blocked hosts were part of a denial of service attack
 which has been going on for a few hours now.  The attack was aimed at
 port 80, although an odd artifact is that no httpd log entries were made
 for any of the hosts attempting to connect on port 80.
 
 A cursory nmap scan of a few of the hosts shows that all hosts had both
 port 25 and 80 open, but none of the hosts accepted connections on
 either of those ports.  Any idea what kind of attack this could be?
 
 -- 
 Jez Hancock
  - System Administrator / PHP Developer
 
 http://munk.nu/
 ___
 [EMAIL PROTECTED] mailing list
 http://lists.freebsd.org/mailman/listinfo/freebsd-questions
 To unsubscribe, send any mail to [EMAIL PROTECTED]
 

You are probably seeing the supposedly blocked packets on the outside of
network. Observe them on inside, i.e., on the interface not fxp0.

What you are seeing are packets ipfilter is just about to handle.


I don't understand your second question. Are you thinking about tcp wrapper,
reset feature of snort, etc ?



horio shoichi

___
[EMAIL PROTECTED] mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-questions
To unsubscribe, send any mail to [EMAIL PROTECTED]


Re: ipfilter traffic blocking and tcpdump snort etc

2003-12-05 Thread Jez Hancock
Hi Horio,

Cheers for reply.

On Fri, Dec 05, 2003 at 07:33:49PM +0900, horio shoichi wrote:
 On Fri, 5 Dec 2003 00:24:12 +
 Jez Hancock [EMAIL PROTECTED] wrote:
  Hi,
  
  I've blocked a dozen or so addresses using ipfilter:
  
  block in quick on fxp0 from 208.186.60.116 to any
  block in quick on fxp0 from 216.230.149.11 to any
  
  etc
  
  but I still see a lot of traffic those hosts in trafshow, snort and
  other packet capturing utils.  Why is this?
 
 You are probably seeing the supposedly blocked packets on the outside of
 network. Observe them on inside, i.e., on the interface not fxp0.
Not sure what you mean here, what command would you issue via tcpdump or
snort to do what you suggest?  

 What you are seeing are packets ipfilter is just about to handle.
Right - it's just I would have thought that ipfilter handled packets
before they reached any traffic dumping utils.  I see what you're
getting at.  Presumably snort for example uses the bpf driver via pcap(?) to
capture network traffic...

actually reading bpf(4) clears things up a little:

 Associated with each open instance of a bpf file is a user-settable
 packet filter.  Whenever a packet is received by an interface, all file
 descriptors listening on that interface apply their filter.  Each
 descriptor that accepts the packet receives its own copy.

  Is there any alternative method of blocking access from certain hosts
  so that this traffic is not 'seen' by higher level /userland apps?
 I don't understand your second question. Are you thinking about tcp wrapper,
 reset feature of snort, etc ?
Let me rephrase that one :P  I meant is there a method - for example
such as adding some kind of routing via arp - so that packets are
dropped on the floor even quicker than they would be via the firewall
method?

-- 
Jez Hancock
 - System Administrator / PHP Developer

http://munk.nu/
___
[EMAIL PROTECTED] mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-questions
To unsubscribe, send any mail to [EMAIL PROTECTED]


Re: ipfilter traffic blocking and tcpdump snort etc

2003-12-05 Thread Melvyn Sopacua
On Friday 05 December 2003 11:58, Jez Hancock wrote:

 Let me rephrase that one :P  I meant is there a method - for example
 such as adding some kind of routing via arp - so that packets are
 dropped on the floor even quicker than they would be via the firewall
 method?

You could bind the ip's to the loopback interface, but I think the firewall 
setup is quicker.

