Recently, I have encountered a problem of limiting Internet Access to specific programs. Could anybody recommend a good way of doing that, without using any particular software?
The solution for me happened to be straight forward.
- Create, validate new group; add required users to this group:
grep no-internet /etc/group
- Add user:
useradd -g no-internet username
Note: If you're modifying already existing user you should run:
usermod -a -G no-internet userNamecheck with :
sudo groups userName
- Create a script in your path and make it executable:
chmod 755 /home/username/.local/bin/no-internet
#!/bin/bash sg no-internet "$@"
Add iptables rule for dropping network activity for group no-internet:
iptables -I OUTPUT 1 -m owner --gid-owner no-internet -j DROP
Note: Don't forget to make the changes permanent, so it would be applied automatically after reboot. Doing it, depends on your Linux distribution.
Check it, for example on Firefox by running:
In case you would want to make an exception and allow a program to access local network:
iptables -A OUTPUT -m owner --gid-owner no-internet -d 192.168.1.0/24 -j ACCEPT
iptables -A OUTPUT -m owner --gid-owner no-internet -d 127.0.0.0/8 -j ACCEPT
iptables -A OUTPUT -m owner --gid-owner no-internet -j DROP
NOTE: In case of spawning the rules will be maintained. For example, if you run a program with no-internet rule and that program will open browser window, still the rules will be applied.
A more straightforward possibility: use firejail. It runs the application inside sandbox. At the sandbox, you can control the access of the application to any network or folder in your computer.
To execute a certain application without network access do following:
firejail --net=none <application>
In that case, "The sandbox looks like a computer without any network interfaces." (See Network Section in documentation)
firejail --net=none firefox will start firefox without any network connection.
See the Installation documentation.
You should install from the package system in your distribution, or better get the latest version LTS. (For example, this latest LTS version,
9.56.2, works also in Ubuntu 16.04.)
Then, starting a process without network access is as simple as:
unshare -n program ...
This creates an empty network namespace for the process. That is, it is run with no network interfaces, including no loopback. In below example we add -r to run the program only after the current effective user and group IDs have been mapped to the superuser ones (avoid sudo):
unshare -r -n ping google.com
nftables can set firewall rules based on cgroups (it can be bound to each program), the syntax:
socket cgroupv2 level NUM PATH
'PATH' is the one in
/sys/fs/cgroup, note that the 'PATH' is relative, so you should exclude
/sys/fs/cgroup/ from the absolute path. For example,
/sys/fs/cgroup/system.slice/nginx.service is the cgroup for nginx and its 'PATH' should be
Most systemd services are confined in cgroups. Use
systemd-cgls to view the cgroup tree and use
systemd-cgls -u XXX.service to get the cgroup for a specific service. For example,
systemd-cgls -u nginx.service returns
Note that you should remove the first / , so the 'PATH' should be
system.slice/nginx.service. If your path contains the symbol '@' please use
nft --interactive to set up your firewall rules, other wise, the symbol would be misinterpreted.
Although cgroups will be setup for every systemd unit automatically, you can modify or create systemd unit files to fine tune the cgroup settings if you have special needs. Theoretically, you can set up rules for all programs in a systemd enabled operating system. See systemd.unit(5) and systemd.slice(5) for more information.
Even the most basic systemd unit is attached to a cgroup:
[Unit] Description="My Program" [Service] ExecStart="/usr/bin/my_program"
If your do not use systemd, you can also add program to cgroups manually. See cgroups kernel documentation for more info.
'NUM' indicates the level of cgroups that will be matched in the firewall rule.
In our previous example,
socket cgroupv2 level 2 "system.slice/nginx.service" matches exactly the nginx service. While
socket cgroupv2 level 1 "system.slice/nginx.service" matches all the cgroups under "system.slice/", thus not only nginx but other programs in that directory will also be matched, such as "snapd.service".
A very basic example:
To prevent nginx from accessing the Internet (assume you already have a inet table named 'filter' and a chain named 'output'):
nft add rule inet filter output socket cgroupv2 level 2 "system.slice/nginx.service" drop