Suchen

Dieser Inhalt ist in der von Ihnen ausgewählten Sprache nicht verfügbar.

Chapter 21. Securing networks

download PDF

21.1. Using secure communications between two systems with OpenSSH

SSH (Secure Shell) is a protocol which provides secure communications between two systems using a client-server architecture and allows users to log in to server host systems remotely. Unlike other remote communication protocols, such as FTP or Telnet, SSH encrypts the login session, which prevents intruders to collect unencrypted passwords from the connection.

Red Hat Enterprise Linux includes the basic OpenSSH packages: the general openssh package, the openssh-server package and the openssh-clients package. Note that the OpenSSH packages require the OpenSSL package openssl-libs, which installs several important cryptographic libraries that enable OpenSSH to provide encrypted communications.

21.1.1. SSH and OpenSSH

SSH (Secure Shell) is a program for logging into a remote machine and executing commands on that machine. The SSH protocol provides secure encrypted communications between two untrusted hosts over an insecure network. You can also forward X11 connections and arbitrary TCP/IP ports over the secure channel.

The SSH protocol mitigates security threats, such as interception of communication between two systems and impersonation of a particular host, when you use it for remote shell login or file copying. This is because the SSH client and server use digital signatures to verify their identities. Additionally, all communication between the client and server systems is encrypted.

A host key authenticates hosts in the SSH protocol. Host keys are cryptographic keys that are generated automatically when OpenSSH is first installed, or when the host boots for the first time.

OpenSSH is an implementation of the SSH protocol supported by Linux, UNIX, and similar operating systems. It includes the core files necessary for both the OpenSSH client and server. The OpenSSH suite consists of the following user-space tools:

  • ssh is a remote login program (SSH client).
  • sshd is an OpenSSH SSH daemon.
  • scp is a secure remote file copy program.
  • sftp is a secure file transfer program.
  • ssh-agent is an authentication agent for caching private keys.
  • ssh-add adds private key identities to ssh-agent.
  • ssh-keygen generates, manages, and converts authentication keys for ssh.
  • ssh-copy-id is a script that adds local public keys to the authorized_keys file on a remote SSH server.
  • ssh-keyscan gathers SSH public host keys.

Two versions of SSH currently exist: version 1, and the newer version 2. The OpenSSH suite in RHEL supports only SSH version 2. It has an enhanced key-exchange algorithm that is not vulnerable to exploits known in version 1.

OpenSSH, as one of core cryptographic subsystems of RHEL, uses system-wide crypto policies. This ensures that weak cipher suites and cryptographic algorithms are disabled in the default configuration. To modify the policy, the administrator must either use the update-crypto-policies command to adjust the settings or manually opt out of the system-wide crypto policies.

The OpenSSH suite uses two sets of configuration files: one for client programs (that is, ssh, scp, and sftp), and another for the server (the sshd daemon).

System-wide SSH configuration information is stored in the /etc/ssh/ directory. User-specific SSH configuration information is stored in ~/.ssh/ in the user’s home directory. For a detailed list of OpenSSH configuration files, see the FILES section in the sshd(8) man page.

Additional resources

21.1.2. Configuring and starting an OpenSSH server

Use the following procedure for a basic configuration that might be required for your environment and for starting an OpenSSH server. Note that after the default RHEL installation, the sshd daemon is already started and server host keys are automatically created.

Prerequisites

  • The openssh-server package is installed.

Procedure

  1. Start the sshd daemon in the current session and set it to start automatically at boot time:

    # systemctl start sshd
    # systemctl enable sshd
  2. To specify different addresses than the default 0.0.0.0 (IPv4) or :: (IPv6) for the ListenAddress directive in the /etc/ssh/sshd_config configuration file and to use a slower dynamic network configuration, add the dependency on the network-online.target target unit to the sshd.service unit file. To achieve this, create the /etc/systemd/system/sshd.service.d/local.conf file with the following content:

    [Unit]
    Wants=network-online.target
    After=network-online.target
  3. Review if OpenSSH server settings in the /etc/ssh/sshd_config configuration file meet the requirements of your scenario.
  4. Optionally, change the welcome message that your OpenSSH server displays before a client authenticates by editing the /etc/issue file, for example:

    Welcome to ssh-server.example.com
    Warning: By accessing this server, you agree to the referenced terms and conditions.

    Ensure that the Banner option is not commented out in /etc/ssh/sshd_config and its value contains /etc/issue:

    # less /etc/ssh/sshd_config | grep Banner
    Banner /etc/issue

    Note that to change the message displayed after a successful login you have to edit the /etc/motd file on the server. See the pam_motd man page for more information.

  5. Reload the systemd configuration and restart sshd to apply the changes:

    # systemctl daemon-reload
    # systemctl restart sshd

Verification

  1. Check that the sshd daemon is running:

    # systemctl status sshd
    ● sshd.service - OpenSSH server daemon
       Loaded: loaded (/usr/lib/systemd/system/sshd.service; enabled; vendor preset: enabled)
       Active: active (running) since Mon 2019-11-18 14:59:58 CET; 6min ago
         Docs: man:sshd(8)
               man:sshd_config(5)
     Main PID: 1149 (sshd)
        Tasks: 1 (limit: 11491)
       Memory: 1.9M
       CGroup: /system.slice/sshd.service
               └─1149 /usr/sbin/sshd -D -oCiphers=aes128-ctr,aes256-ctr,aes128-cbc,aes256-cbc -oMACs=hmac-sha2-256,>
    
    Nov 18 14:59:58 ssh-server-example.com systemd[1]: Starting OpenSSH server daemon...
    Nov 18 14:59:58 ssh-server-example.com sshd[1149]: Server listening on 0.0.0.0 port 22.
    Nov 18 14:59:58 ssh-server-example.com sshd[1149]: Server listening on :: port 22.
    Nov 18 14:59:58 ssh-server-example.com systemd[1]: Started OpenSSH server daemon.
  2. Connect to the SSH server with an SSH client.

    # ssh user@ssh-server-example.com
    ECDSA key fingerprint is SHA256:dXbaS0RG/UzlTTku8GtXSz0S1++lPegSy31v3L/FAEc.
    Are you sure you want to continue connecting (yes/no/[fingerprint])? yes
    Warning: Permanently added 'ssh-server-example.com' (ECDSA) to the list of known hosts.
    
    user@ssh-server-example.com's password:

Additional resources

  • sshd(8) and sshd_config(5) man pages.

21.1.3. Setting an OpenSSH server for key-based authentication

To improve system security, enforce key-based authentication by disabling password authentication on your OpenSSH server.

Prerequisites

  • The openssh-server package is installed.
  • The sshd daemon is running on the server.

Procedure

  1. Open the /etc/ssh/sshd_config configuration in a text editor, for example:

    # vi /etc/ssh/sshd_config
  2. Change the PasswordAuthentication option to no:

    PasswordAuthentication no

    On a system other than a new default installation, check that PubkeyAuthentication no has not been set and the ChallengeResponseAuthentication directive is set to no. If you are connected remotely, not using console or out-of-band access, test the key-based login process before disabling password authentication.

  3. To use key-based authentication with NFS-mounted home directories, enable the use_nfs_home_dirs SELinux boolean:

    # setsebool -P use_nfs_home_dirs 1
  4. Reload the sshd daemon to apply the changes:

    # systemctl reload sshd

Additional resources

  • sshd(8), sshd_config(5), and setsebool(8) man pages.

21.1.4. Generating SSH key pairs

Use this procedure to generate an SSH key pair on a local system and to copy the generated public key to an OpenSSH server. If the server is configured accordingly, you can log in to the OpenSSH server without providing any password.

Important

If you complete the following steps as root, only root is able to use the keys.

Procedure

  1. To generate an ECDSA key pair for version 2 of the SSH protocol:

    $ ssh-keygen -t ecdsa
    Generating public/private ecdsa key pair.
    Enter file in which to save the key (/home/joesec/.ssh/id_ecdsa):
    Enter passphrase (empty for no passphrase):
    Enter same passphrase again:
    Your identification has been saved in /home/joesec/.ssh/id_ecdsa.
    Your public key has been saved in /home/joesec/.ssh/id_ecdsa.pub.
    The key fingerprint is:
    SHA256:Q/x+qms4j7PCQ0qFd09iZEFHA+SqwBKRNaU72oZfaCI joesec@localhost.example.com
    The key's randomart image is:
    +---[ECDSA 256]---+
    |.oo..o=++        |
    |.. o .oo .       |
    |. .. o. o        |
    |....o.+...       |
    |o.oo.o +S .      |
    |.=.+.   .o       |
    |E.*+.  .  . .    |
    |.=..+ +..  o     |
    |  .  oo*+o.      |
    +----[SHA256]-----+

    You can also generate an RSA key pair by using the -t rsa option with the ssh-keygen command or an Ed25519 key pair by entering the ssh-keygen -t ed25519 command.

  2. To copy the public key to a remote machine:

    $ ssh-copy-id joesec@ssh-server-example.com
    /usr/bin/ssh-copy-id: INFO: attempting to log in with the new key(s), to filter out any that are already installed
    joesec@ssh-server-example.com's password:
    ...
    Number of key(s) added: 1
    
    Now try logging into the machine, with: "ssh 'joesec@ssh-server-example.com'" and check to make sure that only the key(s) you wanted were added.

