Spades wrote:

I got this error when i tried to type for some of those.
"sysctl: unknown oid...." any idea..

my server seems to be very lagged, where else
the network connection seems fine, i think BSD
itself as my other redhat box is fine.

What else can i do to get optimum protection.


----- Original Message ----- From: "Per Engelbrecht" <[EMAIL PROTECTED]>
Sent: Saturday, February 07, 2004 5:58 PM
Subject: Re: SYN Attacks - how i cant stop it



all nights. Check this.

Feb  6 11:54:24 TCP: port scan detected [port 6667] from [ports 63432,63453,63466,63499,63522,...]
Feb  6 11:58:09 TCP: port scan mode expired for -


It's hard to get rid of shit-heads like this - I'm talking about the person doing this attac, that is. You send a looong output of a log, but no info on your system or any adjustments you have made (or not made) on your system i.e. kernel (options), sysctl (tweaks) and ipfw (rules). If the problem is out-of-bandwith (and your system already has been optimized) then the only real solution is more 'pipe' a.k.a the Microsoft-solution. So fare I've only been guessing, but here is what I normally do with my setup. I'm not telling you that this is the solution! just adwises!

options      SC_DISABLE_REBOOT
options      IPFIREWALL
options      IPDIVERT
options      IPFILTER
options      IPFILTER_LOG
options      IPSTEALTH   (don't touch the ttl/can't see the wall)
options      TCP_DROP_SYNFIN   (drop tcp packet with syn+fin/scanner)
options      RANDOM_IP_ID   (hard to do calculate ip frekv. number)
options      DUMMYNET   (e.g. 40% for web, 30% for mail and so on)
options      DEVICE_POLLING    (can't do this short and not with SMP)
options      HZ=1000   (can't do this short and not with SMP)

kern.ipc.somaxconn=1024      #this is set high!
kern.ipc.nmbclusters=65536   #this is set high!
kern.polling.enable=1         #remember kernel options
kern.polling.user_frac=50>90  #remember kernel options
(if you use dynamic rules in ipfw [stateful] you can tweak this)
net.inet.ip.fw.dyn_ack_lifetime=200 #shorte timeout on connection
net.inet.ip.fw.dyn_short_lifetime=10 #longer timeout for e.g. icmp
net.inet.ip.fw.dyn_max=1500 #higher number of dynamic rules
net.inet.ip.fw.dyn_count:   #count of number of dynamic rules

There's a zillion ways to set it up. start with a few rules regarding
lo0 and icmp. Then use stateful inspection and dynamic rules for the
rest of the wall.

... and by the way, I could see that a few of the scan came from RIPE
ranges. Do some digging and report it!
Even if the boxes are use without the owners awareness, you can [we all
can] bring this part to an end.



Most important, you did turn on syncookies, did you not?

FreeBSD is pretty immune to syn floods. As for out of bandwidth, this has to do with your uplink and how much you pay for your traffic.

root# sysctl net.inet.tcp.syncookies

If it is not set to one, then do:
root# sysctl net.inet.tcp.syncookies=1

Also edit /etc/sysctl.conf to contain net.inet.tcp.syncookies=1.

A reboot would clear the tcp stack. You can't reboot remotely if kernel securelevel is enabled in /etc/rc.conf.

If you don't have firewall support compiled in the kernel, kldload ipfw.

Might be a good lesson to mirror back all incoming syn packets from the attacker's IP to him. To port 80, or 22, or to some any other open port. You can do that easely with ipfw.

Alin-Adrian Anton
Reversed Hell Networks
GPG keyID 0x1E2FFF2E (2963 0C11 1AF1 96F6 0030 6EE9 D323 639D 1E2F FF2E)
gpg --keyserver --recv-keys 1E2FFF2E
[EMAIL PROTECTED] mailing list
To unsubscribe, send any mail to "[EMAIL PROTECTED]"

Reply via email to