last modified: 2017-10-11
Server ordered on Hetzner.de (based in Germany, dirt cheap, but without management.)
Remember to install the Linux version not from the rescue system in the console but from https://robot.your-server.de/server/index in the "Linux" tab.
(installing from the rescue system provided with the bare server causes a ssh key mess)
I use Debian, version 8.7 (why?)
Vi is used as a text editor in the following
we are logged as root first
apt-get update && sudo apt-get upgrade
→ refreshes the repositories and fetches information about packages that are available online.
→ downloads and installs updates for all installed packages - as long as it doesn’t bother dependencies (install new packages, remove old ones or crosses a repo source (switch a package from one repo to another)).
aptitude install ntp
Then define your time zone (the one where your server is located):
This step helps when your server needs to be synchronized with other servers.
The kernel is the software at the closest of the machine: it is provided by the Linux distribution you use.
A configuration file offers parameters which tune the kernel to make things harder for an intruder. Here I rely exactly on the tutorial by Pontikis:
Create a new file, so as to preserve / not to mess up the original file:
//ST: ! - Paste the contents of link:resources/kernel_config.txt[this file]: - Close the file - reboot the server
Add this line if not already present:
Then, rebuild aliases:
By default, loggging to the server via SSH is done on the port 22. Knowing that, attackers scan the port 22. Changing the port to a different one makes the attacker’s job more difficult. To do that:
Text to change in the file: change port SSH 22 by a new port (let’s say 1234), write the new port down somewhere
service sshd restart
We should now disable root login via SSH. Why? Because attackers would know that a "root" user is available to log in, and it just remains to attack its password.
With the root user disabled at the SSH login step, the attackers must guess both the username and its password to access the connection, and that’s much harder.
Of course, an attacker who aims at you or your server specifically (a "targeted" attack) would expect a series of usernames (in my case "seinecle", the name I use on all social media), so don’t use it either.
So the logic is the following: we will create a user with much less priviledges than the root user. Only this weak user will have the right to connect to the server.
The user will be "enough" for regular tasks on the server. If we need the admin rights of root to install stuff or else, we will temporarily switch from this weak user to root to execute what we need, but then revert back to this weak user.
This way, we limit greatly the exposure of root privileges to the outside.
making sure we have installed the "sudo" command that will allow us to switch from a weak user to root.
creating a weak user
giving rights to this user to establish a connection to the server (not just act on it once logged)
removing the rights of root to connect to the server.
apt-get install sudo
Have a strong password ready
adduser myUser -shell /bin/bash adduser myUser sudo
text to add to this file sshd_config:
Then restart the SSH service:
service sshd restart
Text to change in the file:
From there on, you cannot login to the server from root, only from myUser!
Let’s try it. Create a new SSH session with myUser. Then:
Switch to root privileges:
(you must enter the root password at this step)
Password authentication is less secure than SSH public key. A password transits through the Internet for the auhtentication, it can be hacked at this step.
A SSH private key is not transmitted on the wire. So, it can’t be hacked this way.
A detailed explanation is available here.
Logging through SSH rather than passwords can be hair rising because there are so many tiny details that can go wrong. There is a good chance that if you do it for the first time you will lock yourself outside the server.
So, do this when you can still erase the server, of if you are confortable waiting that your provider will unlock it for you.
LogLevel DEBUG3 (this should be added, the parameter is not listed by default)
Save the file, then:
service sshd restart
In your user home folder:
mkdir ~/.ssh chmod 700 ~/.ssh cd ~/.ssh vi authorized_keys
//ST: ! If you already have a .ssh directory, how to find it and the file `authorized_keys` in it? The `.ssh` directory is *hidden by default* because it starts with a `.`
To find it, you need to navigate with root privileges directly to the
authorized_keys file, like this:
//ST: ! Things to check:
make sure you have put the public key in the .ssh folder of the user in /home/myUser/.ssh/authorized_keys (not in the .ssh folder of the root user)
make sure your key starts with "the "ssh-rsa" (with a space after it, check the first "s" might be missing …)
triple check the key doesn’t break in several lines
service sshd restart after each modif to load your new ssh key
Your private key will probably not be recognized the first time because of some problems above not completely fixed.
Keep trying to log with your SSH key. To find the cause of your issues, inspect the log for auth operations:
tail -f /var/log/auth.log
Some useful answers to questions from developers lost in making SSH keys works:
/etc/ssh/sshd_config, you can disable password authentification:
service sshd restart
Now only connecions via a public / private key is possible.
A firewall gives you control on what can enter and leave your server.
But there is an even easier alternative.
apt-get install ufw
The firewall is now installed, but is is not active yet.
We add a rule to block all incoming traffic, except for SSH connections through the port we defined: ufw default deny incoming ufw allow 1234/tcp
Now, we can activate the firewall
this part builds on: http://www.pontikis.net/blog/psad-install-config-debian-wheezy
Psad is an app which bans users which scan ports. Before installing it, we need to make sure the firewall logs traffic:
iptables -A INPUT -j LOG iptables -A FORWARD -j LOG
Then we install Psad:
apt-get install psad
Now we configure Psad by modifying this file:
Possible values for some interesting parameters (and the source for this section), are here:
Then we must edit this file to add the address of the server to the whitelist:
where I put just 2 values:
127.0.0.1 0; # localhost xx.xx.xxx.xxx 0; # Server IP (replace xx.xx.xxx.xxx by your actual server IP)
Restart psan with this config:
sudo psad --sig-update sudo service psad restart
This is an app which bans users which fail to login after a number of times - typically bots trying to break in.
fail2ban can actually scan logs from a list of apps you decide (MongoDB, Apache server, GlassFish, etc.) and ban ips mentioned in logs showing a failed access. You need to setup a regex rule specific for each log format, though.
Documentation on failtoban: http://www.pontikis.net/blog/fail2ban-install-config-debian-wheezy
This is an application running on your machine, generating security audits and making suggestions.
apt-get install lynis
Run it: (from any directory)
lynis audit system
The report will appear on screen (hit Enter to move on), and in this file:
Tutorial by Clement Levallois.
Discover other tutorials and courses in data / tech for business: http://www.clementlevallois.net
Or get in touch via Twitter: @seinecle