OpenSearch Security Part 1: Concepts

OpenSearch comes with built-in security. The role-based security model makes it possible to implement fine-grained access control.

Reading time: 7 minutes
By Eliatra
For many years Elasticsearch had no built-in security. This led to numerous security breaches with millions of sensitive data leaked. Luckily, OpenSearch provides a strong and reliable security model out-of-the-box.
This post is part of a multi-article series that walks you through the setup of TLS encryption, users, roles, and permissions. We will first explain some of the core concepts of OpenSearch security, and in the next part show you how to set up a basic security model using the internal user database.

OpenSearch Security Basic Concepts

The security model of OpenSearch is based on the following core concepts:
    TLS encryption makes sure that no one can sniff or modify any data in motion.
    Users define who has access to an OpenSearch cluster. Before interacting with the cluster, any user has to authenticate first.
    Roles are used to implement authorisation. Any authenticated user can have one or more roles. Roles define what permissions a user has for particular indices.
    Permissions define what a user is allowed to do. For example, the permissions of a user may grant READ access to data, but do not allow WRITE or DELETE operations.
Apart from these basic security controls, OpenSearch also supports advanced features like Document- and Field-level security. We will cover those in follow-up articles as well.

TLS Encryption - REST Layer

OpenSearch uses TLS to secure and encrypt traffic between users and OpenSearch on the REST layer. This is much like accessing your online bank account with a browser using HTTPS: All traffic between your browser and the online banking server is encrypted. In addition, the browser checks and validates the server’s TLS certificate. By validating the TLS certificate you can be sure you are sending the data to the right entity, and not a malicious website.
The same happens when you send a request to OpenSearch, for example by using a browser or tools like curl. The HTTP traffic is encrypted, so no one can sniff or modify your data while it is being transmitted.

TLS Encryption - Transport Layer

Besides the REST layer which uses the HTTP protocol for communication, there is another internal layer where data is exchanged. This is the so-called transport layer. The transport layer is responsible for any traffic that is exchanged between the nodes in the cluster (“inter-node traffic”).
OpenSearch is a distributed system, so most operations like indexing data or querying for data affect multiple nodes in the cluster. As with the REST layer, OpenSearch uses TLS on the transport layer to encrypt and secure traffic, and to make sure only that trusted nodes are allowed to join your cluster.
By combining TLS encryption on both layers, all data in motion is end-to-end encrypted.
%%{init: { 'logLevel': 'debug' } }%% flowchart LR Z[Browser] <-- HTTP TLS encryption --> a[Cluster Node A] <-- Transport TLS encryption --> b[Cluster Node B]

Users, Roles, and Permissions

Request Flow

OpenSearch implements a role-based security model and supports multiple types of authentication and authorisation backends. The specific details on how a request is handled may vary depending on what technology you use, e.g. Active Directory, LDAP, OIDC, SAML, etc. However, the general request flow always follows the same pattern:
Any request that hits OpenSearch has to carry some sort of user credentials. For example, username and password as HTTP Basic Authentication header, a JSON web token or a Kerberos ticket. OpenSearch extracts this information from the request and validates it using an authentication domain. For example, the internal user database, an LDAP server, or Active Directory. If the credentials are valid, OpenSearch fetches the so-called backend roles of the user from an authorisation domain. Again, this could be LDAP roles, JWT claims, SAML assertions, or roles from the internal user database. This step is optional. Based on the user name and/or the backend roles, the user is assigned one or more security roles The security roles finally define what the user is allowed to do. You can set up access permissions for each index separately, or use wildcards and regular expressions to apply permissions to multiple indices at once.
sequenceDiagram autonumber participant User participant OpenSearch User->>OpenSearch: action: myindex/_search, user: john, pwd: doe participant Authentication Domain OpenSearch->>Authentication Domain: validate credentials Authentication Domain-->>OpenSearch: credentials ok participant Authorization Domain OpenSearch->>Authorization Domain: get roles of user john Authorization Domain-->>OpenSearch: ou=devops,dc=example,dc=com OpenSearch->>OpenSearch: map roles for ou=devops,dc=example,dc=com OpenSearch->>OpenSearch: result: OpenSearchDevOpsRole OpenSearch->>OpenSearch: get permissions for OpenSearchDevOpsRole OpenSearch->>OpenSearch: myindex/_search allowed for OpenSearchDevOpsRole? alt action allowed OpenSearch-->>User: Search Results else action not allowed OpenSearch-->>User: Security Exception end