    If you do not use the ssh-agent program in your session, the previous command copies the most recently modified ~/.ssh/id*.pub public key if it is not yet installed. To specify another public-key file or to prioritize keys in files over keys cached in memory by ssh-agent, use the ssh-copy-id command with the -i option.

Note

If you reinstall your system and want to keep previously generated key pairs, back up the ~/.ssh/ directory. After reinstalling, copy it back to your home directory. You can do this for all users on your system, including root.

Verification

  1. Log in to the OpenSSH server without providing any password:

    $ ssh joesec@ssh-server-example.com
    Welcome message.
    ...
    Last login: Mon Nov 18 18:28:42 2019 from ::1

Additional resources

  • ssh-keygen(1) and ssh-copy-id(1) man pages.

21.1.5. Using SSH keys stored on a smart card

Red Hat Enterprise Linux enables you to use RSA and ECDSA keys stored on a smart card on OpenSSH clients. Use this procedure to enable authentication using a smart card instead of using a password.

Prerequisites

  • On the client side, the opensc package is installed and the pcscd service is running.

Procedure

  1. List all keys provided by the OpenSC PKCS #11 module including their PKCS #11 URIs and save the output to the keys.pub file:

    $ ssh-keygen -D pkcs11: > keys.pub
    $ ssh-keygen -D pkcs11:
    ssh-rsa AAAAB3NzaC1yc2E...KKZMzcQZzx pkcs11:id=%02;object=SIGN%20pubkey;token=SSH%20key;manufacturer=piv_II?module-path=/usr/lib64/pkcs11/opensc-pkcs11.so
    ecdsa-sha2-nistp256 AAA...J0hkYnnsM= pkcs11:id=%01;object=PIV%20AUTH%20pubkey;token=SSH%20key;manufacturer=piv_II?module-path=/usr/lib64/pkcs11/opensc-pkcs11.so
  2. To enable authentication using a smart card on a remote server (example.com), transfer the public key to the remote server. Use the ssh-copy-id command with keys.pub created in the previous step:

    $ ssh-copy-id -f -i keys.pub username@example.com
  3. To connect to example.com using the ECDSA key from the output of the ssh-keygen -D command in step 1, you can use just a subset of the URI, which uniquely references your key, for example:

    $ ssh -i "pkcs11:id=%01?module-path=/usr/lib64/pkcs11/opensc-pkcs11.so" example.com
    Enter PIN for 'SSH key':
    [example.com] $
  4. You can use the same URI string in the ~/.ssh/config file to make the configuration permanent:

    $ cat ~/.ssh/config
    IdentityFile "pkcs11:id=%01?module-path=/usr/lib64/pkcs11/opensc-pkcs11.so"
    $ ssh example.com
    Enter PIN for 'SSH key':
    [example.com] $

    Because OpenSSH uses the p11-kit-proxy wrapper and the OpenSC PKCS #11 module is registered to PKCS#11 Kit, you can simplify the previous commands:

    $ ssh -i "pkcs11:id=%01" example.com
    Enter PIN for 'SSH key':
    [example.com] $

If you skip the id= part of a PKCS #11 URI, OpenSSH loads all keys that are available in the proxy module. This can reduce the amount of typing required:

$ ssh -i pkcs11: example.com
Enter PIN for 'SSH key':
[example.com] $

Additional resources

21.1.6. Making OpenSSH more secure

The following tips help you to increase security when using OpenSSH. Note that changes in the /etc/ssh/sshd_config OpenSSH configuration file require reloading the sshd daemon to take effect:

# systemctl reload sshd
Important

The majority of security hardening configuration changes reduce compatibility with clients that do not support up-to-date algorithms or cipher suites.

Disabling insecure connection protocols

  • To make SSH truly effective, prevent the use of insecure connection protocols that are replaced by the OpenSSH suite. Otherwise, a user’s password might be protected using SSH for one session only to be captured later when logging in using Telnet. For this reason, consider disabling insecure protocols, such as telnet, rsh, rlogin, and ftp.

Enabling key-based authentication and disabling password-based authentication

  • Disabling passwords for authentication and allowing only key pairs reduces the attack surface and it also might save users’ time. On clients, generate key pairs using the ssh-keygen tool and use the ssh-copy-id utility to copy public keys from clients on the OpenSSH server. To disable password-based authentication on your OpenSSH server, edit /etc/ssh/sshd_config and change the PasswordAuthentication option to no:

    PasswordAuthentication no

Key types

  • Although the ssh-keygen command generates a pair of RSA keys by default, you can instruct it to generate ECDSA or Ed25519 keys by using the -t option. The ECDSA (Elliptic Curve Digital Signature Algorithm) offers better performance than RSA at the equivalent symmetric key strength. It also generates shorter keys. The Ed25519 public-key algorithm is an implementation of twisted Edwards curves that is more secure and also faster than RSA, DSA, and ECDSA.

    OpenSSH creates RSA, ECDSA, and Ed25519 server host keys automatically if they are missing. To configure the host key creation in RHEL, use the sshd-keygen@.service instantiated service. For example, to disable the automatic creation of the RSA key type:

    # systemctl mask sshd-keygen@rsa.service
    Note

    In images with cloud-init enabled, the ssh-keygen units are automatically disabled. This is because the ssh-keygen template service can interfere with the cloud-init tool and cause problems with host key generation. To prevent these problems the etc/systemd/system/sshd-keygen@.service.d/disable-sshd-keygen-if-cloud-init-active.conf drop-in configuration file disables the ssh-keygen units if cloud-init is running.

  • To exclude particular key types for SSH connections, comment out the relevant lines in /etc/ssh/sshd_config, and reload the sshd service. For example, to allow only Ed25519 host keys:

    # HostKey /etc/ssh/ssh_host_rsa_key
    # HostKey /etc/ssh/ssh_host_ecdsa_key
    HostKey /etc/ssh/ssh_host_ed25519_key

Non-default port

  • By default, the sshd daemon listens on TCP port 22. Changing the port reduces the exposure of the system to attacks based on automated network scanning and therefore increase security through obscurity. You can specify the port using the Port directive in the /etc/ssh/sshd_config configuration file.

    You also have to update the default SELinux policy to allow the use of a non-default port. To do so, use the semanage tool from the policycoreutils-python-utils package:

    # semanage port -a -t ssh_port_t -p tcp port_number

    Furthermore, update firewalld configuration:

    # firewall-cmd --add-port port_number/tcp
    # firewall-cmd --runtime-to-permanent

    In the previous commands, replace port_number with the new port number specified using the Port directive.

No root login

  • If your particular use case does not require the possibility of logging in as the root user, you should consider setting the PermitRootLogin configuration directive to no in the /etc/ssh/sshd_config file. By disabling the possibility of logging in as the root user, the administrator can audit which users run what privileged commands after they log in as regular users and then gain root rights.

    Alternatively, set PermitRootLogin to prohibit-password:

    PermitRootLogin prohibit-password

    This enforces the use of key-based authentication instead of the use of passwords for logging in as root and reduces risks by preventing brute-force attacks.

Using the X Security extension

  • The X server in Red Hat Enterprise Linux clients does not provide the X Security extension. Therefore, clients cannot request another security layer when connecting to untrusted SSH servers with X11 forwarding. Most applications are not able to run with this extension enabled anyway.

    By default, the ForwardX11Trusted option in the /etc/ssh/ssh_config.d/05-redhat.conf file is set to yes, and there is no difference between the ssh -X remote_machine (untrusted host) and ssh -Y remote_machine (trusted host) command.

    If your scenario does not require the X11 forwarding feature at all, set the X11Forwarding directive in the /etc/ssh/sshd_config configuration file to no.

Restricting access to specific users, groups, or domains

  • The AllowUsers and AllowGroups directives in the /etc/ssh/sshd_config configuration file server enable you to permit only certain users, domains, or groups to connect to your OpenSSH server. You can combine AllowUsers and AllowGroups to restrict access more precisely, for example:

    AllowUsers *@192.168.1.*,*@10.0.0.*,!*@192.168.1.2
    AllowGroups example-group

    The previous configuration lines accept connections from all users from systems in 192.168.1.* and 10.0.0.* subnets except from the system with the 192.168.1.2 address. All users must be in the example-group group. The OpenSSH server denies all other connections.

    Note that using allowlists (directives starting with Allow) is more secure than using blocklists (options starting with Deny) because allowlists block also new unauthorized users or groups.

Changing system-wide cryptographic policies

  • OpenSSH uses RHEL system-wide cryptographic policies, and the default system-wide cryptographic policy level offers secure settings for current threat models. To make your cryptographic settings more strict, change the current policy level:

    # update-crypto-policies --set FUTURE
    Setting system policy to FUTURE
  • To opt-out of the system-wide crypto policies for your OpenSSH server, uncomment the line with the CRYPTO_POLICY= variable in the /etc/sysconfig/sshd file. After this change, values that you specify in the Ciphers, MACs, KexAlgoritms, and GSSAPIKexAlgorithms sections in the /etc/ssh/sshd_config file are not overridden. Note that this task requires deep expertise in configuring cryptographic options.
  • See Using system-wide cryptographic policies in the Security hardening title for more information.

