File-based access control#

To secure access to data in your cluster, you can implement file-based access control where access to data and operations is defined by rules declared in manually-configured JSON files.

There are two types of file-based access control:

System-level access control uses the access control plugin with a single JSON file that specifies authorization rules for the whole cluster.
Catalog-level access control uses individual JSON files for each catalog for granular control over the data in that catalog, including column-level authorization.

System-level access control files#

The access control plugin allows you to specify authorization rules for the cluster in a single JSON file.

Configuration#

To use the access control plugin, add an etc/access-control.properties file containing two required properties: access-control.name, which must be set to file, and security.config-file, which must be set to the location of the config file. The configuration file location can either point to the local disc or to a http endpoint. For example, if a config file named rules.json resides in etc, add an etc/access-control.properties with the following contents:

access-control.name=file
security.config-file=etc/rules.json

If the config should be loaded via the http endpoint http://trino-test/config and is wrapped into a JSON object and available via the data key etc/access-control.properties should look like this:

access-control.name=file
security.config-file=http://trino-test/config
security.json-pointer=/data

The config file is specified in JSON format. It contains rules that define which users have access to which resources. The rules are read from top to bottom and the first matching rule is applied. If no rule matches, access is denied. A JSON pointer (RFC 6901) can be specified using the security.json-pointer property to specify a nested object inside the JSON content containing the rules. Per default, the file is assumed to contain a single object defining the rules rendering the specification of security.json-pointer unnecessary in that case.

Refresh#

By default, when a change is made to the JSON rules file, Trino must be restarted to load the changes. There is an optional property to refresh the properties without requiring a Trino restart. The refresh period is specified in the etc/access-control.properties:

security.refresh-period=1s

Catalog, schema, and table access#

Access to catalogs, schemas, tables, and views is controlled by the catalog, schema, and table rules. The catalog rules are coarse-grained rules used to restrict all access or write access to catalogs. They do not explicitly grant any specific schema or table permissions. The table and schema rules are used to specify who can create, drop, alter, select, insert, delete, etc. for schemas and tables.

For each rule set, permission is based on the first matching rule, read from the top to the bottom of the configuration file. If no rule matches, access is denied.

If no rules are provided at all, then access is granted. You can remove access grant by adding a section with an empty set of rules at that particular level, for example:

{
  "schemas": []
}

At the catalog level you have to add a single “dummy” rule for each accessible catalog.

Note

These rules do not apply to system-defined tables in the information_schema schema.

The following table summarizes the permissions required for each SQL command:

SQL command	Catalog	Schema	Table	Note
SHOW CATALOGS				Always allowed
SHOW SCHEMAS	read-only	any*	any*	Allowed if catalog is visible
SHOW TABLES	read-only	any*	any*	Allowed if schema visible
CREATE SCHEMA	read-only	owner
DROP SCHEMA	all	owner
SHOW CREATE SCHEMA	all	owner
ALTER SCHEMA … RENAME TO	all	owner*		Ownership is required on both old and new schemas
ALTER SCHEMA … SET AUTHORIZATION	all	owner
CREATE TABLE	all		owner
DROP TABLE	all		owner
ALTER TABLE … RENAME TO	all		owner*	Ownership is required on both old and new tables
ALTER TABLE … SET PROPERTIES	all		owner
CREATE VIEW	all		owner
DROP VIEW	all		owner
ALTER VIEW … RENAME TO	all		owner*	Ownership is required on both old and new views
REFRESH MATERIALIZED VIEW	all		update
COMMENT ON TABLE	all		owner
COMMENT ON COLUMN	all		owner
ALTER TABLE … ADD COLUMN	all		owner
ALTER TABLE … DROP COLUMN	all		owner
ALTER TABLE … RENAME COLUMN	all		owner
SHOW COLUMNS	all		any
SELECT FROM table	read-only		select
SELECT FROM view	read-only		select, grant_select
INSERT INTO	all		insert
DELETE FROM	all		delete
UPDATE	all		update

Permissions required for executing functions:

SQL command	Catalog	Function permission	Note
`SELECT function()`		`execute`, `grant_execute*`	`grant_execute` is required when the function is used in a `SECURITY DEFINER` view.
`CREATE FUNCTION`	`all`	`ownership`	Not all connectors support catalog routines.
`DROP FUNCTION`	`all`	`ownership`	Not all connectors support catalog routines.

