Using snort to detect if your users are doing interesting things?

Mailman List Mirror (Read Only)
1

Howdy, I am not sure if this is the proper place, if not I’ve noticed you guys know what to do so I’ll put the fire retardant suit on now. Recently due to growth we have seen an influx of “different” and “interesting” types of characters ending up on our network. They like to do all sorts of things, port scan /8s spam, setup botnets with the controllers hosted on my network… etc. I’m wondering what is the best way to detect people doing these things on my end. I realize there are methods to protect myself from people attacking from the outside but I’m not real sure how to pinpoint who is really being loud on the inside.

I did have one somewhat silly question… if you look at the statistics of a Fast Ethernet port, and it is doing both 2000 pps out, and 2000 pps in (pretty much equal in/out) but hardly any bandwidth at all can anyone think of a single application that would mimic that behavior?

Sorry if this is elementary network school knowledge.

-Drew

2

I'm wondering what is the best way to detect people doing these things
on my end. I realize there are methods to protect myself from people
attacking from the outside but I'm not real sure how to pinpoint who is
really being loud on the inside.

Any IDS ought to be able to do this. The problem will be figuring out
where to connect its taps, and how to provide enough capacity at those
points to do so without negatively impacting your overall network
performance.

You should be lauded for doing this. If all providers did it the
Internet would be a much, much safer place.

I did have one somewhat silly question.. if you look at the statistics
of a Fast Ethernet port, and it is doing both 2000 pps out, and 2000 pps
in (pretty much equal in/out) but hardly any bandwidth at all can anyone
think of a single application that would mimic that behavior?

VoIP with a low-rate codec, or some quantitatively similar multimedia
or gaming application?

3

I'm wondering what is the best way to detect people doing these things on my end. I realize there are methods to protect myself from people attacking from the outside but I'm not real sure how to pinpoint who is really being loud on the inside.

One of the best things we did was setup a snort box with barnyard logging to a mysql server. The snort box has an IP out of each ARIN allocation we have.

On a schedule, we purge the logs in the mysql server that did not come from our IP space and if there are X number of things from one of our IPs, open an abuse ticket which then looks up what type of connection that IP is and finds the specific user. Its then a manual process to hit a 'turn off and note their account' button or notify a downstream ISP.

This setup appears to catch a ton of the worms that scan a /8. I'm sure there is probably a better way of doing this, but without throwing a box at each network access point or better utilizing cflow, I couldn't come up with it.

sam

4

We just finished deploying a Snort IDS system on our network. The task of doing so was well worth the effort, and quite a bit of effort and resources were needed for our deployment. Due to the fact that we have a sustained 5Gbps of traffic to monitor in our Tampa data center alone, a simple server running Snort was just not going to cut it and rather than deploying off of our core routers in Tampa, which would catch inbound and outbound traffic, we decided after our testing that placing our tap points on our core routers was just not going to be sufficient due to the amount of abuse we saw in testing between customers in our facility. We decided to build a single server for each of our distribution switches at all of our locations that would communicate to a central server running the ACID console. This deployment has allowed us to gather so much information about what TRULY is and has been going on, that we wonder why we didn’t do this sooner.

Please keep in mind that there are many right ways to deploy an IDS system, however only one is really going to fit most of your needs initially. With some time, patience, and quite a bit of caffine, you should be well on your way to dropping your abusive traffic on your network. Good luck to you!

5

And when you do set up such an arrangement, depending on the number of rules you turn on, you can generate truly massive volumes of data to be analyzed by ACID or other tools. It is relatively easy to deploy snort for large volume, small number of rules type deployments. Aside from scaling the collectors and management console themselves, it can even be a challenge to aggregate all that data in a WAN deployment.

IDS has to be aimed carefully and then fired. And then one needs to ask what the derived value is, and just how you’re going to deal with the info. The latter being a magnificent operational challenge.

Or that’s at least been my experience. YMMV.

6

Drew Weaver wrote:

            Howdy, I am not sure if this is the proper place, if not
I've noticed you guys know what to do so I'll put the fire retardant
suit on now. Recently due to growth we have seen an influx of
"different" and "interesting" types of characters ending up on our
network. They like to do all sorts of things, port scan /8s spam, setup
botnets with the controllers hosted on my network.. etc. I'm wondering

There are two basic methods, I guess:

1. You search for specific patterns. For example for somebody pinging
more than n addresses in a specific time frame. If you know what you are
looking for, you can set something up to do it easily.

2. You look for something strange. You will need some kind of
statistical method then. They have a tendency to produce false positives
from time to time, so you better look at the results closely.

I did have one somewhat silly question.. if you look at the statistics
of a Fast Ethernet port, and it is doing both 2000 pps out, and 2000 pps
in (pretty much equal in/out) but hardly any bandwidth at all can anyone
think of a single application that would mimic that behavior?

DNS-Servers?

Nils

7

As it was already noted, you need to be very careful about how you set your IDS up, specifically if you choose snort.
Snort is a very powerful tool, when used correctly. Unfortunately, when used incorrectly, it can hose your network over
completely.

My suggestion, in the case that you’ll use snort, is to do some extensive testing on a non-production network.
Take the time to learn and understand its functionality and intended purpose.

Tim

Thor Lancelot Simon tls@NetBSD.ORG
Sent by: owner-nanog@merit.edu

06/09/2005 11:33 AM

Please respond to
tls@rek.tjls.com

To
Drew Weaver drew.weaver@thenap.com

cc
nanog@merit.edu

Subject
Re: Using snort to detect if your users are doing interesting things?

On Thu, Jun 09, 2005 at 11:45:54AM -0400, Drew Weaver wrote:
> I'm wondering what is the best way to detect people doing these things
> on my end. I realize there are methods to protect myself from people
> attacking from the outside but I'm not real sure how to pinpoint who is
> really being loud on the inside.

Any IDS ought to be able to do this. The problem will be figuring out
where to connect its taps, and how to provide enough capacity at those
points to do so without negatively impacting your overall network
performance.

You should be lauded for doing this. If all providers did it the
Internet would be a much, much safer place.

> I did have one somewhat silly question.. if you look at the statistics
> of a Fast Ethernet port, and it is doing both 2000 pps out, and 2000 pps
> in (pretty much equal in/out) but hardly any bandwidth at all can anyone
> think of a single application that would mimic that behavior?

VoIP with a low-rate codec, or some quantitatively similar multimedia
or gaming application?

--
Thor Lancelot Simon tls@rek.tjls.com

"The inconsistency is startling, though admittedly, if consistency is to be
abandoned or transcended, there is no problem." - Noam Chomsky