IMPORTANT: No additional bug fixes or documentation updates will be released for this version. For the latest information, see the current release documentation.
« SAML authentication Integrating with other authentication systems »
Elastic Docs ›Elasticsearch Guide [7.11] ›Secure a cluster ›User authentication

Kerberos authentication

edit

Kerberos authentication

edit

You can configure the Elastic Stack security features to support Kerberos V5 authentication, an industry standard protocol to authenticate users in Elasticsearch.

You cannot use the Kerberos realm to authenticate on the transport network layer.

To authenticate users with Kerberos, you need to configure a Kerberos realm and map users to roles. For more information on realm settings, see Kerberos realm settings.

Key concepts

edit

There are a few terms and concepts that you’ll encounter when you’re setting up Kerberos realms:

kdc
Key Distribution Center. A service that issues Kerberos tickets.
principal

A Kerberos principal is a unique identity to which Kerberos can assign tickets. It can be used to identify a user or a service provided by a server.

Kerberos V5 principal names are of format primary/instance@REALM, where primary is a user name.

instance is an optional string that qualifies the primary and is separated by a slash(/) from the primary. For a user, usually it is not used; for service hosts, it is the fully qualified domain name of the host.

REALM is the Kerberos realm. Usually it is the domain name in upper case. An example of a typical user principal is user@ES.DOMAIN.LOCAL. An example of a typical service principal is HTTP/es.domain.local@ES.DOMAIN.LOCAL.

realm
Realms define the administrative boundary within which the authentication server has authority to authenticate users and services.
keytab
A file that stores pairs of principals and encryption keys.

Anyone with read permissions to this file can use the credentials in the network to access other services so it is important to protect it with proper file permissions.

krb5.conf
A file that contains Kerberos configuration information such as the default realm name, the location of Key distribution centers (KDC), realms information, mappings from domain names to Kerberos realms, and default configurations for realm session key encryption types.
ticket granting ticket (TGT)
A TGT is an authentication ticket generated by the Kerberos authentication server. It contains an encrypted authenticator.

Configuring a Kerberos realm

edit

Kerberos is used to protect services and uses a ticket-based authentication protocol to authenticate users. You can configure Elasticsearch to use the Kerberos V5 authentication protocol, which is an industry standard protocol, to authenticate users. In this scenario, clients must present Kerberos tickets for authentication.

In Kerberos, users authenticate with an authentication service and later with a ticket granting service to generate a TGT (ticket-granting ticket). This ticket is then presented to the service for authentication. Refer to your Kerberos installation documentation for more information about obtaining TGT. Elasticsearch clients must first obtain a TGT then initiate the process of authenticating with Elasticsearch.

Before you begin

edit

  1. Deploy Kerberos.

    You must have the Kerberos infrastructure set up in your environment.

    Kerberos requires a lot of external services to function properly, such as time synchronization between all machines and working forward and reverse DNS mappings in your domain. Refer to your Kerberos documentation for more details.

    These instructions do not cover setting up and configuring your Kerberos deployment. Where examples are provided, they pertain to an MIT Kerberos V5 deployment. For more information, see MIT Kerberos documentation

  2. Configure Java GSS.

    Elasticsearch uses Java GSS framework support for Kerberos authentication. To support Kerberos authentication, Elasticsearch needs the following files:

    • krb5.conf, a Kerberos configuration file
    • A keytab file that contains credentials for the Elasticsearch service principal

    The configuration requirements depend on your Kerberos setup. Refer to your Kerberos documentation to configure the krb5.conf file.

    For more information on Java GSS, see Java GSS Kerberos requirements

  3. Enable TLS for HTTP.

    If your Elasticsearch cluster is operating in production mode, you must configure the HTTP interface to use SSL/TLS before you can enable Kerberos authentication. For more information, see Encrypting HTTP client communications.

    This step is necessary to support Kerberos authentication via Kibana. It is not required for Kerberos authentication directly against the Elasticsearch Rest API.