-- 
Melvyn

===
FreeBSD sarevok.idg.nl 5.2-BETA FreeBSD 5.2-BETA #1: Thu Dec  4 12:47:31 CET 
2003 [EMAIL PROTECTED]:/usr/obj/usr/src/sys/SAREVOK_NOAPM_NODEBUG  i386
===


pgp0.pgp
Description: signature


Re: ipfilter traffic blocking and tcpdump snort etc

2003-12-05 Thread Jez Hancock
On Fri, Dec 05, 2003 at 01:10:16PM +0100, Melvyn Sopacua wrote:
 On Friday 05 December 2003 11:58, Jez Hancock wrote:
 
  Let me rephrase that one :P  I meant is there a method - for example
  such as adding some kind of routing via arp - so that packets are
  dropped on the floor even quicker than they would be via the firewall
  method?
 
 You could bind the ip's to the loopback interface, but I think the firewall 
 setup is quicker.
Interesting(!) idea but kind of does the DOS'ers job for 'em!

I'm really curious as to what type of attack it actually was.  Right now
I know:

- it was aimed at a single address on port 80
- global apache errorlog was relatively quiet in the run up to the
  exhaustion of apache with only a small hint that a larger number of
  requests were being made:

[Thu Dec  4 18:47:46 2003] [info] server seems busy, (you may need to increase 
StartServers, or Min/MaxSpareServers), spawning 8 children, there are 0 idle, and 146 
total children
[Thu Dec  4 18:47:47 2003] [error] server reached MaxClients setting, consider raising 
the MaxClients setting
[Thu Dec  4 18:52:34 2003] [notice] child pid 91863 exit signal Segmentation fault (11)
snip same error log line repeated around 4,500 times!
[Fri Dec  5 00:13:04 2003] [notice] child pid 38280 exit signal Segmentation fault (11)
[Fri Dec  5 01:35:52 2003] [info] server seems busy, (you may need to increase 
StartServers, or Min/MaxSpareServers), spawning 8 children, there are 0 idle, and 17 
total children

  note the 5min gap between the server reaching the MaxClients setting
  and the server collapsing with no err log entries in between

- no HTTP requests were logged by apache from any of the dozen or so
  attacking hosts

- snort captured only SYN packets from the attacking hosts (I suppose
  this explains why no requests were logged by apache)
  
- all the attacking hosts had both port 25 and 80 open, although none of
  those hosts accepted inbound connections to those ports

Would appear someone had control over a few zombie hosts and was able to
coordinate a distributed attack - thankfully it was only a dozen or so
hosts :P

-- 
Jez Hancock
 - System Administrator / PHP Developer

http://munk.nu/
___
[EMAIL PROTECTED] mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-questions
To unsubscribe, send any mail to [EMAIL PROTECTED]


RE: ipfilter traffic blocking and tcpdump snort etc

2003-12-05 Thread David
Maybee an upgrade of apache would be a good start?.  and have a look at
mod_bandwidth http://www.cohprog.com/mod_bandwidth.html and mod_dosevasive
http://www.nuclearelephant.com/projects/dosevasive/

.

-david

 -Original Message-
 From: Jez Hancock [mailto:[EMAIL PROTECTED]
 Sent: Friday, 5 December 2003 23:41
 To: [EMAIL PROTECTED]
 Subject: Re: ipfilter traffic blocking and tcpdump snort etc
 
 
 On Fri, Dec 05, 2003 at 01:10:16PM +0100, Melvyn Sopacua wrote:
  On Friday 05 December 2003 11:58, Jez Hancock wrote:
  
   Let me rephrase that one :P  I meant is there a method - 
 for example
   such as adding some kind of routing via arp - so that packets are
   dropped on the floor even quicker than they would be via 
 the firewall
   method?
  
  You could bind the ip's to the loopback interface, but I 
 think the firewall 
  setup is quicker.
 Interesting(!) idea but kind of does the DOS'ers job for 'em!
 