Additional resources

  • sshd_config(5), ssh-keygen(1), crypto-policies(7), and update-crypto-policies(8) man pages.

21.1.7. Connecting to a remote server using an SSH jump host

Use this procedure for connecting your local system to a remote server through an intermediary server, also called jump host.

Prerequisites

  • A jump host accepts SSH connections from your local system.
  • A remote server accepts SSH connections only from the jump host.

Procedure

  1. Define the jump host by editing the ~/.ssh/config file on your local system, for example:

    Host jump-server1
      HostName jump1.example.com
    • The Host parameter defines a name or alias for the host you can use in ssh commands. The value can match the real host name, but can also be any string.
    • The HostName parameter sets the actual host name or IP address of the jump host.
  2. Add the remote server jump configuration with the ProxyJump directive to ~/.ssh/config file on your local system, for example:

    Host remote-server
      HostName remote1.example.com
      ProxyJump jump-server1
  3. Use your local system to connect to the remote server through the jump server:

    $ ssh remote-server

    The previous command is equivalent to the ssh -J jump-server1 remote-server command if you omit the configuration steps 1 and 2.

Note

You can specify more jump servers and you can also skip adding host definitions to the configurations file when you provide their complete host names, for example:

$ ssh -J jump1.example.com,jump2.example.com,jump3.example.com remote1.example.com

Change the host name-only notation in the previous command if the user names or SSH ports on the jump servers differ from the names and ports on the remote server, for example:

$ ssh -J johndoe@jump1.example.com:75,johndoe@jump2.example.com:75,johndoe@jump3.example.com:75 joesec@remote1.example.com:220

Additional resources

  • ssh_config(5) and ssh(1) man pages.

21.1.8. Connecting to remote machines with SSH keys using ssh-agent

To avoid entering a passphrase each time you initiate an SSH connection, you can use the ssh-agent utility to cache the private SSH key. The private key and the passphrase remain secure.

Prerequisites

  • You have a remote host with SSH daemon running and reachable through the network.
  • You know the IP address or hostname and credentials to log in to the remote host.
  • You have generated an SSH key pair with a passphrase and transferred the public key to the remote machine.

Procedure

  1. Optional: Verify you can use the key to authenticate to the remote host:

    1. Connect to the remote host using SSH:

      $ ssh example.user1@198.51.100.1 hostname
    2. Enter the passphrase you set while creating the key to grant access to the private key.

      $ ssh example.user1@198.51.100.1 hostname
       host.example.com
  2. Start the ssh-agent.

    $ eval $(ssh-agent)
    Agent pid 20062
  3. Add the key to ssh-agent.

    $ ssh-add ~/.ssh/id_rsa
    Enter passphrase for ~/.ssh/id_rsa:
    Identity added: ~/.ssh/id_rsa (example.user0@198.51.100.12)

Verification

  • Optional: Log in to the host machine using SSH.

    $ ssh example.user1@198.51.100.1
    
    Last login: Mon Sep 14 12:56:37 2020

    Note that you did not have to enter the passphrase.

21.1.9. Additional resources

21.2. Planning and implementing TLS

TLS (Transport Layer Security) is a cryptographic protocol used to secure network communications. When hardening system security settings by configuring preferred key-exchange protocols, authentication methods, and encryption algorithms, it is necessary to bear in mind that the broader the range of supported clients, the lower the resulting security. Conversely, strict security settings lead to limited compatibility with clients, which can result in some users being locked out of the system. Be sure to target the strictest available configuration and only relax it when it is required for compatibility reasons.

21.2.1. SSL and TLS protocols

The Secure Sockets Layer (SSL) protocol was originally developed by Netscape Corporation to provide a mechanism for secure communication over the Internet. Subsequently, the protocol was adopted by the Internet Engineering Task Force (IETF) and renamed to Transport Layer Security (TLS).

The TLS protocol sits between an application protocol layer and a reliable transport layer, such as TCP/IP. It is independent of the application protocol and can thus be layered underneath many different protocols, for example: HTTP, FTP, SMTP, and so on.

Protocol versionUsage recommendation

SSL v2

Do not use. Has serious security vulnerabilities. Removed from the core crypto libraries since RHEL 7.

SSL v3

Do not use. Has serious security vulnerabilities. Removed from the core crypto libraries since RHEL 8.

TLS 1.0

Not recommended to use. Has known issues that cannot be mitigated in a way that guarantees interoperability, and does not support modern cipher suites. In RHEL 8, enabled only in the LEGACY system-wide cryptographic policy profile.

TLS 1.1

Use for interoperability purposes where needed. Does not support modern cipher suites. In RHEL 8, enabled only in the LEGACY policy.

TLS 1.2

Supports the modern AEAD cipher suites. This version is enabled in all system-wide crypto policies, but optional parts of this protocol contain vulnerabilities and TLS 1.2 also allows outdated algorithms.

TLS 1.3

Recommended version. TLS 1.3 removes known problematic options, provides additional privacy by encrypting more of the negotiation handshake and can be faster thanks usage of more efficient modern cryptographic algorithms. TLS 1.3 is also enabled in all system-wide crypto policies.

21.2.2. Security considerations for TLS in RHEL 8

In RHEL 8, cryptography-related considerations are significantly simplified thanks to the system-wide crypto policies. The DEFAULT crypto policy allows only TLS 1.2 and 1.3. To allow your system to negotiate connections using the earlier versions of TLS, you need to either opt out from following crypto policies in an application or switch to the LEGACY policy with the update-crypto-policies command. See Using system-wide cryptographic policies for more information.

The default settings provided by libraries included in RHEL 8 are secure enough for most deployments. The TLS implementations use secure algorithms where possible while not preventing connections from or to legacy clients or servers. Apply hardened settings in environments with strict security requirements where legacy clients or servers that do not support secure algorithms or protocols are not expected or allowed to connect.

The most straightforward way to harden your TLS configuration is switching the system-wide cryptographic policy level to FUTURE using the update-crypto-policies --set FUTURE command.

Warning

Algorithms disabled for the LEGACY cryptographic policy do not conform to Red Hat’s vision of RHEL 8 security, and their security properties are not reliable. Consider moving away from using these algorithms instead of re-enabling them. If you do decide to re-enable them, for example for interoperability with old hardware, treat them as insecure and apply extra protection measures, such as isolating their network interactions to separate network segments. Do not use them across public networks.

If you decide to not follow RHEL system-wide crypto policies or create custom cryptographic policies tailored to your setup, use the following recommendations for preferred protocols, cipher suites, and key lengths on your custom configuration:

21.2.2.1. Protocols

The latest version of TLS provides the best security mechanism. Unless you have a compelling reason to include support for older versions of TLS, allow your systems to negotiate connections using at least TLS version 1.2.

Note that even though RHEL 8 supports TLS version 1.3, not all features of this protocol are fully supported by RHEL 8 components. For example, the 0-RTT (Zero Round Trip Time) feature, which reduces connection latency, is not yet fully supported by the Apache web server.

21.2.2.2. Cipher suites

Modern, more secure cipher suites should be preferred to old, insecure ones. Always disable the use of eNULL and aNULL cipher suites, which do not offer any encryption or authentication at all. If at all possible, ciphers suites based on RC4 or HMAC-MD5, which have serious shortcomings, should also be disabled. The same applies to the so-called export cipher suites, which have been intentionally made weaker, and thus are easy to break.

While not immediately insecure, cipher suites that offer less than 128 bits of security should not be considered for their short useful life. Algorithms that use 128 bits of security or more can be expected to be unbreakable for at least several years, and are thus strongly recommended. Note that while 3DES ciphers advertise the use of 168 bits, they actually offer 112 bits of security.

Always prefer cipher suites that support (perfect) forward secrecy (PFS), which ensures the confidentiality of encrypted data even in case the server key is compromised. This rules out the fast RSA key exchange, but allows for the use of ECDHE and DHE. Of the two, ECDHE is the faster and therefore the preferred choice.

You should also prefer AEAD ciphers, such as AES-GCM, over CBC-mode ciphers as they are not vulnerable to padding oracle attacks. Additionally, in many cases, AES-GCM is faster than AES in CBC mode, especially when the hardware has cryptographic accelerators for AES.

Note also that when using the ECDHE key exchange with ECDSA certificates, the transaction is even faster than a pure RSA key exchange. To provide support for legacy clients, you can install two pairs of certificates and keys on a server: one with ECDSA keys (for new clients) and one with RSA keys (for legacy ones).

21.2.2.3. Public key length

When using RSA keys, always prefer key lengths of at least 3072 bits signed by at least SHA-256, which is sufficiently large for true 128 bits of security.

Warning

The security of your system is only as strong as the weakest link in the chain. For example, a strong cipher alone does not guarantee good security. The keys and the certificates are just as important, as well as the hash functions and keys used by the Certification Authority (CA) to sign your keys.

Additional resources

21.2.3. Hardening TLS configuration in applications

In RHEL, system-wide crypto policies provide a convenient way to ensure that your applications using cryptographic libraries do not allow known insecure protocols, ciphers, or algorithms.

If you want to harden your TLS-related configuration with your customized cryptographic settings, you can use the cryptographic configuration options described in this section, and override the system-wide crypto policies just in the minimum required amount.