User Management

OpenSearch supports multiple ways of managing users. Probably the easiest way is to use the built-in internal user database. The internal user database stores users and hashed passwords directly in a protected OpenSearch index. You can use the OpenSearch securityadmin CLI or the REST API to directly create, modify and delete users.
OpenSearch also supports a wide variety of other authentication domains, like:
    LDAP / Active Directory
    JSON web tokens
    Proxy authentication
    TLS client certificates
OpenSearch also provides the capability of chaining or combining authentication domains. For example, you can configure LDAP as first authentication domain, and the internal user database as second authentication domain. In this case, if the LDAP user authentication fails, OpenSearch will try the internal user database next. This provides for great flexibility, making it possible to implement most basic to very complex use-cases.
Authentication and authorization domains are configured in the file config.yml. After making changes here, do not forget to apply the changes by using the securityadmin CLI to upload changes.

Defining Security Roles

A security role defines what the bearer of this role is allowed to do. A role can grant access permissions for:
    Cluster actions
    Dashboard tenants
Cluster actions are actions that are not tied to specific indices, but rather operate on the cluster level. For example, health checks, cluster monitor actions, or cluster admin actions. They also include msearch or bulk actions.
Index-level permissions can be granted by index or by index pattern. For example, you can assign READ permissions for all indices starting with logstash like:
  - index_patterns:
    - "logstash-*"
    - "READ"
OpenSearch comes with a number of pre-defined sets of permissions, like READ, WRITE, CRUD, etc. These permission sets are called action groups and should cover most use cases.
If required, you can also use single actions and even define your own action groups for re-use across roles.
Security roles are configured in the file roles.yml. After making changes here, do not forget to apply them by using the securityadmin CLI.

Mapping Security Roles to Users

We have now discussed user management and role management. The last question is how to assign our roles to users.
If you use the internal user database, then you can directly assign security roles to users. Again, you can use the securityadmin CLI or the REST API. For example:
  hash: "..."
  - "my-role"
  - "my-other-role"  
Here, user myuser has two roles, my-role and my-other-role.
This is nice and straightforward. However, if your user base grows, or if you use LDAP or Active Directory, you may want to leverage the user mapping feature of OpenSearch.
Think of it like adding users to groups first, and then assign one or more security roles to those groups. This introduces a layer of indirection but provides more flexibility when changing permissions. Let’s say you have an LDAP and all employees that work in DevOps are members of this group:
You can use the name of this group to automatically assign security roles to all members. To do so, add an entry in the role_mapping.yml configuration file like:
Now all members of the LDAP group cn=devops,ou=it,dc=example,dc=com will be assigned to the security role my-devops-security-role.

Admin TLS Certificates

The last concept we want to cover is admin TLS certificates. In contrast to other security solutions, OpenSearch does not have the notion of a root user that has unlimited permissions. Instead, OpenSearch uses a client TLS certificate, the so-called admin TLS certificate.
This is a normal TLS certificate, signed by your root CA or intermediate CA. If a request to OpenSearch sends an admin TLS certificate, all permissions to the cluster are granted.
Using a TLS certificate for granting full access to the cluster provides a higher level of security than using a username/password combination.

Wrap Up

In this article, we introduced the building blocks of OpenSearch security: TLS encryption, users, roles, and permissions. In the next article, we will walk you through a basic security setup using the built-in demo configuration. We will then cover TLS certificates in detail before we go into more advanced security controls like document- and field-level security. Stay tuned!
Ready to get started?!
Let's work together to navigate your OpenSearch journey. Send us a message and talk to the team today!
Get in touch