Visibility#

For a catalog, schema, or table to be visible in a SHOW command, the user must have at least one permission on the item or any nested item. The nested items do not need to already exist as any potential permission makes the item visible. Specifically:

catalog: Visible if user is the owner of any nested schema, has permissions on any nested table or function, or has permissions to set session properties in the catalog.
schema: Visible if the user is the owner of the schema, or has permissions on any nested table or function.
table: Visible if the user has any permissions on the table.

Catalog rules#

Each catalog rule is composed of the following fields:

user (optional): regex to match against user name. Defaults to .*.
role (optional): regex to match against role names. Defaults to .*.
group (optional): regex to match against group names. Defaults to .*.
catalog (optional): regex to match against catalog name. Defaults to .*.
allow (required): string indicating whether a user has access to the catalog. This value can be all, read-only or none, and defaults to none. Setting this value to read-only has the same behavior as the read-only system access control plugin.

In order for a rule to apply the user name must match the regular expression specified in user attribute.

For role names, a rule can be applied if at least one of the currently enabled roles matches the role regular expression.

For group names, a rule can be applied if at least one group name of this user matches the group regular expression.

The all value for allow means these rules do not restrict access in any way, but the schema and table rules can restrict access.

Note

By default, all users have access to the system catalog. You can override this behavior by adding a rule.

Boolean true and false are also supported as legacy values for allow, to support backwards compatibility. true maps to all, and false maps to none.

For example, if you want to allow only the role admin to access the mysql and the system catalog, allow users from the finance and human_resources groups access to postgres catalog, allow all users to access the hive catalog, and deny all other access, you can use the following rules:

{
  "catalogs": [
    {
      "role": "admin",
      "catalog": "(mysql|system)",
      "allow": "all"
    },
    {
      "group": "finance|human_resources",
      "catalog": "postgres",
      "allow": true
    },
    {
      "catalog": "hive",
      "allow": "all"
    },
    {
      "user": "alice",
      "catalog": "postgresql",
      "allow": "read-only"
    },
    {
      "catalog": "system",
      "allow": "none"
    }
  ]
}

For group-based rules to match, users need to be assigned to groups by a Group provider.

Schema rules#

Each schema rule is composed of the following fields:

user (optional): regex to match against user name. Defaults to .*.
role (optional): regex to match against role names. Defaults to .*.
group (optional): regex to match against group names. Defaults to .*.
catalog (optional): regex to match against catalog name. Defaults to .*.
schema (optional): regex to match against schema name. Defaults to .*.
owner (required): boolean indicating whether the user is to be considered an owner of the schema. Defaults to false.

For example, to provide ownership of all schemas to role admin, treat all users as owners of the default.default schema and prevent user guest from ownership of any schema, you can use the following rules:

{
  "schemas": [
    {
      "role": "admin",
      "schema": ".*",
      "owner": true
    },
    {
      "user": "guest",
      "owner": false
    },
    {
      "catalog": "default",
      "schema": "default",
      "owner": true
    }
  ]
}

Table rules#

Each table rule is composed of the following fields:

user (optional): regex to match against user name. Defaults to .*.
role (optional): regex to match against role names. Defaults to .*.
group (optional): regex to match against group names. Defaults to .*.
catalog (optional): regex to match against catalog name. Defaults to .*.
schema (optional): regex to match against schema name. Defaults to .*.
table (optional): regex to match against table names. Defaults to .*.
privileges (required): zero or more of SELECT, INSERT, DELETE, UPDATE, OWNERSHIP, GRANT_SELECT
columns (optional): list of column constraints.
filter (optional): boolean filter expression for the table.
filter_environment (optional): environment use during filter evaluation.