 I'm really curious as to what type of attack it actually was. 
  Right now
 I know:
 
 - it was aimed at a single address on port 80
 - global apache errorlog was relatively quiet in the run up to the
   exhaustion of apache with only a small hint that a larger number of
   requests were being made:
 
 [Thu Dec  4 18:47:46 2003] [info] server seems busy, (you may 
 need to increase StartServers, or Min/MaxSpareServers), 
 spawning 8 children, there are 0 idle, and 146 total children
 [Thu Dec  4 18:47:47 2003] [error] server reached MaxClients 
 setting, consider raising the MaxClients setting
 [Thu Dec  4 18:52:34 2003] [notice] child pid 91863 exit 
 signal Segmentation fault (11)
 snip same error log line repeated around 4,500 times!
 [Fri Dec  5 00:13:04 2003] [notice] child pid 38280 exit 
 signal Segmentation fault (11)
 [Fri Dec  5 01:35:52 2003] [info] server seems busy, (you may 
 need to increase StartServers, or Min/MaxSpareServers), 
 spawning 8 children, there are 0 idle, and 17 total children
 
   note the 5min gap between the server reaching the MaxClients setting
   and the server collapsing with no err log entries in between
 
 - no HTTP requests were logged by apache from any of the dozen or so
   attacking hosts
 
 - snort captured only SYN packets from the attacking hosts (I suppose
   this explains why no requests were logged by apache)
   
 - all the attacking hosts had both port 25 and 80 open, 
 although none of
   those hosts accepted inbound connections to those ports
 
 Would appear someone had control over a few zombie hosts and 
 was able to
 coordinate a distributed attack - thankfully it was only a dozen or so
 hosts :P
 
 -- 
 Jez Hancock
  - System Administrator / PHP Developer
 
 http://munk.nu/
 ___
 [EMAIL PROTECTED] mailing list
 http://lists.freebsd.org/mailman/listinfo/freebsd-questions
 To unsubscribe, send any mail to 
 [EMAIL PROTECTED]
 
___
[EMAIL PROTECTED] mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-questions
To unsubscribe, send any mail to [EMAIL PROTECTED]


Re: ipfilter traffic blocking and tcpdump snort etc

2003-12-05 Thread Jez Hancock
On Sat, Dec 06, 2003 at 12:01:09AM +1100, David wrote:
 Maybee an upgrade of apache would be a good start?.  and have a look at
 mod_bandwidth http://www.cohprog.com/mod_bandwidth.html and mod_dosevasive
 http://www.nuclearelephant.com/projects/dosevasive/
I upgrade manually using portupgrade where necessary every weekend after
the weekly periodic run.  Without doubt apache is up to date - 
unless any changes to the ports in the last few days :P

Server: Apache/1.3.29 (Unix) mod_accounting/0.5 PHP/4.3.4 mod_perl/1.28
mod_throttle/3.1.2

Gets me that something as simple as a flood of packets can just
cripple a service so easily given enough bandwidth (although adding ipf
rules helped a lot).

I've not actually checked out mod_bandwidth, I use mod_throttle - fwiw,
it's not great for multiple vhosts :( - should check that out, thanks.

mod_dosevasive sounds even more interesting.  Heading toward that link
now...

Very interesting, particularly this feature:

The blacklist can/should be configured to talk to your network's
firewalls and/or routers to push the attack out to the front lines, but
this is not required.

This is something I could do with for Exim as well :P

Not sure that it would have helped last night actually - no hits
were actually registered by apache during the attack from any of the
attacking hosts.  As I said in another post, all the packets I captured
from the attacking hosts with snort during the packet attack only had the SYN
flag set - it appeared to be the sheer volume of these packets to port
80 that was causing apache child procs to die rapidly in succession.

The hardest part over the few hours the attack lasted was working out
from the snort logs which _bad_ hosts I'd already blocked with ipf, which
hosts were legit and which hosts I still had to block  - over time more
hosts joined in the attack.  The last 20mins or so consisted of a flood
from a single host I'd missed in my blocking spree!

