Securing your Node

The goal of this guide is to walk you through steps you can take to secure your node against malicious actors. Whether you're running a local server at home or a VPS server / virtual machine on the cloud, the tips here will help you harden your node against outside attack and help protect it during its lifespan.

This section will describe both essential actions, which you must take, and nice-to-have actions, which are helpful but not required.

NOTE

This guide is meant to be an introduction to some of the things you can do to harden your node machine. If you are comfortable with the command-line terminal and want to go even further in protecting your node, take a look at the popular imthenachoman/How-To-Secure-A-Linux-Server guide.

Assumptions in This Guide

This guide assumes your node runs Ubuntu 20.04 LTS. The concepts will carry over to other systems but the example commands may not.

As with all of the commands in this guide, we assume that you are connecting remotely to your node's command terminal using ssh. If you need a refresher on how to use ssh, take a look at the Intro to Secure Shell guide first.

ESSENTIAL: Keep your Client Machine Secure

NOTE

If you use your Smartnode locally (by physically logging into it with a keyboard and monitor directly attached to it), then this section is not relevant to you - you can skip it.

Most Smartnode operators interact with their node remotely by connecting to its terminal from another computer using ssh:

  • The machine you connect to (in this case, your node machine) is called the server.
  • The machine you connect from (such as your laptop, desktop, or even your phone) is the client.

One of the most important things you can do to secure your Smartnode is to keep your client machine secure. If your client machine is compromised and you use it to log into your node, then most of the security settings you apply to the node can be bypassed.

For example: if you use a laptop as an SSH client, and it has a keylogger installed, then any secret things you type on the node (such as your password or recovery mnemonic) when connected via SSH will be stolen.

There is no definitive guide to keeping your client machine secure, but being aware that it is a factor in your security is a good first step. Make sure that your client machine is as secure as it can be.

Here are a few tips:

  • Don't use your client machine for risky activities (such as visiting untrustworthy websites or installing unnecessary programs)
  • Keep your client machine updated with the latest security patches
  • If possible, use a malware and antivirus protection program for your Operating System

For maximum security, you may want to use a dedicated machine as your SSH client, though this may not be practical for you.

ESSENTIAL: Secure your SSH Access

NOTE

If you use your Smartnode locally (by physically logging into it with a keyboard and monitor directly attached to it), then this section is not relevant to you - you can skip it.

Whether you run your Smartnode at home or you use a VPS in a remote datacenter, it is likely that either you access it through SSH, or that SSH is enabled even if you do not use it.

SSH connections are based on secure cryptography, but as with any secure system, the real security comes from using it correctly. There are two main things to do for your SSH settings:

  1. Use an SSH key to log in remotely instead of a username and password
  2. Disable password-based authentication entirely, so SSH keys are the only remote login option

As you are probably familiar with now, the default way to log into your node via SSH is with a username and password. The downside of this is that your password is typically something rather "short" and susceptible to brute-force attacks.

Luckily, there is an alternative way to log in via SSH: an SSH key pair.

SSH key pairs work similarly to blockchain wallets; they come with a public part (such as your wallet address) and a private part (the private key for your wallet address):

  • You provide the public part to your node. This way, the node knows you're allowed to connect to it, and it knows that it's really you trying to connect.
  • You keep the private part to yourself on your client machine. This way, you (and only you) can connect to your node.
    • You can (and should!) protect the private part with a password, so someone who steals your key can't use it.
  • From a computer's perspective, the private key is exponentially harder to crack than a password is. This mitigates the risk of a brute-force attack against your node.
TIP

If you'd like to learn more about SSH key pairs before creating your own, take a look at these links:

Creating an SSH Key Pair

Let's start by creating a new SSH key pair on your client machine. There are many varieties of key out there, but we're going to use a key type called ed25519 which provides excellent security.

Run the following command on your client machine (i.e., you should not run this while already SSH'd into your node machine - if you are, exit out of SSH first):

ssh-keygen -t ed25519 -C "[email protected]"

You will see the following:

Generating public/private ed25519 key pair. Enter file in which to save the key (/home/username/.ssh/id_ed25519):

This asks you where you would like to save your private key file. SSH is compatible with the provided default and will automatically use it for you if you select it. However, you have the option of changing it to something else if you wish.

::: tip NOTE The path /home/username/.ssh/id_ed25519 is just an example, assuming your username is username. You likely have a different username. Whenever you see a path like the above in this guide, replace it with whatever path your system actually prints with your actual username.

If you are comfortable with the default setting, simply press Enter.

Otherwise, type your desired location for the key. It must be an absolute path (e.g. /home/username/.ssh/rocketpool_key on Linux, or /Users/username/.ssh/rocketpool_key on OSX). Press Enter when you're done.

