Most Read This Week
Cover Story: Stealth Firewalling with Linux
Ever Needed to Set Up a Firewall in a Network Without Anyone Knowing It?
Feb. 27, 2005 12:00 AM
Have you ever needed to set up a firewall in a network without anyone knowing it was there or so that it wouldn't require you to change your network? Welcome to the world of stealth firewalling. We dedicated one chapter in our book, Troubleshooting Linux Firewalls, to it but honestly the topic has so many uses, to do it justice it really needs its own book (or books!).
What Is Stealth Firewalling with Linux?Simply put, a stealth firewall is an Ethernet bridge with filtering capabilities. This means that it's a firewall that operates at Layer 2 of the OSI model, leveraging netfilter rules and chains (Linux's firewall system) applied to the bridge. For those not familiar with what a bridge is: an Ethernet bridge is a means of connecting two or more networks/devices at the Data Link layer. The Data Link layer is the layer of the OSI model before the Network Layer (Layer 3). Layer 3 is where things like the IP in TCP/IP come into the picture. This means that a bridge operates before we get into things like protocols, so you can apply firewalling to protocols other than IP (IPX, DECnet, SNA, etc.) and you don't have to worry about routing issues either. You're moving raw frames from one interface to another, which lets you deploy your stealth firewall without anyone being the wiser. Another advantage of a stealth firewall is that if the hardware fails or malfunctions, you can bypass it with nothing more than a network cable.
How Does a Bridge Work?A bridge operates in promiscuous mode, grabbing all the packets it sees on its interfaces, learns which MAC addresses apply to which interfaces, and moves packets between those interfaces. One example of a bridge that many are familiar with is the humble network switch, wherein all the interfaces on a switch comprise one bridge. With Linux, when you create a bridge you are combining two or more interfaces into one bridge. For instance, we'll create a bridge, br0, that is made up of two interfaces, eth0 and eth1. Think of br0 as a set that contains two elements, eth0 and eth1. When the two interfaces are added to the bridge, the kernel will then move packets back and forth automatically between those two interfaces as if they were both part of the same physical network.
A Brief Explanation of Netfilter
Thanks to the flexibility of the netfilter subsystem and the way it handles packets, we can use it to apply firewall rules to network traffic, even if the device is not operating at the IP layer. The subsystem doesn't actually care; it can apply firewall rules to whatever traffic passes through it. To help explain this, Figure 1 shows which netfilter rules are applied to the bridge interface, and the order in which they are applied for traffic flowing from eth0 to eth1 over the bridge.
What This Means in Practical TermsYou can put a firewall in place without anyone knowing it's there or making any IP topology changes. For example, at the California Community Colocation Project (www.communitycolo.net), we used a bridging/stealth firewall to filter out worm traffic entering and exiting the network. This was done in such a way that there was no interruption for the users of the network, no need to change the routing tables or to even muck about with IP issues, and it required only some basic physical wiring changes. The use of a stealthy firewall such as this leads to all sorts of interesting alternative applications outside of just basic firewalling, such as stealth proxies, traffic-shaping devices, anti-spam/virus filtering, and stealth IDS/IPS systems. The key advantage for the firewall administrator is that you can deploy a fairly paranoid firewall setup without having to make any changes to your network, and failing over, as previously mentioned, is as simple as plugging in a patch cable.
How Do I Set One Up?We're using Fedora Core 3 for our demonstration environment, but the following should apply to nearly any modern Linux distribution running a 2.6 kernel. You can also do this with a 2.4 kernel, but for brevity's sake we're going to stick with the screaming edge. It's more fun that way. First, we need to make sure our kernels support both bridging and Layer 2 filtering in netfilter. If you're running FC3 with the default kernel, you can skip this next bit; this is for those of you running custom 2.6 kernels or other distributions that don't include this functionality by default. To enable bridging and Layer 2 filtering, you'll need to make sure the following settings are configured in your kernel to support ebtables.
Device Drivers Networking Support Networking Options ---> <M> 802.1d Ethernet Bridging [*] Network packet filtering (replaces ipchains) ---> [*] Bridged IP/ARP packets filtering Bridge: Netfilter Configuration ---> <M> Ethernet Bridge tables (ebtables) support <M> ebt: broute table support (NEW) <M> ebt: filter table support (NEW) <M> ebt: nat table support (NEW) <M> ebt: 802.3 filter support (NEW) <M> ebt: among filter support (NEW) <M> ebt: ARP filter support (NEW) <M> ebt: IP filter support (NEW) <M> ebt: limit match support (NEW) <M> ebt: mark filter support (NEW) <M> ebt: packet type filter support (NEW) <M> ebt: STP filter support (NEW) <M> ebt: 802.1Q VLAN filter support (NEW) <M> ebt: arp reply target support (NEW) <M> ebt: dnat target support (NEW) <M> ebt: mark target support (NEW) <M> ebt: redirect target support (NEW) <M> ebt: snat target support (NEW) <M> ebt: log support (NEW)
You may also need to install ebtables for your distribution. The userspace ebtables utility and 2.4 kernel patches are available at http://sf.net/projects/ebtables and from our Web site - www.gotroot.com.
