Re: External IP addresses on internal network

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Morning all-


I've seen this before. An internal client goes to access a Web site (say www.fubar.org) and the authoritative NS is actually a load balancer. It spews suspicious looking traffic at the requesting NS in order to generate performance metrics to figure out what IP to serve back (assumption being the client is close to the NS).


So if this is the case, you should see a query for a host within the AOL domain (owner of the address space) just prior to this traffic.

I haven't logged that traffic, but the user is generally checking an AOL mail account and could be following links or browsing.




As for seeing the firewall's private IP in the log entry, are you running DNAT on the reply traffic? If so that would explain why it shows up as private.

I run SNAT on POSTROUTING to masq the internal LAN but don't do any DNAT.



As mentioned I would verify by capturing the traffic. Something like:


windump -nn -s 1500 -w weird-dns.cap "src port 137 and dst port 53"

from a system inside of your firewall. If you get the log entry but no capture, you know its received from outside.

I'm a cup of coffee short of a full pot today. I'll have to work on this later :-)



I *totally* agree. I write the material and teach SANS perimeter security track and one thing I am uber big on is filtering outbound as well. Stuff like echo-replies, type 3's, type 11's, NetBIOS/IP, SNMP, TFTP, etc. etc. should never be allowed to leave your network. Defense in-depth and all of that. If the attacker's stimulus gets in, at least you have a shot at blocking the reply.


Since I'm an IT lightweight, I generally use sample scripts and default configs whenever I can. I knew from a philosophical standpoint that I wanted outbound filtering. I didn't really know how to do it and I didn't find any examples that could help me, so I used the brute force method: reject everything and see what it breaks, then open those ports only. This method works, sort of, but it can be a problem since many programs assign ports dynamically. Anyway, I caught me a big ole worm this way so I must be doing something right.



[Index of Archives]     [Linux Netfilter Development]     [Linux Kernel Networking Development]     [Netem]     [Berkeley Packet Filter]     [Linux Kernel Development]     [Advanced Routing & Traffice Control]     [Bugtraq]

  Powered by Linux