loading...
Cover image for Securing your Linux desktop using iptables firewall rules

Securing your Linux desktop using iptables firewall rules

dandyvica profile image Alain Viguier Updated on ・4 min read

Today, I'll be tackling not a specific development topic, but rather a way of securing your Linux desktop access to the internet. I'm using a Linux Mint 19.1 version, but it should be the same across all Linux distributions as soon as the iptables package is installed.

iptables is a Linux command to manage firewall rules. It basically filters network packets coming from or going to the external world.

There're a lot of awesome resources describing the nuts and bolts of iptables, I'm not going to detail this here. I'm rather focusing on how to set up simple rules in order to secure your internet connection from your desktop. This comes up in addition to other safeguarding techniques like using privacy minded browser extensions like uBlock origin or uMatrix, or using AppArmor confinement framework to name a few.

The first thing to understand when dealing with iptables are chains. Chains are like channels where network packets are exchanged (not talking about NAT and MANGLE tables, not the subject here), in the FILTER table. A chain contains of a list of rules which are applied sequentially, in the order which they were created. If a packet doesn't match a rule, it's either rejected, accepted or dropped.

To define a rule, you can use the iptables command (root access needed). Or create a rules file and import all rules using the iptables-restore command.

There're 3 chains:

  • the INPUT chain is for all incoming packets from the outside world
  • the FORWARD chain is used when the host is considered as a gateway forwarding packet to another host
  • the OUTPUT chain is for all outgoing packets from my Linux desktop

The firewall strategy to harden you desktop is to deny everything, and then only accept the packets you want.

My setup is the following:

# load firewall config with iptables-restore < iptables.rules

*filter

# 1: set default DROP policy
:INPUT DROP [0:0]
:FORWARD DROP [0:0]
:OUTPUT DROP [0:0]

# 2: accept any related or established connection
-A INPUT -m state --state ESTABLISHED,RELATED -j ACCEPT
-A OUTPUT -m state --state ESTABLISHED,RELATED -j ACCEPT

# 3: allow all traffic on the loopback interface
-A INPUT -i lo -j ACCEPT
-A OUTPUT -o lo -j ACCEPT

# 4: allow outbound DHCP requests
-A OUTPUT -p udp --dport 67:68 --sport 67:68 -j ACCEPT

# 5: allow outbound DNS lookups
-A OUTPUT -p udp -m udp --dport 53 -j ACCEPT

# 6: allow outbound ping requests
-A OUTPUT -p icmp -j ACCEPT

# 7: allow outbound NTP requests
-A OUTPUT -p udp --dport 123 --sport 123 -j ACCEPT

# 8: allow outbound http/https requests
-A OUTPUT -p tcp -m tcp --dport 80 -m state --state NEW -j ACCEPT
-A OUTPUT -p tcp -m tcp --dport 443 -m state --state NEW -j ACCEPT

# 9: allow SMTP
-A OUTPUT -p tcp --dport 25 -m state --state NEW -j ACCEPT

# 10: allow incoming IMAP/IMAPS
-A OUTPUT -p tcp --dport 143 -m state --state NEW -j ACCEPT
-A OUTPUT -p tcp --dport 993 -m state --state NEW -j ACCEPT

# 11: access SSH server
-A OUTPUT -p tcp -m tcp --dport 22 -j ACCEPT

# commit changes
COMMIT

The previous rules are basically all the parameters which are passed to the iptables command. Mainly:

  • -A: append to the chain name
  • -p: the protocol name
  • -m: additional option passed to an iptables module
  • --sport: the source port
  • --dport: the destination port
  • -j: the action whenever such a packet is matched
  • --state: the connection state module options

Let's breakdown those rules:

  1. by default, drop all packets for all chains
  2. accept, either incoming or outgoing, any packet where its state is either ESTABLISHED (connection is known and tracked) or RELATED (initiated from an already established connection)
  3. the loopback interface is 127.0.0.1 or localhost. It's a local one, so accept all connexions to it
  4. allow dhcp requests, from tcp source port 67 or 68 to destination port 67 or 68
  5. allow DNS requests to port 53
  6. if you want to ping an address, you need this rule
  7. if you're using the time protocol to set your clock, you need this rule
  8. the core one: only allow http or https protocol from your box, those initiating the connection (--state NEW)
  9. if you're using a mail software like Thunderbird to read and send emails, this adds a rule for SMTP server
  10. the same for IMAP/IMAPS servers
  11. If you have to access a remote SSH server (for example your AWS ionstance), add this rule

You can also fine tune those rules. You can for example list a couple of ip addresses or urls to restrict the destination address using the -s parameter. You can also define a rule for a specific interface name like -i eht0.

Then simply load those rules by:

$ sudo iptables-restore < iptables.rules

These will be deleted during the next reboot. To make them permanent, install the iptables-persistent packed:

$ sudo apt-get install iptables-persistent

Hope this helps !

Photo by Andy Art on Unsplash

Posted on by:

dandyvica profile

Alain Viguier

@dandyvica

Senior software engineer & sysadmin, technical architect, Linux expert. Always willing to learn new stuff.

Discussion

pic
Editor guide
 

Dear sir, Can you consider Rule #4 and #7. I think we could ignore --sport. So we can modify like this:

4: allow outbound DHCP requests

-A OUTPUT -p udp --dport 67:68 -j ACCEPT

7: allow outbound NTP requests

-A OUTPUT -p udp --dport 123 -j ACCEPT

Because source port will be randomed when establish new connection.

What do you think?

 

This is great.. Would love to see one for pfctl.. for all the MacOS devs.