  4. Enable the token service

    The Elasticsearch Kerberos implementation makes use of the Elasticsearch token service. If you configure TLS on the HTTP interface, this service is automatically enabled. It can be explicitly configured by adding the following setting in your elasticsearch.yml file:

    xpack.security.authc.token.enabled: true

    This step is necessary to support Kerberos authentication via Kibana. It is not required for Kerberos authentication directly against the Elasticsearch Rest API.

Create a Kerberos realm

edit

To configure a Kerberos realm in Elasticsearch:

  1. Configure the JVM to find the Kerberos configuration file.

    Elasticsearch uses Java GSS and JAAS Krb5LoginModule to support Kerberos authentication using a Simple and Protected GSSAPI Negotiation Mechanism (SPNEGO) mechanism. The Kerberos configuration file (krb5.conf) provides information such as the default realm, the Key Distribution Center (KDC), and other configuration details required for Kerberos authentication. When the JVM needs some configuration properties, it tries to find those values by locating and loading this file. The JVM system property to configure the file path is java.security.krb5.conf. To configure JVM system properties see Setting JVM options. If this system property is not specified, Java tries to locate the file based on the conventions.

    It is recommended that this system property be configured for Elasticsearch. The method for setting this property depends on your Kerberos infrastructure. Refer to your Kerberos documentation for more details.

    For more information, see krb5.conf

  2. Create a keytab for the Elasticsearch node.

    A keytab is a file that stores pairs of principals and encryption keys. Elasticsearch uses the keys from the keytab to decrypt the tickets presented by the user. You must create a keytab for Elasticsearch by using the tools provided by your Kerberos implementation. For example, some tools that create keytabs are ktpass.exe on Windows and kadmin for MIT Kerberos.

  3. Put the keytab file in the Elasticsearch configuration directory.

    Make sure that this keytab file has read permissions. This file contains credentials, therefore you must take appropriate measures to protect it.

    Elasticsearch uses Kerberos on the HTTP network layer, therefore there must be a keytab file for the HTTP service principal on every Elasticsearch node. The service principal name must have the format HTTP/es.domain.local@ES.DOMAIN.LOCAL. The keytab files are unique for each node since they include the hostname. An Elasticsearch node can act as any principal a client requests as long as that principal and its credentials are found in the configured keytab.

  4. Create a Kerberos realm.

    To enable Kerberos authentication in Elasticsearch, you must add a Kerberos realm in the realm chain.

    You can configure only one Kerberos realm on Elasticsearch nodes.

    To configure a Kerberos realm, there are a few mandatory realm settings and other optional settings that you need to configure in the elasticsearch.yml configuration file. Add a realm configuration under the xpack.security.authc.realms.kerberos namespace.

    The most common configuration for a Kerberos realm is as follows:

    xpack.security.authc.realms.kerberos.kerb1:
  order: 3
  keytab.path: es.keytab
  remove_realm_name: false

    The username is extracted from the ticket presented by user and usually has the format username@REALM. This username is used for mapping roles to the user. If realm setting remove_realm_name is set to true, the realm part (@REALM) is removed. The resulting username is used for role mapping.

    For detailed information of available realm settings, see Kerberos realm settings.

  5. Restart Elasticsearch

  6. Map Kerberos users to roles.

    The kerberos realm enables you to map Kerberos users to roles. You can configure these role mappings by using the create or update role mappings API. You identify users by their username field.

    The following example uses the role mapping API to map user@REALM to the roles monitoring and user:

    POST /_security/role_mapping/kerbrolemapping
{
  "roles" : [ "monitoring_user" ],
  "enabled": true,
  "rules" : {
    "field" : { "username" : "user@REALM" }
  }
}

    In case you want to support Kerberos cross realm authentication you may need to map roles based on the Kerberos realm name. For such scenarios following are the additional user metadata available for role mapping: - kerberos_realm will be set to Kerberos realm name. - kerberos_user_principal_name will be set to user principal name from the Kerberos ticket.

    For more information, see Mapping users and groups to roles.

    The Kerberos realm supports authorization realms as an alternative to role mapping.

Configure Kibana for Kerberos

edit

If you want to use Kerberos to authenticate via your browser and Kibana, you need to enable the relevant authentication provider in Kibana configuration. See kerberos single sign-on

« SAML authentication Integrating with other authentication systems »