Regardless of the configuration you choose to use, always ensure that your server application enforces server-side cipher order, so that the cipher suite to be used is determined by the order you configure.

21.2.3.1. Configuring the Apache HTTP server to use TLS

The Apache HTTP Server can use both OpenSSL and NSS libraries for its TLS needs. RHEL 8 provides the mod_ssl functionality through eponymous packages:

# yum install mod_ssl

The mod_ssl package installs the /etc/httpd/conf.d/ssl.conf configuration file, which can be used to modify the TLS-related settings of the Apache HTTP Server.

Install the httpd-manual package to obtain complete documentation for the Apache HTTP Server, including TLS configuration. The directives available in the /etc/httpd/conf.d/ssl.conf configuration file are described in detail in the /usr/share/httpd/manual/mod/mod_ssl.html file. Examples of various settings are described in the /usr/share/httpd/manual/ssl/ssl_howto.html file.

When modifying the settings in the /etc/httpd/conf.d/ssl.conf configuration file, be sure to consider the following three directives at the minimum:

SSLProtocol
Use this directive to specify the version of TLS or SSL you want to allow.
SSLCipherSuite
Use this directive to specify your preferred cipher suite or disable the ones you want to disallow.
SSLHonorCipherOrder
Uncomment and set this directive to on to ensure that the connecting clients adhere to the order of ciphers you specified.

For example, to use only the TLS 1.2 and 1.3 protocol:

SSLProtocol             all -SSLv3 -TLSv1 -TLSv1.1

See the Configuring TLS encryption on an Apache HTTP Server chapter in the Deploying different types of servers document for more information.

21.2.3.2. Configuring the Nginx HTTP and proxy server to use TLS

To enable TLS 1.3 support in Nginx, add the TLSv1.3 value to the ssl_protocols option in the server section of the /etc/nginx/nginx.conf configuration file:

server {
    listen 443 ssl http2;
    listen [::]:443 ssl http2;
    ....
    ssl_protocols TLSv1.2 TLSv1.3;
    ssl_ciphers
    ....
}

See the Adding TLS encryption to an Nginx web server chapter in the Deploying different types of servers document for more information.

21.2.3.3. Configuring the Dovecot mail server to use TLS

To configure your installation of the Dovecot mail server to use TLS, modify the /etc/dovecot/conf.d/10-ssl.conf configuration file. You can find an explanation of some of the basic configuration directives available in that file in the /usr/share/doc/dovecot/wiki/SSL.DovecotConfiguration.txt file, which is installed along with the standard installation of Dovecot.

When modifying the settings in the /etc/dovecot/conf.d/10-ssl.conf configuration file, be sure to consider the following three directives at the minimum:

ssl_protocols
Use this directive to specify the version of TLS or SSL you want to allow or disable.
ssl_cipher_list
Use this directive to specify your preferred cipher suites or disable the ones you want to disallow.
ssl_prefer_server_ciphers
Uncomment and set this directive to yes to ensure that the connecting clients adhere to the order of ciphers you specified.

For example, the following line in /etc/dovecot/conf.d/10-ssl.conf allows only TLS 1.1 and later:

ssl_protocols = !SSLv2 !SSLv3 !TLSv1

21.3. Configuring a VPN with IPsec

In RHEL 8, a virtual private network (VPN) can be configured using the IPsec protocol, which is supported by the Libreswan application.

21.3.1. Libreswan as an IPsec VPN implementation

In RHEL, a Virtual Private Network (VPN) can be configured using the IPsec protocol, which is supported by the Libreswan application. Libreswan is a continuation of the Openswan application, and many examples from the Openswan documentation are interchangeable with Libreswan.

The IPsec protocol for a VPN is configured using the Internet Key Exchange (IKE) protocol. The terms IPsec and IKE are used interchangeably. An IPsec VPN is also called an IKE VPN, IKEv2 VPN, XAUTH VPN, Cisco VPN or IKE/IPsec VPN. A variant of an IPsec VPN that also uses the Layer 2 Tunneling Protocol (L2TP) is usually called an L2TP/IPsec VPN, which requires the xl2tpd package provided by the optional repository.

Libreswan is an open-source, user-space IKE implementation. IKE v1 and v2 are implemented as a user-level daemon. The IKE protocol is also encrypted. The IPsec protocol is implemented by the Linux kernel, and Libreswan configures the kernel to add and remove VPN tunnel configurations.

The IKE protocol uses UDP port 500 and 4500. The IPsec protocol consists of two protocols:

  • Encapsulated Security Payload (ESP), which has protocol number 50.
  • Authenticated Header (AH), which has protocol number 51.

The AH protocol is not recommended for use. Users of AH are recommended to migrate to ESP with null encryption.

The IPsec protocol provides two modes of operation:

  • Tunnel Mode (the default)
  • Transport Mode.

You can configure the kernel with IPsec without IKE. This is called Manual Keying. You can also configure manual keying using the ip xfrm commands, however, this is strongly discouraged for security reasons. Libreswan interfaces with the Linux kernel using netlink. Packet encryption and decryption happen in the Linux kernel.

Libreswan uses the Network Security Services (NSS) cryptographic library. Both Libreswan and NSS are certified for use with the Federal Information Processing Standard (FIPS) Publication 140-2.

Important

IKE/IPsec VPNs, implemented by Libreswan and the Linux kernel, is the only VPN technology recommended for use in RHEL. Do not use any other VPN technology without understanding the risks of doing so.

In RHEL, Libreswan follows system-wide cryptographic policies by default. This ensures that Libreswan uses secure settings for current threat models including IKEv2 as a default protocol. See Using system-wide crypto policies for more information.

Libreswan does not use the terms "source" and "destination" or "server" and "client" because IKE/IPsec are peer to peer protocols. Instead, it uses the terms "left" and "right" to refer to end points (the hosts). This also allows you to use the same configuration on both end points in most cases. However, administrators usually choose to always use "left" for the local host and "right" for the remote host.

The leftid and rightid options serve as identification of the respective hosts in the authentication process. See the ipsec.conf(5) man page for more information.

21.3.2. Authentication methods in Libreswan

Libreswan supports several authentication methods, each of which fits a different scenario.

Pre-Shared key (PSK)

Pre-Shared Key (PSK) is the simplest authentication method. For security reasons, do not use PSKs shorter than 64 random characters. In FIPS mode, PSKs must comply with a minimum-strength requirement depending on the integrity algorithm used. You can set PSK by using the authby=secret connection.

Raw RSA keys

Raw RSA keys are commonly used for static host-to-host or subnet-to-subnet IPsec configurations. Each host is manually configured with the public RSA keys of all other hosts, and Libreswan sets up an IPsec tunnel between each pair of hosts. This method does not scale well for large numbers of hosts.

You can generate a raw RSA key on a host using the ipsec newhostkey command. You can list generated keys by using the ipsec showhostkey command. The leftrsasigkey= line is required for connection configurations that use CKA ID keys. Use the authby=rsasig connection option for raw RSA keys.

X.509 certificates

X.509 certificates are commonly used for large-scale deployments with hosts that connect to a common IPsec gateway. A central certificate authority (CA) signs RSA certificates for hosts or users. This central CA is responsible for relaying trust, including the revocations of individual hosts or users.

For example, you can generate X.509 certificates using the openssl command and the NSS certutil command. Because Libreswan reads user certificates from the NSS database using the certificates' nickname in the leftcert= configuration option, provide a nickname when you create a certificate.

If you use a custom CA certificate, you must import it to the Network Security Services (NSS) database. You can import any certificate in the PKCS #12 format to the Libreswan NSS database by using the ipsec import command.

Warning

Libreswan requires an Internet Key Exchange (IKE) peer ID as a subject alternative name (SAN) for every peer certificate as described in section 3.1 of RFC 4945. Disabling this check by changing the require-id-on-certificated= option can make the system vulnerable to man-in-the-middle attacks.

Use the authby=rsasig connection option for authentication based on X.509 certificates using RSA with SHA-1 and SHA-2. You can further limit it for ECDSA digital signatures using SHA-2 by setting authby= to ecdsa and RSA Probabilistic Signature Scheme (RSASSA-PSS) digital signatures based authentication with SHA-2 through authby=rsa-sha2. The default value is authby=rsasig,ecdsa.

The certificates and the authby= signature methods should match. This increases interoperability and preserves authentication in one digital-signature system.

NULL authentication

NULL authentication is used to gain mesh encryption without authentication. It protects against passive attacks but not against active attacks. However, because IKEv2 allows asymmetric authentication methods, NULL authentication can also be used for internet-scale opportunistic IPsec. In this model, clients authenticate the server, but servers do not authenticate the client. This model is similar to secure websites using TLS. Use authby=null for NULL authentication.

Protection against quantum computers

In addition to the previously mentioned authentication methods, you can use the Post-quantum Pre-shared Key (PPK) method to protect against possible attacks by quantum computers. Individual clients or groups of clients can use their own PPK by specifying a PPK ID that corresponds to an out-of-band configured pre-shared key.

Using IKEv1 with pre-shared keys provides protection against quantum attackers. The redesign of IKEv2 does not offer this protection natively. Libreswan offers the use of Post-quantum Pre-shared Key (PPK) to protect IKEv2 connections against quantum attacks.