Column constraint#

These constraints can be used to restrict access to column data.

name: name of the column.
allow (optional): if false, column can not be accessed.
mask (optional): mask expression applied to column.
mask_environment (optional): environment use during mask evaluation.

Filter and mask environment#

user (optional): username for checking permission of subqueries in mask.

Note

These rules do not apply to information_schema.

mask can contain conditional expressions such as IF or CASE, which achieves conditional masking.

The example below defines the following table access policy:

Role admin has all privileges across all tables and schemas
User banned_user has no privileges
All users have SELECT privileges on default.hr.employees, but the table is filtered to only the row for the current user.
All users have SELECT privileges on all tables in the default.default schema, except for the address column which is blocked, and ssn which is masked.

{
  "tables": [
    {
      "role": "admin",
      "privileges": ["SELECT", "INSERT", "DELETE", "UPDATE", "OWNERSHIP"]
    },
    {
      "user": "banned_user",
      "privileges": []
    },
    {
      "catalog": "default",
      "schema": "hr",
      "table": "employee",
      "privileges": ["SELECT"],
      "filter": "user = current_user",
      "filter_environment": {
        "user": "system_user"
      }
    },
    {
      "catalog": "default",
      "schema": "default",
      "table": ".*",
      "privileges": ["SELECT"],
      "columns" : [
         {
            "name": "address",
            "allow": false
         },
         {
            "name": "SSN",
            "mask": "'XXX-XX-' + substring(credit_card, -4)",
            "mask_environment": {
              "user": "system_user"
            }
         }
      ]
    }
  ]
}

Function rules#

These rules control the ability of a user to create, drop, and execute functions.

When these rules are present, the authorization is based on the first matching rule, processed from top to bottom. If no rules match, the authorization is denied. If function rules are not present, only functions insystem.builtin can be executed.

Note

Users always have access to functions in the system.builtin schema, and you cannot override this behavior by adding a rule.

Each function rule is composed of the following fields:

user (optional): regular expression to match against user name. Defaults to .*.
role (optional): regular expression to match against role names. Defaults to .*.
group (optional): regular expression to match against group names. Defaults to .*.
catalog (optional): regular expression to match against catalog name. Defaults to .*.
schema (optional): regular expression to match against schema name. Defaults to .*.
function (optional): regular expression to match against function names. Defaults to .*.
privileges (required): zero or more of EXECUTE, GRANT_EXECUTE, OWNERSHIP.

Care should be taken when granting permission to the system schema of a catalog, as this is the schema Trino uses for table function such as query. These table functions can be used to access or modify the underlying data of the catalog.

The following example allows the admin user to execute system.query table function in any catalog, and allows all users to create, drop, and execute functions (including SECURITY DEFINER views) in the hive.function schema:

{
  "functions": [
    {
      "user": "admin",
      "schema": "system",
      "function": "query",
      "privileges": [
        "EXECUTE"
      ]
    },
    {
      "catalog": "hive",
      "schema": "function",
      "privileges": [
        "EXECUTE", "GRANT_EXECUTE", "OWNERSHIP"
      ]
    }
  ]
}

Verify configuration#

To verify the system-access control file is configured properly, set the rules to completely block access to all users of the system:

{
  "catalogs": [
    {
      "catalog": "system",
      "allow": "none"
    }
  ]
}

Restart your cluster to activate the rules for your cluster. With the Trino CLI run a query to test authorization:

trino> SELECT * FROM system.runtime.nodes;
Query 20200824_183358_00000_c62aw failed: Access Denied: Cannot access catalog system

Remove these rules and restart the Trino cluster.

Session property rules#

These rules control the ability of a user to set system and catalog session properties. The user is granted or denied access, based on the first matching rule, read from top to bottom. If no rules are specified, all users are allowed set any session property. If no rule matches, setting the session property is denied. System session property rules are composed of the following fields:

user (optional): regex to match against user name. Defaults to .*.
role (optional): regex to match against role names. Defaults to .*.
group (optional): regex to match against group names. Defaults to .*.
property (optional): regex to match against the property name. Defaults to .*.
allow (required): boolean indicating if the setting the session property should be allowed.

