Create more secure ssh keys

Create a key using elliptic curve cryptography (more secure):

ssh-keygen -a 100 -t ed25519

generate RSA key of length 4096 to file my_key

ssh-keygen -t rsa -b 4096 -C "" -f my_key

Generate md5 fingerprint of the key (works in newer ubuntu, 16):

ssh-keygen -lf ./my_key -E md5

see also: Upgrade Your SSH Key to Ed25519

How to prevent SSH from scanning all the keys in .ssh directory

$ ssh -o IdentitiesOnly=yes -F /dev/null -i ~/path/to/some_id_rsa

The options are as follows:

  • -o IdentitiesOnly=yes - tells SSH to only use keys that are provided via the CLI and none from the $HOME/.ssh or via ssh-agent
  • -F /dev/null - disables the use of $HOME/.ssh/config
  • -i ~/path/to/some_id_rsa - the key that you explicitly want to use for the connection


No strict server checking

To disable confirmation about unknown server fingerprints add this to command line: -o StrictHostKeyChecking=no

Set correct permissions for .ssh directory

You may get and error while connecting to ssh if you have wrond permissions for .ssh directory on server for a given user.

Error message on client while ssh user-on-server@server:

debug1: Authentications that can continue: publickey
debug1: Next authentication method: publickey
debug1: Offering RSA public key: /home/user-on-client/.ssh/YOUR_KEY
debug1: Authentications that can continue: publickey
debug1: No more authentication methods to try.
Permission denied (publickey).

Error message on server:

Aug 11 10:55:21 comp sshd[3446]: Connection closed by port 47794 [preauth]

And nothing more in logs. The reason was that .ssh directory on the server had too strict permissions:

dr-------- 2 user-on-server user-on-server 4096 Aug 11 10:54 .ssh/

Read only is not enough. It must be rwx:

chmod 700 .ssh

For authorized_keys file use the following premissions:

chmod 644 ~/.ssh/authorized_keys 

Port forwarding

Remote port forwarding

Take all the connections to of the remote machine ( and forward it to localhost:80 on the local machine:

ssh -R -N

The command above creates a general IPv4-only bind, which means that the port is accessible on all interfaces via IPv4.

This version binds to all interfaces individually:

ssh -R \*:8080:localhost:80 -N

Another option:

ssh -R "[::]:8080:localhost:80" -N

The third version is probably technically equivalent to the first, but again it creates only a single bind to ::, which means that the port is accessible via IPv6 natively and via IPv4 through IPv4-mapped IPv6 addresses (doesn’t work on Windows, OpenBSD). (You need the quotes because [::] could be interpreted as a glob otherwise.)

Full list of specifications for the remote forwarding.

     -R [bind_address:]port:host:hostport
     -R [bind_address:]port:local_socket
     -R remote_socket:host:hostport
     -R remote_socket:local_socket

Specifies that connections to the given TCP port or Unix socket on the remote (server) host are to be forwarded to the given host and port, or Unix socket, on the local side.

By default, TCP listening sockets on the server will be bound to the loopback interface only. This may be overridden by specifying a bind_address. An empty bind_address, or the address ‘*’, indicates that the remote socket should listen on all interfaces. Specifying a remote bind_address will only succeed if the server’s GatewayPorts option is enabled (see sshd_config(5)).

If the port argument is ‘0’, the listen port will be dynamically allocated on the server and reported to the client at run time.
When used together with -O forward the allocated port will be printed to the standard output.

When bind_address is omitted (as in your example), the port is bound on the loopback interface only.

Note that if you use OpenSSH sshd server, the server’s GatewayPorts option needs to be enabled (set to yes or clientspecified) for arbitrary interface to work (check file /etc/ssh/sshd_config on the server). Otherwise (default value for this option is no), the server will always force the port to be bound on the loopback interface only.

See also: 1, 2

Local port forwarding

ssh -L localhost🔢localhost:5678 user@

With this command

  • all the connections to the localhost:1234 of the host machine
  • are forwarded to localhost:5678 at the server (

Port forwarding between two machines that don’t have external ip’s using a server

On machine 1:

ssh -L localhost:10002:localhost:10001 user@

On machine 2:

ssh -R localhost:10001:localhost:10000 user@

Now when I access localhost:10002 on machine 1 the request is redirected to localhost:10000 of machine 2. The connection goes through port 10001 of host

config for multiple port forwarding

How to forward multiple ports with one command. It is possible with a config file.

Imagine we want to forward local ports 22, 9999, 9998 to the corresponding ports of machine using ssh on machine In order to do that add the following to ~/.ssh/config:

Host all-port-forwards
  User user
  LocalForward 22
  LocalForward 9999
  LocalForward 9998

Now you can run a single command:

ssh all-port-forwards

Port forwarding on system startup with retry

We want our port to be forwarded on the startup of the system. Also we want to deal with failures: retry on disconnect etc.

Initial version - it doesn’t work

In my experience the approach didn’t really work. The built-in mechanisms of systemctl didn’t allow me to achieve robust restarting.