To enable optional PPK support, add ppk=yes to the connection definition. To require PPK, add ppk=insist. Then, each client can be given a PPK ID with a secret value that is communicated out-of-band (and preferably quantum safe). The PPK’s should be very strong in randomness and not based on dictionary words. The PPK ID and PPK data are stored in ipsec.secrets, for example:

@west @east : PPKS "user1" "thestringismeanttobearandomstr"

The PPKS option refers to static PPKs. This experimental function uses one-time-pad-based Dynamic PPKs. Upon each connection, a new part of the one-time pad is used as the PPK. When used, that part of the dynamic PPK inside the file is overwritten with zeros to prevent re-use. If there is no more one-time-pad material left, the connection fails. See the ipsec.secrets(5) man page for more information.

Warning

The implementation of dynamic PPKs is provided as an unsupported Technology Preview. Use with caution.

21.3.3. Installing Libreswan

This procedure describes the steps for installing and starting the Libreswan IPsec/IKE VPN implementation.

Prerequisites

  • The AppStream repository is enabled.

Procedure

  1. Install the libreswan packages:

    # yum install libreswan
  2. If you are re-installing Libreswan, remove its old database files and create a new database:

    # systemctl stop ipsec
    # rm /etc/ipsec.d/*db
    # ipsec initnss
  3. Start the ipsec service, and enable the service to be started automatically on boot:

    # systemctl enable ipsec --now
  4. Configure the firewall to allow 500 and 4500/UDP ports for the IKE, ESP, and AH protocols by adding the ipsec service:

    # firewall-cmd --add-service="ipsec"
    # firewall-cmd --runtime-to-permanent

21.3.4. Creating a host-to-host VPN

To configure [application]Libreswan to create a host-to-host IPsec VPN between two hosts referred to as left and right using authentication by raw RSA keys, enter the following commands on both of the hosts:

Prerequisites

  • Libreswan is installed and the ipsec service is started on each node.

Procedure

  1. Generate a raw RSA key pair on each host:

    # ipsec newhostkey
  2. The previous step returned the generated key’s ckaid. Use that ckaid with the following command on left, for example:

    # ipsec showhostkey --left --ckaid 2d3ea57b61c9419dfd6cf43a1eb6cb306c0e857d

    The output of the previous command generated the leftrsasigkey= line required for the configuration. Do the same on the second host (right):

    # ipsec showhostkey --right --ckaid a9e1f6ce9ecd3608c24e8f701318383f41798f03
  3. In the /etc/ipsec.d/ directory, create a new my_host-to-host.conf file. Write the RSA host keys from the output of the ipsec showhostkey commands in the previous step to the new file. For example:

    conn mytunnel
        leftid=@west
        left=192.1.2.23
        leftrsasigkey=0sAQOrlo+hOafUZDlCQmXFrje/oZm [...] W2n417C/4urYHQkCvuIQ==
        rightid=@east
        right=192.1.2.45
        rightrsasigkey=0sAQO3fwC6nSSGgt64DWiYZzuHbc4 [...] D/v8t5YTQ==
        authby=rsasig
  4. After importing keys, restart the ipsec service:

    # systemctl restart ipsec
  5. Load the connection:

    # ipsec auto --add mytunnel
  6. Establish the tunnel:

    # ipsec auto --up mytunnel
  7. To automatically start the tunnel when the ipsec service is started, add the following line to the connection definition:

    auto=start

21.3.5. Configuring a site-to-site VPN

To create a site-to-site IPsec VPN, by joining two networks, an IPsec tunnel between the two hosts, is created. The hosts thus act as the end points, which are configured to permit traffic from one or more subnets to pass through. Therefore you can think of the host as gateways to the remote portion of the network.

The configuration of the site-to-site VPN only differs from the host-to-host VPN in that one or more networks or subnets must be specified in the configuration file.

Prerequisites

Procedure

  1. Copy the file with the configuration of your host-to-host VPN to a new file, for example:

    # cp /etc/ipsec.d/my_host-to-host.conf /etc/ipsec.d/my_site-to-site.conf
  2. Add the subnet configuration to the file created in the previous step, for example:

    conn mysubnet
         also=mytunnel
         leftsubnet=192.0.1.0/24
         rightsubnet=192.0.2.0/24
         auto=start
    
    conn mysubnet6
         also=mytunnel
         leftsubnet=2001:db8:0:1::/64
         rightsubnet=2001:db8:0:2::/64
         auto=start
    
    # the following part of the configuration file is the same for both host-to-host and site-to-site connections:
    
    conn mytunnel
        leftid=@west
        left=192.1.2.23
        leftrsasigkey=0sAQOrlo+hOafUZDlCQmXFrje/oZm [...] W2n417C/4urYHQkCvuIQ==
        rightid=@east
        right=192.1.2.45
        rightrsasigkey=0sAQO3fwC6nSSGgt64DWiYZzuHbc4 [...] D/v8t5YTQ==
        authby=rsasig

21.3.6. Configuring a remote access VPN

Road warriors are traveling users with mobile clients and a dynamically assigned IP address. The mobile clients authenticate using X.509 certificates.

The following example shows configuration for IKEv2, and it avoids using the IKEv1 XAUTH protocol.

On the server:

conn roadwarriors
    ikev2=insist
    # support (roaming) MOBIKE clients (RFC 4555)
    mobike=yes
    fragmentation=yes
    left=1.2.3.4
    # if access to the LAN is given, enable this, otherwise use 0.0.0.0/0
    # leftsubnet=10.10.0.0/16
    leftsubnet=0.0.0.0/0
    leftcert=gw.example.com
    leftid=%fromcert
    leftxauthserver=yes
    leftmodecfgserver=yes
    right=%any
    # trust our own Certificate Agency
    rightca=%same
    # pick an IP address pool to assign to remote users
    # 100.64.0.0/16 prevents RFC1918 clashes when remote users are behind NAT
    rightaddresspool=100.64.13.100-100.64.13.254
    # if you want remote clients to use some local DNS zones and servers
    modecfgdns="1.2.3.4, 5.6.7.8"
    modecfgdomains="internal.company.com, corp"
    rightxauthclient=yes
    rightmodecfgclient=yes
    authby=rsasig
    # optionally, run the client X.509 ID through pam to allow or deny client
    # pam-authorize=yes
    # load connection, do not initiate
    auto=add
    # kill vanished roadwarriors
    dpddelay=1m
    dpdtimeout=5m
    dpdaction=clear

On the mobile client, the road warrior’s device, use a slight variation of the previous configuration:

conn to-vpn-server
    ikev2=insist
    # pick up our dynamic IP
    left=%defaultroute
    leftsubnet=0.0.0.0/0
    leftcert=myname.example.com
    leftid=%fromcert
    leftmodecfgclient=yes
    # right can also be a DNS hostname
    right=1.2.3.4
    # if access to the remote LAN is required, enable this, otherwise use 0.0.0.0/0
    # rightsubnet=10.10.0.0/16
    rightsubnet=0.0.0.0/0
    fragmentation=yes
    # trust our own Certificate Agency
    rightca=%same
    authby=rsasig
    # allow narrowing to the server’s suggested assigned IP and remote subnet
    narrowing=yes
    # support (roaming) MOBIKE clients (RFC 4555)
    mobike=yes
    # initiate connection
    auto=start

21.3.7. Configuring a mesh VPN

A mesh VPN network, which is also known as an any-to-any VPN, is a network where all nodes communicate using IPsec. The configuration allows for exceptions for nodes that cannot use IPsec. The mesh VPN network can be configured in two ways:

  • To require IPsec.
  • To prefer IPsec but allow a fallback to clear-text communication.

Authentication between the nodes can be based on X.509 certificates or on DNS Security Extensions (DNSSEC).

The following procedure uses X.509 certificates. These certificates can be generated using any kind of Certificate Authority (CA) management system, such as the Dogtag Certificate System. Dogtag assumes that the certificates for each node are available in the PKCS #12 format (.p12 files), which contain the private key, the node certificate, and the Root CA certificate used to validate other nodes' X.509 certificates.

Each node has an identical configuration with the exception of its X.509 certificate. This allows for adding new nodes without reconfiguring any of the existing nodes in the network. The PKCS #12 files require a "friendly name", for which we use the name "node" so that the configuration files referencing the friendly name can be identical for all nodes.

Prerequisites

  • Libreswan is installed, and the ipsec service is started on each node.