After pressing Enter, you will see:

Enter passphrase (empty for no passphrase):

This will become the password for the private key itself. Whenever you use the key to connect to your node, you will need to enter this password first.

WARNING

You should not leave this blank - otherwise, anyone with the SSH key file will be able to use it! Pick a good password that you (and only you) will know.

Also, don't forget your password - there is no way to recover this password if you lose it.

Once you finish typing the password, press Enter. It will ask you to retype it for confirmation.

After that, you will see something like the following output:

Your identification has been saved in /home/username/.ssh/id_ed25519 Your public key has been saved in /home/username/.ssh/id_ed25519.pub The key fingerprint is: SHA256:CASbPZETiQ83lLhpUO2aoT05TxMVLwqiWtdsRtoPt4s [email protected] The key's randomart image is: +--[ED25519 256]--+ | .o*==.. | |. +=O... | |..+B++o . | |..=.+X o | |.+.=+.O S | |o.B.oo + . | |. = . o | | . . . | | E . | +----[SHA256]-----+

The first line states the location of the private key, which is called id_ed25519 by default (notice that it does not have a file extension). Ubuntu will load this key for you automatically when you use ssh if this private key file is in the default location.

The second line states the location of the public key, which is called id_ed25519.pub by default. We'll need the public key for the next step.

NOTE

Ubuntu should load this new key automatically. However, some systems (such as macOS machines) will not load it automatically - you will have to tell it to do this with the following command on your client machine:

ssh-add $HOME/.ssh/id_ed25519

Note that this is the path of the private key that we generated in the previous step, not the public key. Replace the path with whatever your system printed in that previous step.

If you get an error saying that the ssh-agent is not running, start it by running the following command on your client machine:

eval $(ssh-agent)

If you don't want to type these two commands every time you open the terminal, you can create a shortcut for adding your key by adding an alias to your ~/.bashrc file.

Open the file using the text editor:

nano ~/.bashrc

Add this line to the end (assuming you used the default path for the private key - update as necessary):

alias loadkey='ssh-add $HOME/.ssh/id_ed25519'

Save and exit with Ctrl+O and Enter, then Ctrl+X. Next, close and open your terminal for the changes to take effect.

You can now type loadkey on your client machine to load the key.

Adding the Public Key to your Node

Once you have your SSH key pair, you can now add the public key to your node. This will let you connect to it over ssh using the private key you just generated, instead of your username and password.

There are two ways to do this - if one doesn't work, try the other way:

Using ssh-copy-id
Manually Adding the Key

Note: if your client machine is running Windows, ssh-copy-id is not yet available. Please follow the instructions in the "Manually Adding the Key" tab.

Run the following command on your client machine:

ssh-copy-id -i $HOME/.ssh/id_ed25519.pub [email protected]

For example, if my username on the node was staker and my node's IP address was 192.168.1.10, I would run the following command:

ssh-copy-id -i $HOME/.ssh/id_ed25519.pub [email protected]

You will see some messages like the following:

/usr/bin/ssh-copy-id: INFO: Source of key(s) to be installed: "/home/username/.ssh/id_ed25519.pub" /usr/bin/ssh-copy-id: INFO: attempting to log in with the new key(s), to filter out any that are already installed /usr/bin/ssh-copy-id: INFO: 1 key(s) remain to be installed -- if you are prompted now it is to install the new keys

This tells you that it's trying to log in with your key first to make sure it isn't already there. Once it fails logging in, it knows that it's OK to add the new public key to the node machine.

It will then prompt you for the password of the user on your node machine. (Note that this is not the password of the SSH key!)

Enter your user's password, and you will see the following output:

Number of key(s) added: 1 Now try logging into the machine with: "ssh '[email protected]'" and check to make sure that only the key(s) you wanted were added.

That means it worked!

You should now be able to ssh into the node like you normally would, but now you won't have to type the password of the user account.

Instead, you will have to type the password of your SSH private key. Depending on your system settings, you may only have to do this once per restart, or you may have to do it every time you use the key to connect to your node.

Disable Login-via-Password

Even though you have an SSH key pair set up, your node will still allow other machines to try to log in using the username and password method. This defeats the entire purpose of using SSH keys in the first place, so the next step is to disable those.

NOTE

You are about to modify the SSH server's configuration.existing SSH sessions will be preserved. However, if you make a mistake, then it's possible that you will not be able to create new SSH sessions anymore and effectively lock yourself out of the machine.

All of your

To prevent this, we strongly recommend that you create 2 SSH sessions for the next steps - one for editing things and testing, and one as a backup so you can revert any breaking changes.

Start by logging into your machine using ssh as usual:

As a reminder, you should do this twice on two separate terminals so you have a backup session just in case. You can ignore the backup session for now - we'll tell you when you need it. Run the following commands only in the first session.