Create a file /etc/systemd/system/tunnelssh.service:

Description=tunnel ssh to server

ExecStart=/usr/bin/ssh -NT -o ServerAliveInterval=60 -o ExitOnForwardFailure=yes -i PATH_TO_YOUR_PRIVATE_KEY -R REMOTE_SERVER_INTERFACE:PORT_ON_REMOTE_SERVER:localhost:LOCAL_PORT user_at_remote_server@REMOTE_SERVER



Then enable and start

systemctl enable tunnelssh.service
systemctl start tunnelssh.service
systemctl status tunnelssh.service

Unfortunately it doesn’t work on startup. It says that the unit entered failed state and doesn’t restart. The problem is that

  1. our service may start before network connection is up. You may add or as suggested here but it doesn’t save you from the next issue.
  2. our server can be down for some time
  3. retry mechanism of systemd is somehow broken. I couldn’t make it work with configuration of burst intervals from here, here and here. And it seems I am not the only one.

(in the spirit of this solution)

Second version

So we have to implement retry on our own without systemd. Fortunately there is a tool autossh. See for example [1] or [2]

Description=tunnel ssh to server

ExecStart=/usr/bin/autossh -M 0 -NT -o ServerAliveInterval=60 -o ExitOnForwardFailure=yes -i PATH_TO_YOUR_PRIVATE_KEY -R REMOTE_SERVER_INTERFACE:PORT_ON_REMOTE_SERVER:localhost:LOCAL_PORT user_at_remote_server@REMOTE_SERVER



Configuring SSH server

Enable only ssh v2:

Protocol 2

Disable password authentication. Replace corresponding values in /etc/ssh/sshd_config:

ChallengeResponseAuthentication no
PasswordAuthentication no


Restrict SSH access to only user accounts that should have it. For example, you may create a group called “sshlogin” and add the group name as the value associated with the AllowGroups variable located in the file /etc/ssh/sshd_config.

AllowGroups sshlogin

Then add your permitted SSH users to the group “sshlogin”, and restart the SSH service.

sudo adduser username sshlogin
sudo systemctl restart sshd.service


Restarting ssh service

after each change in sshd_config you have to restart ssh service to enable the changes

## Ubuntu/debian user ##
sudo service ssh restart
# only for systemd based Ubuntu/Debian 8.x+ users #
sudo systemctl restart ssh
#### RHEL/CentOS/Fedora Linux user type ####
sudo service sshd restart
# only for systemd based RHEL/CentOS v7+ users #
sudo systemctl restart sshd

Add user for ssh tunnel only

useradd sshtunnel -m -d /home/sshtunnel -s /bin/rbash
passwd sshtunnel

in .profile in the home directory of the user (in our example it is /home/sshtunnel/):


Forbid changes:

chmod 555 /home/sshtunnel/
cd /home/sshtunnel/
chmod 444 .bash_logout .bashrc .profile


Restrictions in sshd config:

Match User that-restricted-guy
  AllowTcpForwarding yes
  X11Forwarding no
  AllowAgentForwarding no
  ForceCommand /bin/false


Restricting root user

For security reason you should always block access to root user and group on a Linux or Unix-like systems.

First, make sure at least one user is allowed to use ‘su -‘ or ‘sudo’ command on the server.

Then add to /etc/ssh/sshd_config

DenyUsers root
DenyGroups root

set PermitRootLogin to no:

PermitRootLogin no

Save the file and restart the ssh service.

How To Configure SSH Key-Based Authentication on a Linux Server

See How To Configure SSH Key-Based Authentication on a Linux Server. Should I put public or private key to the server?

Use ssh authentication by key instead of password

Setup the SSH server to use keys for authentication

How to load ssh key without placing it to ~/.ssh

You must load a key to the ssh agent running in the background.

To start the ssh-agent in the background:

eval "$(ssh-agent -s)"
Agent pid 59566

You can add your SSH private key to the ssh-agent using ssh-add and full path to the key:

    ssh-add ~/path_to_your_private_key/id_rsa

Now you can for example do git clone from github using ssh key.

see also 1

mount remote file system using ssh

sudo apt-get install sshfs
sudo mkdir /mnt/droplet
sudo sshfs -o allow_other,IdentityFile=~/.ssh/id_rsa /mnt/droplet

With identity files only:

sudo sshfs -o allow_other,IdentitiesOnly=yes,IdentityFile=~/.ssh/id_rsa user@ /mnt/local/path/


Specific settings for a website

I have a private key with a custom name in ~/.ssh/ add the .pub is uploaded to github. When I try to clone my repo, git returns

Cloning into 'repo'...
Permission denied (publickey).
fatal: Could not read from remote repository.

Please make sure you have the correct access rights
and the repository exists.

Probably git reads only keys with standard names (e.g. id_rsa). To make it pick up your custome key you can create a config file ~/.ssh/config:

  IdentityFile ~/.ssh/your_custom_private_key
  IdentitiesOnly yes

More here

See also

SOCKS proxy

How to run socks proxy. It’s a bit easier than using VPN because it doesn’t require sudo to launch. Also it doesn’t require to install anything on the server except openssh server.

ssh -D 1337 -q -C -N user@your_server.ip

-D defines the port on the local machine,

-q is quiet mode,

-C compresses,

-N causes remote commands not to be executed, just straight port forwarding.

Then one can set up a socks proxy in the settings of the browser.


Faster retry after disconnecting

In order to make recovery faster after network issues use options -o "ServerAliveInterval 2" -o "ServerAliveCountMax 2".

Alternatively one can add the following to ~/.ssh/config:

Host *
  ServerAliveInterval 2
  ServerAliveCountMax 2

See also