Procedure

  1. On each node, import PKCS #12 files. This step requires the password used to generate the PKCS #12 files:

    # ipsec import nodeXXX.p12
  2. Create the following three connection definitions for the IPsec required (private), IPsec optional (private-or-clear), and No IPsec (clear) profiles:

    # cat /etc/ipsec.d/mesh.conf
    conn clear
    	auto=ondemand
    	type=passthrough
    	authby=never
    	left=%defaultroute
    	right=%group
    
    conn private
    	auto=ondemand
    	type=transport
    	authby=rsasig
    	failureshunt=drop
    	negotiationshunt=drop
    	# left
    	left=%defaultroute
    	leftcert=nodeXXXX
    	leftid=%fromcert
            leftrsasigkey=%cert
    	# right
    	rightrsasigkey=%cert
    	rightid=%fromcert
    	right=%opportunisticgroup
    
    conn private-or-clear
    	auto=ondemand
    	type=transport
    	authby=rsasig
    	failureshunt=passthrough
    	negotiationshunt=passthrough
    	# left
    	left=%defaultroute
    	leftcert=nodeXXXX
    	leftid=%fromcert
            leftrsasigkey=%cert
    	# right
    	rightrsasigkey=%cert
    	rightid=%fromcert
    	right=%opportunisticgroup
  3. Add the IP address of the network in the proper category. For example, if all nodes reside in the 10.15.0.0/16 network, and all nodes should mandate IPsec encryption:

    # echo "10.15.0.0/16" >> /etc/ipsec.d/policies/private
  4. To allow certain nodes, for example, 10.15.34.0/24, to work with and without IPsec, add those nodes to the private-or-clear group using:

    # echo "10.15.34.0/24" >> /etc/ipsec.d/policies/private-or-clear
  5. To define a host, for example, 10.15.1.2, that is not capable of IPsec into the clear group, use:

    # echo "10.15.1.2/32" >> /etc/ipsec.d/policies/clear

    The files in the /etc/ipsec.d/policies directory can be created from a template for each new node, or can be provisioned using Puppet or Ansible.

    Note that every node has the same list of exceptions or different traffic flow expectations. Two nodes, therefore, might not be able to communicate because one requires IPsec and the other cannot use IPsec.

  6. Restart the node to add it to the configured mesh:

    # systemctl restart ipsec
  7. Once you finish with the addition of nodes, a ping command is sufficient to open an IPsec tunnel. To see which tunnels a node has opened:

    # ipsec trafficstatus

21.3.8. Deploying a FIPS-compliant IPsec VPN

Use this procedure to deploy a FIPS-compliant IPsec VPN solution based on Libreswan. The following steps also enable you to identify which cryptographic algorithms are available and which are disabled for Libreswan in FIPS mode.

Prerequisites

  • The AppStream repository is enabled.

Procedure

  1. Install the libreswan packages:

    # yum install libreswan
  2. If you are re-installing Libreswan, remove its old NSS database:

    # systemctl stop ipsec
    # rm /etc/ipsec.d/*db
  3. Start the ipsec service, and enable the service to be started automatically on boot:

    # systemctl enable ipsec --now
  4. Configure the firewall to allow 500 and 4500/UDP ports for the IKE, ESP, and AH protocols by adding the ipsec service:

    # firewall-cmd --add-service="ipsec"
    # firewall-cmd --runtime-to-permanent
  5. Switch the system to FIPS mode:

    # fips-mode-setup --enable
  6. Restart your system to allow the kernel to switch to FIPS mode:

    # reboot

Verification

  1. To confirm Libreswan is running in FIPS mode:

    # ipsec whack --fipsstatus
    000 FIPS mode enabled
  2. Alternatively, check entries for the ipsec unit in the systemd journal:

    $ journalctl -u ipsec
    ...
    Jan 22 11:26:50 localhost.localdomain pluto[3076]: FIPS Product: YES
    Jan 22 11:26:50 localhost.localdomain pluto[3076]: FIPS Kernel: YES
    Jan 22 11:26:50 localhost.localdomain pluto[3076]: FIPS Mode: YES
  3. To see the available algorithms in FIPS mode:

    # ipsec pluto --selftest 2>&1 | head -11
    FIPS Product: YES
    FIPS Kernel: YES
    FIPS Mode: YES
    NSS DB directory: sql:/etc/ipsec.d
    Initializing NSS
    Opening NSS database "sql:/etc/ipsec.d" read-only
    NSS initialized
    NSS crypto library initialized
    FIPS HMAC integrity support [enabled]
    FIPS mode enabled for pluto daemon
    NSS library is running in FIPS mode
    FIPS HMAC integrity verification self-test passed
  4. To query disabled algorithms in FIPS mode:

    # ipsec pluto --selftest 2>&1 | grep disabled
    Encryption algorithm CAMELLIA_CTR disabled; not FIPS compliant
    Encryption algorithm CAMELLIA_CBC disabled; not FIPS compliant
    Encryption algorithm SERPENT_CBC disabled; not FIPS compliant
    Encryption algorithm TWOFISH_CBC disabled; not FIPS compliant
    Encryption algorithm TWOFISH_SSH disabled; not FIPS compliant
    Encryption algorithm NULL disabled; not FIPS compliant
    Encryption algorithm CHACHA20_POLY1305 disabled; not FIPS compliant
    Hash algorithm MD5 disabled; not FIPS compliant
    PRF algorithm HMAC_MD5 disabled; not FIPS compliant
    PRF algorithm AES_XCBC disabled; not FIPS compliant
    Integrity algorithm HMAC_MD5_96 disabled; not FIPS compliant
    Integrity algorithm HMAC_SHA2_256_TRUNCBUG disabled; not FIPS compliant
    Integrity algorithm AES_XCBC_96 disabled; not FIPS compliant
    DH algorithm MODP1024 disabled; not FIPS compliant
    DH algorithm MODP1536 disabled; not FIPS compliant
    DH algorithm DH31 disabled; not FIPS compliant
  5. To list all allowed algorithms and ciphers in FIPS mode:

    # ipsec pluto --selftest 2>&1 | grep ESP | grep FIPS | sed "s/^.*FIPS//"
    {256,192,*128}  aes_ccm, aes_ccm_c
    {256,192,*128}  aes_ccm_b
    {256,192,*128}  aes_ccm_a
    [*192]  3des
    {256,192,*128}  aes_gcm, aes_gcm_c
    {256,192,*128}  aes_gcm_b
    {256,192,*128}  aes_gcm_a
    {256,192,*128}  aesctr
    {256,192,*128}  aes
    {256,192,*128}  aes_gmac
    sha, sha1, sha1_96, hmac_sha1
    sha512, sha2_512, sha2_512_256, hmac_sha2_512
    sha384, sha2_384, sha2_384_192, hmac_sha2_384
    sha2, sha256, sha2_256, sha2_256_128, hmac_sha2_256
    aes_cmac
    null
    null, dh0
    dh14
    dh15
    dh16
    dh17
    dh18
    ecp_256, ecp256
    ecp_384, ecp384
    ecp_521, ecp521

21.3.9. Protecting the IPsec NSS database by a password

By default, the IPsec service creates its Network Security Services (NSS) database with an empty password during the first start. Add password protection by using the following steps.

Note

In the previous releases of RHEL up to version 6.6, you had to protect the IPsec NSS database with a password to meet the FIPS 140-2 requirements because the NSS cryptographic libraries were certified for the FIPS 140-2 Level 2 standard. In RHEL 8, NIST certified NSS to Level 1 of this standard, and this status does not require password protection for the database.

Prerequisites

  • The /etc/ipsec.d/ directory contains NSS database files.

Procedure

  1. Enable password protection for the NSS database for Libreswan:

    # certutil -N -d sql:/etc/ipsec.d
    Enter Password or Pin for "NSS Certificate DB":
    Enter a password which will be used to encrypt your keys.
    The password should be at least 8 characters long,
    and should contain at least one non-alphabetic character.
    
    Enter new password:
  2. Create the /etc/ipsec.d/nsspassword file containing the password you have set in the previous step, for example:

    # cat /etc/ipsec.d/nsspassword
    NSS Certificate DB:MyStrongPasswordHere

    Note that the nsspassword file use the following syntax:

    token_1_name:the_password
    token_2_name:the_password

    The default NSS software token is NSS Certificate DB. If your system is running in FIPS mode, the name of the token is NSS FIPS 140-2 Certificate DB.

  3. Depending on your scenario, either start or restart the ipsec service after you finish the nsspassword file:

    # systemctl restart ipsec

Verification

  1. Check that the ipsec service is running after you have added a non-empty password to its NSS database:

    # systemctl status ipsec
    ● ipsec.service - Internet Key Exchange (IKE) Protocol Daemon for IPsec
       Loaded: loaded (/usr/lib/systemd/system/ipsec.service; enabled; vendor preset: disable>
       Active: active (running)...
  2. Optionally, check that the Journal log contains entries confirming a successful initialization:

    # journalctl -u ipsec
    ...
    pluto[6214]: Initializing NSS using read-write database "sql:/etc/ipsec.d"
    pluto[6214]: NSS Password from file "/etc/ipsec.d/nsspassword" for token "NSS Certificate DB" with length 20 passed to NSS
    pluto[6214]: NSS crypto library initialized
    ...

Additional resources

21.3.10. Configuring an IPsec VPN to use TCP

Libreswan supports TCP encapsulation of IKE and IPsec packets as described in RFC 8229. With this feature, you can establish IPsec VPNs on networks that prevent traffic transmitted via UDP and Encapsulating Security Payload (ESP). You can configure VPN servers and clients to use TCP either as a fallback or as the main VPN transport protocol. Because TCP encapsulation has bigger performance costs, use TCP as the main VPN protocol only if UDP is permanently blocked in your scenario.