Open the configuration file for the SSH server:

sudo nano /etc/ssh/sshd_config

As with all commands that start with sudo, this will prompt you for your user account's password. This is a large file, so you'll have to navigate through it using the arrow keys on your keyboard or Page Up / Page Down.

Make the following changes:

  1. Uncomment #AuthorizedKeysFile if it is commented (by removing the # in front of it)
  2. Change KbdInteractiveAuthentication yes to KbdInteractiveAuthentication no and uncomment (by removing the # in front of it) - note that older versions of SSH call this option ChallengeResponseAuthentication instead of KbdInteractiveAuthentication
  3. Change PasswordAuthentication yes to PasswordAuthentication no and uncomment (by removing the # in front of it)
  4. Change PermitRootLogin yes to PermitRootLogin prohibit-password unless it's already set to that and has a # in front of it

Once you're done, save with Ctrl+O and Enter, then exit with Ctrl+X.

Finally, run sudo sshd -T | grep -i passwordauthentication and make sure that it prints passwordauthentication no.
If it does not, you may need to run sudo nano /etc/ssh/sshd_config.d/50-cloud-init.conf and set PasswordAuthentication yes to PasswordAuthentication no in that file as well. Save and exit as before, with Ctrl+O and Enter, then Ctrl+X

Next, restart the SSH server so it picks up the new settings:

sudo systemctl restart ssh.service

After this, logging into SSH via a username and password should be disabled.

NOTE

At this point, you should exit the SSH session and try to SSH back in.

If you are able to do so successfully, then your SSH configuration is still valid!

If you are not able to get back in, then something has gone wrong with your configuration. Use the backup SSH session you created at the start of this section to modify the /etc/ssh/sshd_config file.

Try to find the mistake or undo your changes, then restart the SSH server using sudo systemctl restart sshd.

Once it's been restarted, try connecting with SSH again on your "other" terminal. Keep doing this until you have it working again and are able to successfully connect.

(Optional) Enable Two-Factor Authentication

Two-factor authentication involves requiring a second security measure in addition to your password or SSH key, usually on a separate device from your primary one.

For example, you may be familiar with logging into a website such as a crypto exchange using both a password and a Google Authenticator code (or an SMS code). This two-step process is an example of two-factor authentication.

SSH can also be configured to require a Google Authenticator code, which means that an attacker that somehow compromised your SSH key and its passphrase would still need the device with the authenticator app on it (presumably your phone). This adds an extra layer of security to your system.

WARNING

We strongly recommend that you open a second terminal with an SSH connection to your node, just in case you misconfigure something. This way, you will have a backup that is still connected in case you lock yourself out, so you can easily undo your mistakes.

If you do manage to lock yourself out, you will need to physically access your node via its local monitor and keyboard to log in and repair the misconfiguration.

Start by installing Google Authenticator (or a compatible equivalent) on your phone if you don't already have it. For Android users, consider andOTP which is an open-source alternative that supports password locking and convenient backups.

Next, install the Google Authenticator module on your node with this command:

sudo apt install -y libpam-google-authenticator

Now tell the PAM (pluggable authentication modules) to use this module. First, open the config file:

sudo nano /etc/pam.d/sshd

Find @include common-auth (it should be at the top) and comment it out by adding a # in front of it, so it looks like this:

# Standard Un*x authentication. #@include common-auth

Next, add these lines to the top of the file:

# Enable Google Authenticator
auth required pam_google_authenticator.so

Then save and exit the file with Ctrl+O, Enter, and Ctrl+X.

Now that PAM knows to use Google Authenticator, the next step is to tell sshd to use PAM. Open the sshd config file:

sudo nano /etc/ssh/sshd_config

Now change the line KbdInteractiveAuthentication no to KbdInteractiveAuthentication yes so it looks like this:

# Change to yes to enable challenge-response passwords (beware issues with # some PAM modules and threads) KbdInteractiveAuthentication yes

(Older versions of SSH call this option ChallengeResponseAuthentication instead of KbdInteractiveAuthentication.)

Add the following line to the bottom of the file, which indicates to sshd that it needs both an SSH key and the Google Authenticator code:

AuthenticationMethods publickey,keyboard-interactive:pam

Then save and exit the file with Ctrl+O, Enter, and Ctrl+X.

Now that sshd is set up, we need to create our 2FA codes. In your terminal, run:

google-authenticator

First, it will ask you about time-based tokens. Say y to this question:

Do you want authentication tokens to be time-based: y

You will now see a big QR code on your screen; scan it with your Google Authenticator app to add it. You will also see your secret and a few backup codes looking like this:

Your new secret key is: IRG2TALMR5U2LK5VQ5AQIG3HA4 Your verification code is 282436 Your emergency scratch codes are: 29778030 86888537 50553659 41403052 82649596
NOTE

Record the emergency scratch codes somewhere safe in case you need to log into the machine but don't have your 2FA app handy.

