Introduction
Outlined here are the minimal security steps the Bonsa Framework uses in server builds.
Allow staff Group to sudo
Use account names associated with a user. For this example, I will be designating adminstrative users under the staff group. To see the list of groups available to you use the command, cat /etc/group.
These steps are assuming that you are logged in with the initial Ubuntu user which as sudo access. If you are using root instead, then you do not need to prefix any command with sudo.
The staff group by default does not normally have sudo access. To grant sudo access to the staff group,
sudo visudo
visudo launches your default editor to a special file. Add the following to the bottom of the file,
# Members of the staff group may gain root privileges %staff ALL=(ALL) ALL
Create Catch-All serveradmin user
The purpose of serveradmin is the catch-all place to setup things like scripts. It may also, depending on requirements for your organization be used to manually setup software like application servers.
Further along these lines, in a more security sensitive environment rather than just one serveradmin, we could setup distinct accounts for running a manual setup of tomcat, manual setup of postgressql etc.
Also, the serveradmin account is limited in that it can not use sudo. If an attacker compromises the application, sudo is still out of reach.
Add the user and assign a password to that user,
sudo addgroup --gid 3000 serveradmin sudo useradd -d /home/serveradmin -m -g serveradmin -u 3000 -c "Admin catch-all" -s /bin/bash serveradmin sudo passwd serveradmin
Create Staff Users
We will also create staff users associated with the built in staff group so we know who is working on the machine. As a policy, our team requires that unless absolutely necessary, staff log in as their own account and then su to serveradmin or use sudo for maintenance work. That way we can have a trail of who does what.
sudo useradd -d /home/bhitch -m -g staff -u 2000 -c "Support Bryan Hitch" -s /bin/bash bhitch sudo useradd -d /home/jcassaday -m -g staff -u 2001 -c "Support John Cassaday" -s /bin/bash jcassaday sudo useradd -d /home/wellis -m -g staff -u 2002 -c "Support Warren Ellis" -s /bin/bash wellis
When adding an existing user to an existing group the user must log out and log back in for changes to take effect.
Notice the -u which set's the user's GUIDs. We found it essential to standardize on the GUID of the accounts across all our systems consistently. Not doing so causes problems when it comes to cloning systems or moving programs across different environments. As a practice, we use the following GUID's ranges,
- Staff 2000-2499
- Guest Staff Users 2500-2999
- Custom services 3000 - 3999
Additionally, we use the following group GUID ranges,
- Custom services 3000 - 3999
- I further broke up my custom services with Web Groups for website management - 3100-3199
Next, we add to the Staff users the following groups,
- adm - so staff can view logs in apps setup without having to use the sudo command
Here is the command,
sudo usermod -a -G adm bhitch sudo usermod -a -G adm jcassaday sudo usermod -a -G adm wellis
The above step could have been done on user create. However, this illustrates user modification as part of the tutorial.
Do not forget to set a passwords for the new accounts,
sudo passwd bhitch Enter new Unix password: Reenter new Unix password: passwd: password updated successfully sudo passwd jcassaday sudo passwd wellis
At this point it is important to logout and log in with your staff account to continue your work.
Disable Direct Login as Root Through SSH
Normally having permit root login in ssh in Ubuntu is not a security issue. Root is simply disabled in the OS. This is a hosted Ubuntu only step where often the root account is enabled. This is dangerous because there are attackers out there looking for Unix/Linux boxes and trying to login via ssh using the username root and then a list of common passwords.
I do not going down the route of disabling the root account as this might break the hosted Ubuntu setup. For example, Slice's or Rackspace special terminal console login might stop working. In any event, the vector of attack is SSH login. To prevent users from using root, well don't provide the root password and provide sudo privileged accounts as shown in this article.
Connect to SSH as a staff user and edit sshd_config,
sudo nano /etc/ssh/sshd_config
Search for the line "PermitRootLogin yes" and change to "PermitRootLogin no". You can still issue su to go in as root but only after logging in as a user belonging to the admin group.
Last restart the SSH service for the changes to take effect.
sudo service ssh restart
In older versions of Ubuntu (to determine) where Upstart is not available use,
sudo /etc/init.d/ssh restart
Prevent SSH Brute Force Dictionary Attacks
As soon as it is on the Internet people will try to brute force attack your server over ssh. Basically they keep on pounding your system trying different passwords.
Just look in /var/log/auth.log to see some attackes,
cat /var/log/auth.log | grep "Invalid user" Jun 19 18:18:33 myra sshd[29346]: Invalid user oracle from 210.83.86.139 Jun 19 18:18:36 myra sshd[29349]: Invalid user test from 210.83.86.139 Jun 19 18:19:02 myra sshd[29381]: Invalid user kylix from 210.83.86.139 Jun 19 18:19:09 myra sshd[29387]: Invalid user www from 210.83.86.139
Fail2ban makes this kind of attack not worthwhile. After a chosen number of failed login attempts from the same ip address, fail2ban blocks that ip address for a set period of time. As constantly changing ip addresses is not a trivial task, the attacker will move on to another system.
sudo apt-get install fail2ban
The fail2ban installer also starts fail2ban as a service right after installation completes.
Most of the how fail2ban works is in /etc/fail2ban/jail.conf and here are the highlights,
maxretry = 6 # under the ssh section you are allowed 6 retries} bantime = 600 # 600 seconds = 10 minutes ignoreip = 127.0.0.1 # do not block list, and CIDR list
The default settigs of fail2ban are usually good enough but you can also customize fail2ban to suit your needs.
After a day or so on the Internet you should start seeing people getting banned in the logs, /var/log/fail2ban.log. Here is an example of an ip getting banned and then after 10 minutes it unbans,
2009-02-15 10:29:24,108 fail2ban.actions: WARNING \[ssh\] Ban 59.63.25.158 2009-02-15 10:39:24,137 fail2ban.actions: WARNING \[ssh\] Unban 59.63.25.158
Unbanning
To unban a user try these instructions. I am hesitant about playing with the ip tables in any way, so I have not tried myself. I usually just wait the 10 minutes.
According to the developers, Fail2ban version 0.9 will include an unban command through it's own client program.
More
This is basic Ubuntu Security. There is more.
This article can be enhanced with a strategy and write up on the following triggers,
- Alert me when a sudo capable account logs in.
- Alert me when serveradmin logs in.
- Alert me when accounts fail sudo attempts.