Prerequisites

Procedure

  1. Add the following option to the /etc/ipsec.conf file in the config setup section:

    listen-tcp=yes
  2. To use TCP encapsulation as a fallback option when the first attempt over UDP fails, add the following two options to the client’s connection definition:

    enable-tcp=fallback
    tcp-remoteport=4500

    Alternatively, if you know that UDP is permanently blocked, use the following options in the client’s connection configuration:

    enable-tcp=yes
    tcp-remoteport=4500

21.3.11. Configuring automatic detection and usage of ESP hardware offload to accelerate an IPsec connection

Offloading Encapsulating Security Payload (ESP) to the hardware accelerates IPsec connections over Ethernet. By default, Libreswan detects if hardware supports this feature and, as a result, enables ESP hardware offload. In case that the feature was disabled or explicitly enabled, you can switch back to automatic detection.

Prerequisites

  • The network card supports ESP hardware offload.
  • The network driver supports ESP hardware offload.
  • The IPsec connection is configured and works.

Procedure

  1. Edit the Libreswan configuration file in the /etc/ipsec.d/ directory of the connection that should use automatic detection of ESP hardware offload support.
  2. Ensure the nic-offload parameter is not set in the connection’s settings.
  3. If you removed nic-offload, restart the ipsec service:

    # systemctl restart ipsec

Verification

If the network card supports ESP hardware offload support, following these steps to verify the result:

  1. Display the tx_ipsec and rx_ipsec counters of the Ethernet device the IPsec connection uses:

    # ethtool -S enp1s0 | egrep "_ipsec"
         tx_ipsec: 10
         rx_ipsec: 10
  2. Send traffic through the IPsec tunnel. For example, ping a remote IP address:

    # ping -c 5 remote_ip_address
  3. Display the tx_ipsec and rx_ipsec counters of the Ethernet device again:

    # ethtool -S enp1s0 | egrep "_ipsec"
         tx_ipsec: 15
         rx_ipsec: 15

    If the counter values have increased, ESP hardware offload works.

Additional resources

21.3.12. Configuring ESP hardware offload on a bond to accelerate an IPsec connection

Offloading Encapsulating Security Payload (ESP) to the hardware accelerates IPsec connections. If you use a network bond for fail-over reasons, the requirements and the procedure to configure ESP hardware offload are different from those using a regular Ethernet device. For example, in this scenario, you enable the offload support on the bond, and the kernel applies the settings to the ports of the bond.

Prerequisites

  • All network cards in the bond support ESP hardware offload.
  • The network driver supports ESP hardware offload on a bond device. In RHEL, only the ixgbe driver supports this feature.
  • The bond is configured and works.
  • The bond uses the active-backup mode. The bonding driver does not support any other modes for this feature.
  • The IPsec connection is configured and works.

Procedure

  1. Enable ESP hardware offload support on the network bond:

    # nmcli connection modify bond0 ethtool.feature-esp-hw-offload on

    This command enables ESP hardware offload support on the bond0 connection.

  2. Reactivate the bond0 connection:

    # nmcli connection up bond0
  3. Edit the Libreswan configuration file in the /etc/ipsec.d/ directory of the connection that should use ESP hardware offload, and append the nic-offload=yes statement to the connection entry:

    conn example
        ...
        nic-offload=yes
  4. Restart the ipsec service:

    # systemctl restart ipsec

Verification

  1. Display the active port of the bond:

    # grep "Currently Active Slave" /proc/net/bonding/bond0
    Currently Active Slave: enp1s0
  2. Display the tx_ipsec and rx_ipsec counters of the active port:

    # ethtool -S enp1s0 | egrep "_ipsec"
         tx_ipsec: 10
         rx_ipsec: 10
  3. Send traffic through the IPsec tunnel. For example, ping a remote IP address:

    # ping -c 5 remote_ip_address
  4. Display the tx_ipsec and rx_ipsec counters of the active port again:

    # ethtool -S enp1s0 | egrep "_ipsec"
         tx_ipsec: 15
         rx_ipsec: 15

    If the counter values have increased, ESP hardware offload works.

Additional resources

21.3.13. Configuring IPsec connections that opt out of the system-wide crypto policies

Overriding system-wide crypto-policies for a connection

The RHEL system-wide cryptographic policies create a special connection called %default. This connection contains the default values for the ikev2, esp, and ike options. However, you can override the default values by specifying the mentioned option in the connection configuration file.

For example, the following configuration allows connections that use IKEv1 with AES and SHA-1 or SHA-2, and IPsec (ESP) with either AES-GCM or AES-CBC:

conn MyExample
  ...
  ikev2=never
  ike=aes-sha2,aes-sha1;modp2048
  esp=aes_gcm,aes-sha2,aes-sha1
  ...

Note that AES-GCM is available for IPsec (ESP) and for IKEv2, but not for IKEv1.

Disabling system-wide crypto policies for all connections

To disable system-wide crypto policies for all IPsec connections, comment out the following line in the /etc/ipsec.conf file:

include /etc/crypto-policies/back-ends/libreswan.config

Then add the ikev2=never option to your connection configuration file.

21.3.14. Troubleshooting IPsec VPN configurations

Problems related to IPsec VPN configurations most commonly occur due to several main reasons. If you are encountering such problems, you can check if the cause of the problem corresponds to any of the following scenarios, and apply the corresponding solution.

Basic connection troubleshooting

Most problems with VPN connections occur in new deployments, where administrators configured endpoints with mismatched configuration options. Also, a working configuration can suddenly stop working, often due to newly introduced incompatible values. This could be the result of an administrator changing the configuration. Alternatively, an administrator may have installed a firmware update or a package update with different default values for certain options, such as encryption algorithms.

To confirm that an IPsec VPN connection is established:

# ipsec trafficstatus
006 #8: "vpn.example.com"[1] 192.0.2.1, type=ESP, add_time=1595296930, inBytes=5999, outBytes=3231, id='@vpn.example.com', lease=100.64.13.5/32

If the output is empty or does not show an entry with the connection name, the tunnel is broken.

To check that the problem is in the connection:

  1. Reload the vpn.example.com connection:

    # ipsec auto --add vpn.example.com
    002 added connection description "vpn.example.com"
  2. Next, initiate the VPN connection:

    # ipsec auto --up vpn.example.com

Firewall-related problems

The most common problem is that a firewall on one of the IPsec endpoints or on a router between the endpoints is dropping all Internet Key Exchange (IKE) packets.

  • For IKEv2, an output similar to the following example indicates a problem with a firewall:

    # ipsec auto --up vpn.example.com
    181 "vpn.example.com"[1] 192.0.2.2 #15: initiating IKEv2 IKE SA
    181 "vpn.example.com"[1] 192.0.2.2 #15: STATE_PARENT_I1: sent v2I1, expected v2R1
    010 "vpn.example.com"[1] 192.0.2.2 #15: STATE_PARENT_I1: retransmission; will wait 0.5 seconds for response
    010 "vpn.example.com"[1] 192.0.2.2 #15: STATE_PARENT_I1: retransmission; will wait 1 seconds for response
    010 "vpn.example.com"[1] 192.0.2.2 #15: STATE_PARENT_I1: retransmission; will wait 2 seconds for
    ...
  • For IKEv1, the output of the initiating command looks like:

    # ipsec auto --up vpn.example.com
    002 "vpn.example.com" #9: initiating Main Mode
    102 "vpn.example.com" #9: STATE_MAIN_I1: sent MI1, expecting MR1
    010 "vpn.example.com" #9: STATE_MAIN_I1: retransmission; will wait 0.5 seconds for response
    010 "vpn.example.com" #9: STATE_MAIN_I1: retransmission; will wait 1 seconds for response
    010 "vpn.example.com" #9: STATE_MAIN_I1: retransmission; will wait 2 seconds for response
    ...

Because the IKE protocol, which is used to set up IPsec, is encrypted, you can troubleshoot only a limited subset of problems using the tcpdump tool. If a firewall is dropping IKE or IPsec packets, you can try to find the cause using the tcpdump utility. However, tcpdump cannot diagnose other problems with IPsec VPN connections.

  • To capture the negotiation of the VPN and all encrypted data on the eth0 interface:

    # tcpdump -i eth0 -n -n esp or udp port 500 or udp port 4500 or tcp port 4500

Mismatched algorithms, protocols, and policies

VPN connections require that the endpoints have matching IKE algorithms, IPsec algorithms, and IP address ranges. If a mismatch occurs, the connection fails. If you identify a mismatch by using one of the following methods, fix it by aligning algorithms, protocols, or policies.

  • If the remote endpoint is not running IKE/IPsec, you can see an ICMP packet indicating it. For example:

    # ipsec auto --up vpn.example.com
    ...
    000 "vpn.example.com"[1] 192.0.2.2 #16: ERROR: asynchronous network error report on wlp2s0 (192.0.2.2:500), complainant 198.51.100.1: Connection refused [errno 111, origin ICMP type 3 code 3 (not authenticated)]
    ...
  • Example of mismatched IKE algorithms:

    # ipsec auto --up vpn.example.com
    ...
    003 "vpn.example.com"[1] 193.110.157.148 #3: dropping unexpected IKE_SA_INIT message containing NO_PROPOSAL_CHOSEN notification; message payloads: N; missing payloads: SA,KE,Ni
  • Example of mismatched IPsec algorithms:

    # ipsec auto --up vpn.example.com
    ...
    182 "vpn.example.com"[1] 193.110.157.148 #5: STATE_PARENT_I2: sent v2I2, expected v2R2 {auth=IKEv2 cipher=AES_GCM_16_256 integ=n/a prf=HMAC_SHA2_256 group=MODP2048}
    002 "vpn.example.com"[1] 193.110.157.148 #6: IKE_AUTH response contained the error notification NO_PROPOSAL_CHOSEN

    A mismatched IKE version could also result in the remote endpoint dropping the request without a response. This looks identical to a firewall dropping all IKE packets.