Without the app, you will no longer be able to SSH into the machine!

Finally, it will ask you for some more parameters; the recommended defaults are as follows:

Do you want me to update your "/<username>/.google_authenticator" file: y Do you want to disallow multiple uses of the same authentication token: y By default... < long story about time skew > ... Do you want to do so: n Do you want to enable rate-limiting: y

Once you're done, restart sshd so it grabs the new settings:

sudo systemctl restart sshd

When you try to SSH into your server with your SSH keys, you should now also be asked for a 2FA verification code, but not for a password.

ESSENTIAL: Enable Automatic Security Updates

Operating System vendors routinely publish updates and security fixes, so it is important that you keep your system up to date with the latest patches. The easiest way to do this is to enable automatic updates.

Run the following commands on your node machine:

sudo apt update
sudo apt install -y unattended-upgrades update-notifier-common

You can change the auto-update settings by editing /etc/apt/apt.conf.d/20auto-upgrades:

sudo nano /etc/apt/apt.conf.d/20auto-upgrades

This is an example of reasonable auto-update settings:

APT::Periodic::Update-Package-Lists "1";
APT::Periodic::Unattended-Upgrade "1";
APT::Periodic::AutocleanInterval "7";
Unattended-Upgrade::Remove-Unused-Dependencies "true";
Unattended-Upgrade::Remove-New-Unused-Dependencies "true";

# This is the most important choice: auto-reboot.
# This should be fine since Rocketpool auto-starts on reboot.
Unattended-Upgrade::Automatic-Reboot "true";
Unattended-Upgrade::Automatic-Reboot-Time "02:00";

When you are done adding your changes, save with Ctrl+O and Enter, then exit with Ctrl+X.

After, make sure to load the new settings:

sudo systemctl restart unattended-upgrades

ESSENTIAL: Enable a Firewall

In general, your machine should only accept network traffic on ports that your Execution client, Consensus client, and Smartnode stack use. To enforce that and prevent any unexpected or undesirable traffic, we can install a firewall on the node.

NOTE

If you selected a different execution/consensus client port during the Rocketpool setup, you need to edit the ports below to reflect your settings.

Ubuntu comes with ufw installed by default (the uncomplicated fire wall), which is a convenient utility for managing your node's firewall settings.

The following commands will set ufw up with a good default configuration for your Smartnode. Run these on your node machine.

Disable connections unless they're explicitly allowed by later rules:

sudo ufw default deny incoming comment 'Deny all incoming traffic'

Allow SSH:

sudo ufw allow "22/tcp" comment 'Allow SSH'

Allow execution client (formerly referred to as ETH1):

sudo ufw allow 30303/tcp comment 'Execution client port, standardized by Rocket Pool'
sudo ufw allow 30303/udp comment 'Execution client port, standardized by Rocket Pool'

Allow consensus client (formerly referred to as ETH2):

sudo ufw allow 9001/tcp comment 'Consensus client port, standardized by Rocket Pool'
sudo ufw allow 9001/udp comment 'Consensus client port, standardized by Rocket Pool'

If you run lighthouse client v4.5.0+, you can use quic protocol to reduce latency / increase bandwitdh, quic protocol uses lighthouse's --port + 1 to listen for quic messages by default: https://lighthouse-blog.sigmaprime.io/Quic,%20Networking.html

sudo ufw allow 8001/udp comment 'Consensus client port, standardised by Rocket Pool'

Finally, enable ufw:

sudo ufw enable
NOTE

iptables experts might note that Docker bypasses ufw settings. Strictly speaking, that means unless you are running in Hybrid mode, you do not need the Execution and Consensus client rules. Adding them, however, has no downside and will make sure that if you ever switch to Hybrid mode you will not run into firewall issues.

(Optional) Enable Brute-Force and DDoS Protection

To protect your server from DDoS attacks and brute-force connection attempts, you can install fail2ban. This program will monitor incoming connections and block IP addresses that try to log in with faulty credentials repeatedly.

See this guide for more information on intrusion prevention.

Run the following commands on your node machine:

Install the service:

sudo apt install -y fail2ban

Next, open /etc/fail2ban/jail.d/ssh.local:

sudo nano /etc/fail2ban/jail.d/ssh.local

Add the following contents to it:

[sshd] enabled = true banaction = ufw port = 22 filter = sshd logpath = %(sshd_log)s maxretry = 5

You can change the maxretry setting, which is the number of attempts it will allow before locking the offending address out.

Once you're done, save and exit with Ctrl+O and Enter, then Ctrl+X.

Finally, restart the service:

sudo systemctl restart fail2ban

And with that, you've just greatly improved the security posture of your node. Congratulations!