Installation Method 1
The hardest part about setting up a Linux firewall is getting Linux installed on your firewall computer, and that is no longer very difficult. Any of the current Linux distributions will work—from Red Hat, Caldera Systems, Mandrake, to SuSE. They support everything you'll need for the firewall in the standard installation.
Caldera's eDesktop 2.4 and Mandrake Linux version 7 are the easiest Linux distributions to install. It doesn't really matter which one you choose, though, because under the hood they all run the same Linux kernel and use the same firewall software.
Most of the current distributions give you many predefined choices, including a standard installation and a server installation. The server installation adds only what is needed to run a server, which speeds up system startup and makes the system more secure overall. If you have a choice and you won't be using the firewall computer for anything else, choose the server installation.
If you are installing Linux onto a more powerful computer, you could choose the standard installation and then use the computer for more than just a firewall. If you do this, you may want to limit use of the computer while you are connected to the Internet; depending on what is being done on the computer, it can affect the performance of your network connection.
Make sure that both network cards are configured properly. You can use the configuration tool that is included with your Linux distribution—such as linuxconf or COAS—or you can use the netcfg tool. The first network card uses a standard setup for connecting to the Internet. The second card will have a private internal IP address, typically 192.168.0.1.
Private network addressing is defined in RFC 1918, which defines a group of IP addresses that will never be assigned and that are not routable through the Internet. These addresses are reserved for use in a home or company internal network.
Enough reserved network addresses have been defined by RFC 1918 to accommodate the needs of any corporation, no matter how large. This means that, no matter how big or small, no network needs its own registered IP address. If you need to connect to Internet, the network address translator can handle the job. The reserved addresses are these:
Make sure that both network cards are working properly by using the command /sbin/ifconfig -a. This will show all of your network devices. You should see eth0 and eth1, and both must be in an UP state.
Also, make sure that you have a working connection. Check the connection using the ping command; for example, ping www.informit.com and make sure that you get a reply.
The Linux firewall computer needs to be secured. First, disable all services that you do not need. For a firewall system, almost everything should be disabled. To see what ports are open, enter the command netstat -atu.
In the /etc/inetd.conf file, put a hash mark in front of all of the lines for services you won't need on the firewall computer. This includes gopher, exec, talk, ntalk, pop2, pop3, imap, uucp, finger, shell, and login. Even ftp and telnet could be disabled if you will be making all of your configuration changes from a keyboard and a monitor that are connected directly to the firewall computer.
When a service is requested from your computer, a utility called TCP wrappers checks whether the connection is allowed. TCP wrappers work with the hosts.allow and hosts.deny files in the /etc directory. Make sure that the hosts.deny file has ALL : ALL on its last line. This means that, unless something has been specifically allowed in the hosts.allow file, the service will be denied. The hosts.allow file should have at least one line in it that reads ALL : localhost; this ensures that you can use these services when you are using the localhost, the firewall computer.
You should also remove all service daemons that are activated at startup, such as sendmail and the Web server daemon. See the instructions on how to do this for your Linux version. In Red Hat, it can be done with linuxconf; in Caldera OpenLinux, use COAS.
For even more security tips, see the resources listed at the end of this article.
All of the computers that will connect to the Internet through the firewall must be connected to your network hub and have TCP/IP networking installed and running. The internal network should use one of the private reserved address blocks.
Finally, all of the computers that will be connecting to the Internet through the Linux firewall must be set up to use the Linux firewall as the default gateway. (For reference, see the graphic in Part 2 to see how the connections are set up.) The firewall computer has two Ethernet cards, one that is connected to the Internet through the cable modem, DSL modem, or other device, and a second card that is connected to your local network hub.
Your Internet service provider (ISP) assigns an IP to the card that is connected to the external hub. This address is either static (permanently assigned) or dynamic (it changes, and the firewall system gets the address using DHCP or some other dynamic addressing method).
The second Ethernet card on the Linux firewall computer is typically assigned the IP address of 192.168.0.1. The other computers on the internal network have IP addresses that use the same subnet: 192.168.0.2, 192.168.0.3, and so on. These computers also need to have the DNS set to your ISP's DNS server. If you would like, you can sometimes speed up response time by setting up and installing a caching DNS server on your Linux firewall, and you can then set your computers to use the Linux firewall as the DNS server. See the Linux DNS how-to for a description of setting up a caching DNS server.
The first step is to create a set of rules for your firewall. On the firewall computer, use a text editor such as vi to create a filed called rc.ipchains in the /etc/rc.d directory.
In this file, put the standard rule set. You can check the resources at the end of this guide for information on adding rules to ipchains. Put these basic firewall rules in a file called rc.ipchains in the /etc/rc.d directory:
#! /bin/sh #Enable IP forwarding echo "1" > /proc/sys/net/ipv4/ip_forward # If you get your IP address
dynamically from SLIP, PPP # # or DHCP, enable this option. # #echo "1" > /proc/sys/net/ipv4/ip_dynaddr # Enable timeouts -
2 hour timeout for TCP session timeouts - 10 sec traffic timeout - # # 60 sec UDP traffic timeout # ipchains -M -S 7200 10 160 # Enable basic IP forwarding and masquerading /sbin/ipchains -P forward DENY /sbin/ipchains -A forward -s
192.168.0.0/255.255.255.0 -j MASQ #Enable proper masquerading of FTP transfers /sbin/insmod ip_masq_ftp #Enable RealAudio masquerading over UDP. /sbin/insmod ip_masq_raudio #Uncomment if you need any of these services. #Enable Quake and QuakeWorld masquerading #/sbin/insmod ip_masq_quake #Enable CuSeeme video conferencing #/sbin/insmod ip_masq_cuseeme #Enable VDO-live video conferencing #/sbin/insmod ip_masq_vdolive #Enable IRC DCC file transfers #/sbin/insmod ip_masq_irc
One type of attack on the Internet involves IP spoofing, a technique where the attacker sends out packets that claim to be from a different computer. Packet filtering firewalls either accept or deny access based on the IP address, so IP spoofing is used to get past packet filters. It is also used to hide the identity of the attackers. The internal address, if 127.*, is already blocked, but any internal address using any of the other reserved internal subnet addresses may be vulnerable.
To block IP spoofing, add the following lines to the rc.ipchains configuration file. This assumes that you are using the IP subnet of 192.168.0.x. If you're using a different subnet, change the rule so that it fits what you are using:
ipchains -A input -i eth0 -s !
192.168.0.0/255.255.255.0 -j DENY
ipchains -A input -i !
eth0 -s 192.168.0.0/255.255.255.0 -j DENY
Next, make the script executable. Run this command:
chmod 700 /etc/rc.d/rc.ipchains.
Then make the ipchains script run every time the system starts. Edit the rc.local file in the /etc/rc.d directory, and add a new line that reads rc.ipchains.
Now, start ipchains with the command /etc/rc.d/rc.ipchains. Test to see if it is working. First, ping the Linux firewall from any computer on your internal network with the command ping 192.168.0.1. If that works, then ping a site on the Internet. Not all active servers on the Internet accept pings. One that will work is the InformIT Web site. Use the command ping www.informit.com. Sometimes pings will time out as a result of network traffic or other reasons.
Run other tests using telnet or a Web browser from one of your computers. If you can get through, your Linux firewall is working.
See the resources for additional information on ipchains rules. The TrintyOS document, in particular, contains a good example of an extra-strong firewall rule set in Section 10.