  • Example of mismatched IP address ranges for IKEv2 (called Traffic Selectors - TS):

    # ipsec auto --up vpn.example.com
    ...
    1v2 "vpn.example.com" #1: STATE_PARENT_I2: sent v2I2, expected v2R2 {auth=IKEv2 cipher=AES_GCM_16_256 integ=n/a prf=HMAC_SHA2_512 group=MODP2048}
    002 "vpn.example.com" #2: IKE_AUTH response contained the error notification TS_UNACCEPTABLE
  • Example of mismatched IP address ranges for IKEv1:

    # ipsec auto --up vpn.example.com
    ...
    031 "vpn.example.com" #2: STATE_QUICK_I1: 60 second timeout exceeded after 0 retransmits.  No acceptable response to our first Quick Mode message: perhaps peer likes no proposal
  • When using PreSharedKeys (PSK) in IKEv1, if both sides do not put in the same PSK, the entire IKE message becomes unreadable:

    # ipsec auto --up vpn.example.com
    ...
    003 "vpn.example.com" #1: received Hash Payload does not match computed value
    223 "vpn.example.com" #1: sending notification INVALID_HASH_INFORMATION to 192.0.2.23:500
  • In IKEv2, the mismatched-PSK error results in an AUTHENTICATION_FAILED message:

    # ipsec auto --up vpn.example.com
    ...
    002 "vpn.example.com" #1: IKE SA authentication request rejected by peer: AUTHENTICATION_FAILED

Maximum transmission unit

Other than firewalls blocking IKE or IPsec packets, the most common cause of networking problems relates to an increased packet size of encrypted packets. Network hardware fragments packets larger than the maximum transmission unit (MTU), for example, 1500 bytes. Often, the fragments are lost and the packets fail to re-assemble. This leads to intermittent failures, when a ping test, which uses small-sized packets, works but other traffic fails. In this case, you can establish an SSH session but the terminal freezes as soon as you use it, for example, by entering the 'ls -al /usr' command on the remote host.

To work around the problem, reduce MTU size by adding the mtu=1400 option to the tunnel configuration file.

Alternatively, for TCP connections, enable an iptables rule that changes the MSS value:

# iptables -I FORWARD -p tcp --tcp-flags SYN,RST SYN -j TCPMSS --clamp-mss-to-pmtu

If the previous command does not solve the problem in your scenario, directly specify a lower size in the set-mss parameter:

# iptables -I FORWARD -p tcp --tcp-flags SYN,RST SYN -j TCPMSS --set-mss 1380

Network address translation (NAT)

When an IPsec host also serves as a NAT router, it could accidentally remap packets. The following example configuration demonstrates the problem:

conn myvpn
    left=172.16.0.1
    leftsubnet=10.0.2.0/24
    right=172.16.0.2
    rightsubnet=192.168.0.0/16
…

The system with address 172.16.0.1 have a NAT rule:

iptables -t nat -I POSTROUTING -o eth0 -j MASQUERADE

If the system on address 10.0.2.33 sends a packet to 192.168.0.1, then the router translates the source 10.0.2.33 to 172.16.0.1 before it applies the IPsec encryption.

Then, the packet with the source address 10.0.2.33 no longer matches the conn myvpn configuration, and IPsec does not encrypt this packet.

To solve this problem, insert rules that exclude NAT for target IPsec subnet ranges on the router, in this example:

iptables -t nat -I POSTROUTING -s 10.0.2.0/24 -d 192.168.0.0/16 -j RETURN

Kernel IPsec subsystem bugs

The kernel IPsec subsystem might fail, for example, when a bug causes a desynchronizing of the IKE user space and the IPsec kernel. To check for such problems:

$ cat /proc/net/xfrm_stat
XfrmInError                 0
XfrmInBufferError           0
...

Any non-zero value in the output of the previous command indicates a problem. If you encounter this problem, open a new support case, and attach the output of the previous command along with the corresponding IKE logs.

Libreswan logs

Libreswan logs using the syslog protocol by default. You can use the journalctl command to find log entries related to IPsec. Because the corresponding entries to the log are sent by the pluto IKE daemon, search for the “pluto” keyword, for example:

$ journalctl -b | grep pluto

To show a live log for the ipsec service:

$ journalctl -f -u ipsec

If the default level of logging does not reveal your configuration problem, enable debug logs by adding the plutodebug=all option to the config setup section in the /etc/ipsec.conf file.

Note that debug logging produces a lot of entries, and it is possible that either the journald or syslogd service rate-limits the syslog messages. To ensure you have complete logs, redirect the logging to a file. Edit the /etc/ipsec.conf, and add the logfile=/var/log/pluto.log in the config setup section.

Additional resources

21.4. Using MACsec to encrypt layer-2 traffic in the same physical network

You can use MACsec to secure the communication between two devices (point-to-point). For example, your branch office is connected over a Metro-Ethernet connection with the central office, you can configure MACsec on the two hosts that connect the offices to increase the security.

Media Access Control security (MACsec) is a layer 2 protocol that secures different traffic types over the Ethernet links including:

  • dynamic host configuration protocol (DHCP)
  • address resolution protocol (ARP)
  • Internet Protocol version 4 / 6 (IPv4 / IPv6) and
  • any traffic over IP such as TCP or UDP

MACsec encrypts and authenticates all traffic in LANs, by default with the GCM-AES-128 algorithm, and uses a pre-shared key to establish the connection between the participant hosts. If you want to change the pre-shared key, you need to update the NM configuration on all hosts in the network that uses MACsec.

A MACsec connection uses an Ethernet device, such as an Ethernet network card, VLAN, or tunnel device, as parent. You can either set an IP configuration only on the MACsec device to communicate with other hosts only using the encrypted connection, or you can also set an IP configuration on the parent device. In the latter case, you can use the parent device to communicate with other hosts using an unencrypted connection and the MACsec device for encrypted connections.

MACsec does not require any special hardware. For example, you can use any switch, except if you want to encrypt traffic only between a host and a switch. In this scenario, the switch must also support MACsec.

In other words, there are 2 common methods to configure MACsec;

  • host to host and
  • host to switch then switch to other host(s)
Important

You can use MACsec only between hosts that are in the same (physical or virtual) LAN.

21.4.1. Configuring a MACsec connection using nmcli

You can configure Ethernet interfaces to use MACsec using the nmcli utility. For example, you can create a MACsec connection between two hosts that are connected over Ethernet.

Procedure

  1. On the first host on which you configure MACsec:

    • Create the connectivity association key (CAK) and connectivity-association key name (CKN) for the pre-shared key:

      1. Create a 16-byte hexadecimal CAK:

        # dd if=/dev/urandom count=16 bs=1 2> /dev/null | hexdump -e '1/2 "%04x"'
        50b71a8ef0bd5751ea76de6d6c98c03a
      2. Create a 32-byte hexadecimal CKN:

        # dd if=/dev/urandom count=32 bs=1 2> /dev/null | hexdump -e '1/2 "%04x"'
        f2b4297d39da7330910a74abc0449feb45b5c0b9fc23df1430e1898fcf1c4550
  2. On both hosts you want to connect over a MACsec connection:
  3. Create the MACsec connection:

    # nmcli connection add type macsec con-name macsec0 ifname macsec0 connection.autoconnect yes macsec.parent enp1s0 macsec.mode psk macsec.mka-cak 50b71a8ef0bd5751ea76de6d6c98c03a macsec.mka-ckn f2b4297d39da7330910a7abc0449feb45b5c0b9fc23df1430e1898fcf1c4550

    Use the CAK and CKN generated in the previous step in the macsec.mka-cak and macsec.mka-ckn parameters. The values must be the same on every host in the MACsec-protected network.

  4. Configure the IP settings on the MACsec connection.

    1. Configure the IPv4 settings. For example, to set a static IPv4 address, network mask, default gateway, and DNS server to the macsec0 connection, enter:

      # nmcli connection modify macsec0 ipv4.method manual ipv4.addresses '192.0.2.1/24' ipv4.gateway '192.0.2.254' ipv4.dns '192.0.2.253'
    2. Configure the IPv6 settings. For example, to set a static IPv6 address, network mask, default gateway, and DNS server to the macsec0 connection, enter:

      # nmcli connection modify macsec0 ipv6.method manual ipv6.addresses '2001:db8:1::1/32' ipv6.gateway '2001:db8:1::fffe' ipv6.dns '2001:db8:1::fffd'
  5. Activate the connection:

    # nmcli connection up macsec0

Verification

  1. Verify that the traffic is encrypted:

    # tcpdump -nn -i enp1s0
  2. Optional: Display the unencrypted traffic:

    # tcpdump -nn -i macsec0
  3. Display MACsec statistics:

    # ip macsec show
  4. Display individual counters for each type of protection: integrity-only (encrypt off) and encryption (encrypt on)

    # ip -s macsec show