Next we need to set up our firewall, in this case it's named Minimoose (kudos to those of you who get this reference), to bridge traffic between our network, 10.10.10.0/24, and our gateway, 10.10.10.1. The goal is to put Minimoose between the network and the gateway device. We start this process by first configuring Minimoose's interfaces to act as a bridge. This is accomplished by using the userspace tool, brctl, to set up the bridged interfaces and to "capture" them. In this hypothetical example, Minimoose has two interfaces. One is pointing at the gateway router, which has an IP address 10.10.10.1 and is reached via ethernet interface eth0. Minimoose's other interface, eth1, is facing the actual 10.10.10.0/24 network. Remember, the bridge does not need any IP addresses. It's going to "bridge" the traffic between these two points, allowing traffic to move from the 10.10.10.0/24 network to the gateway router, 10.10.10.1, and vice versa as defined by the firewall rules on Minimoose. The first step, as already explained, is to set up the interfaces on Minimoose and to configure it as a bridge:
1. Create the logical interface in the kernel, which is called br0.
[root@minimoose root]# brctl addbr br0
2. Add the left interface, eth0, that connects to the 10.10.10.1 gateway.
[root@minimoose root]# brctl addif br0 eth0
3. Add the right interface, eth1, that connects to the 10.10.10.0/24 network.
[root@minimoose root]# brctl addif br0 eth1
4. Activate the bridged interfaces by bringing up the two real interfaces.
[root@minimoose root]# ifconfig eth0 0.0.0.0 promisc up
At this point your bridge is working.
As you have probably surmised, Minimoose doesn't even have an IP address for itself. Although it's probably a good idea to add an IP address to one of those interfaces for management reasons, you don't have to in most instances. For example, you could just as easily connect this device to a serial console, or add another interface and run an "Out of Band" network just for admining your stealth firewall, or something like that if you're über paranoid.
Figure 2 shows our network, 10.10.10.0/ 24, before we put our firewall in place.
Now we put our firewall, Minimoose, in place as shown in Figure 3.
With traffic passing from 10.10.10.0/24 to the gateway through Minimoose, it's time to do something interesting with the firewall. Listing 1 provides a simple example script that sets up the firewall rules and configures them to stealthily drop all SMB traffic moving between the gateway and the hosts on the 10.10.10.0/24 network. These rules also allow traffic to move in both directions, not just from the 10.10.10.0/24 network outward, but also from the gateway back into the 10.10.10.0/24 network. This listing will help illustrate the stealthy nature of this configuration; neither the hosts on the 10.10.10/24 network nor the gateway will "see" any changes in the network, except that SMB traffic is not flowing between the two.
In the previous example you can see that applying firewall rules to Layer 3 (IP) traffic works exactly the same as it does with a normal firewall, that is we need to apply those rules to the FORWARD table, and of course make sure you enable ip_forwarding on our firewall for this example to work:
echo 1 > /proc/sys/net/ipv4/ip_forward
Another application of a stealth firewall would be to act as a transparent proxy server. We frequently add squid to our firewalls for performance reasons and this next example shows how to configure the squid Web proxy cache server for your stealth firewall. The good news is that setting this up is easy; the bad news is that it requires adding an IP address to the stealth firewall and, because of this, your firewall will not be as stealthy as you might like, as Web requests will be seen as if they were coming from the bridge's IP address and, of course, your bridge will now have an IP address. Keep in mind, this doesn't mean that other traffic will appear to come from the bridge's IP address or that you need to change any default routes on your network. The stealth firewall will just be seen as another node on your network, not as a router. The steps to set this up are:
1. Install squid and configure squid.conf to allow connections from localhost. This is typically the default configuration for squid. The steps to do this are beyond the scope of this article, but if you are having problems, please visit our Web site for documentation: www.gotroot.com.
2. Add an IP address to your bridge interface, if you haven't already done so. In the commands that follow, replace the variable $MANAGEMENTIP with the IP address you intend to assign to your bridge, and the $MANAGEMENTGATEWAY variable with the gateway address for your network.
ifconfig br0 $MANAGEMENTIP
3. Finally, add the following two rules to your firewall:
iptables -A INPUT -i <internal interface on bridge, such as eth1> -p tcp -d $MANAGEMENTIP -s
Your stealth firewall is now configured to grab all the port 80 traffic coming from our hypothetical stealth firewall's internal network, 10.10.10.0/24. Redirect it to squid, and then allow squid to request and cache the documents. Again, all Web requests will now come from the bridge's IP address.
Hopefully this brief introduction to stealth firewalls has encouraged you to think up your own ways to use them in your environment. Stealth firewalls are a fantastic and reliable tool for any network or security administrator. They're easy to set up, can be more secure than classic IP firewalls, and because they require no routing and don't have to be integrated into the network's routing scheme, they are actually less of a risk to a network, from a point-of-failure perspective, than a traditional IP firewall.
As we already mentioned, if the stealth firewall fails, just run a patch cable around it. Of course, this eliminates your firewall and we don't want that. The good news is that there's an easy solution to that problem as well, and it allows you to create a fully automatic failover between two or more stealth firewalls. The better news is that this is already built into Linux and is very robust, but that discussion is for another time.
Reader Feedback: Page 1 of 1
Subscribe to the World's Most Powerful Newsletters
Today's Top Reads