Fun and games :=P

-- 
Jez Hancock
 - System Administrator / PHP Developer

http://munk.nu/
___
[EMAIL PROTECTED] mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-questions
To unsubscribe, send any mail to [EMAIL PROTECTED]


Re: ipfilter traffic blocking and tcpdump snort etc

2003-12-05 Thread horio shoichi
On Fri, 5 Dec 2003 10:58:39 +
Jez Hancock [EMAIL PROTECTED] wrote:
 Hi Horio,
 
 Cheers for reply.
 
 On Fri, Dec 05, 2003 at 07:33:49PM +0900, horio shoichi wrote:
  On Fri, 5 Dec 2003 00:24:12 +
  Jez Hancock [EMAIL PROTECTED] wrote:
   Hi,
   
   I've blocked a dozen or so addresses using ipfilter:
   
   block in quick on fxp0 from 208.186.60.116 to any
   block in quick on fxp0 from 216.230.149.11 to any
   
   etc
   
   but I still see a lot of traffic those hosts in trafshow, snort and
   other packet capturing utils.  Why is this?
  
  You are probably seeing the supposedly blocked packets on the outside of
  network. Observe them on inside, i.e., on the interface not fxp0.
 Not sure what you mean here, what command would you issue via tcpdump or
 snort to do what you suggest?  

Um, that's my bad assumption !
I thought your box is a filtering router, and has at least two interfaces.

 
  What you are seeing are packets ipfilter is just about to handle.
 Right - it's just I would have thought that ipfilter handled packets
 before they reached any traffic dumping utils.  I see what you're
 getting at.  Presumably snort for example uses the bpf driver via pcap(?) to
 capture network traffic...
 
 actually reading bpf(4) clears things up a little:
 
  Associated with each open instance of a bpf file is a user-settable
  packet filter.  Whenever a packet is received by an interface, all file
  descriptors listening on that interface apply their filter.  Each
  descriptor that accepts the packet receives its own copy.
 
The log keyword on blocking rules would have helped...

   Is there any alternative method of blocking access from certain hosts
   so that this traffic is not 'seen' by higher level /userland apps?
  I don't understand your second question. Are you thinking about tcp wrapper,
  reset feature of snort, etc ?
 Let me rephrase that one :P  I meant is there a method - for example
 such as adding some kind of routing via arp - so that packets are
 dropped on the floor even quicker than they would be via the firewall
 method?

In my observation, packet filters are the quickest since blocked packets die
in ip_input(), below which is where ethernet interrupt handlers are laid out.


horio shoichi

 
 -- 
 Jez Hancock
  - System Administrator / PHP Developer
 
 http://munk.nu/
 ___
 [EMAIL PROTECTED] mailing list
 http://lists.freebsd.org/mailman/listinfo/freebsd-questions
 To unsubscribe, send any mail to [EMAIL PROTECTED]
 

___
[EMAIL PROTECTED] mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-questions
To unsubscribe, send any mail to [EMAIL PROTECTED]


ipfilter traffic blocking and tcpdump snort etc

2003-12-04 Thread Jez Hancock
Hi,

I've blocked a dozen or so addresses using ipfilter:

block in quick on fxp0 from 208.186.60.116 to any
block in quick on fxp0 from 216.230.149.11 to any

etc

but I still see a lot of traffic those hosts in trafshow, snort and
other packet capturing utils.  Why is this?

Is there any alternative method of blocking access from certain hosts
so that this traffic is not 'seen' by higher level /userland apps?

As background, the blocked hosts were part of a denial of service attack
which has been going on for a few hours now.  The attack was aimed at
port 80, although an odd artifact is that no httpd log entries were made
for any of the hosts attempting to connect on port 80.

A cursory nmap scan of a few of the hosts shows that all hosts had both
port 25 and 80 open, but none of the hosts accepted connections on
either of those ports.  Any idea what kind of attack this could be?

-- 
Jez Hancock
 - System Administrator / PHP Developer

http://munk.nu/
___
[EMAIL PROTECTED] mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-questions
To unsubscribe, send any mail to [EMAIL PROTECTED]