The catalog session property rules have the additional field:

catalog (optional): regex to match against catalog name. Defaults to .*.

The example below defines the following table access policy:

Role admin can set all session property
User banned_user can not set any session properties
All users can set the resource_overcommit system session property, and the bucket_execution_enabled session property in the hive catalog.

{
    "system_session_properties": [
        {
            "role": "admin",
            "allow": true
        },
        {
            "user": "banned_user",
            "allow": false
        },
        {
            "property": "resource_overcommit",
            "allow": true
        }
    ],
    "catalog_session_properties": [
        {
            "role": "admin",
            "allow": true
        },
        {
            "user": "banned_user",
            "allow": false
        },
        {
            "catalog": "hive",
            "property": "bucket_execution_enabled",
            "allow": true
        }
    ]
}

Query rules#

These rules control the ability of a user to execute, view, or kill a query. The user is granted or denied access, based on the first matching rule read from top to bottom. If no rules are specified, all users are allowed to execute queries, and to view or kill queries owned by any user. If no rule matches, query management is denied. Each rule is composed of the following fields:

user (optional): regex to match against user name. Defaults to .*.
role (optional): regex to match against role names. Defaults to .*.
group (optional): regex to match against group names. Defaults to .*.
queryOwner (optional): regex to match against the query owner name. Defaults to .*.
allow (required): set of query permissions granted to user. Values: execute, view, kill

Note

Users always have permission to view or kill their own queries.

A rule that includes queryOwner may not include the execute access mode. Queries are only owned by a user once their execution has begun.

For example, if you want to allow the role admin full query access, allow the user alice to execute and kill queries, allow members of the group contractors to view queries owned by users alice or dave, allow any user to execute queries, and deny all other access, you can use the following rules:

{
  "queries": [
    {
      "role": "admin",
      "allow": ["execute", "kill", "view"]
    },
    {
      "user": "alice",
      "allow": ["execute", "kill"]
    },
    {
      "group": "contractors",
      "queryOwner": "alice|dave",
      "allow": ["view"]
    },
    {
      "allow": ["execute"]
    }
  ]
}

Impersonation rules#

These rules control the ability of a user to impersonate another user. In some environments it is desirable for an administrator (or managed system) to run queries on behalf of other users. In these cases, the administrator authenticates using their credentials, and then submits a query as a different user. When the user context is changed, Trino verifies that the administrator is authorized to run queries as the target user.

When these rules are present, the authorization is based on the first matching rule, processed from top to bottom. If no rules match, the authorization is denied. If impersonation rules are not present but the legacy principal rules are specified, it is assumed impersonation access control is being handled by the principal rules, so impersonation is allowed. If neither impersonation nor principal rules are defined, impersonation is not allowed.

Each impersonation rule is composed of the following fields:

original_user (optional): regex to match against the user requesting the impersonation. Defaults to .*.
original_role (optional): regex to match against role names of the requesting impersonation. Defaults to .*.
new_user (required): regex to match against the user to impersonate. Can contain references to subsequences captured during the match against original_user, and each reference is replaced by the result of evaluating the corresponding group respectively.
allow (optional): boolean indicating if the authentication should be allowed. Defaults to true.

The impersonation rules are a bit different than the other rules: The attribute new_user is required to not accidentally prevent more access than intended. Doing so it was possible to make the attribute allow optional.

The following example allows the admin role, to impersonate any user, except for bob. It also allows any user to impersonate the test user. It also allows a user in the form team_backend to impersonate the team_backend_sandbox user, but not arbitrary users:

{
    "impersonation": [
        {
            "original_role": "admin",
            "new_user": "bob",
            "allow": false
        },
        {
            "original_role": "admin",
            "new_user": ".*"
        },
        {
            "original_user": ".*",
            "new_user": "test"
        },
        {
            "original_user": "team_(.*)",
            "new_user": "team_$1_sandbox",
            "allow": true
        }
    ]
}

Principal rules#

Warning

Principal rules are deprecated. Instead, use User mapping which specifies how a complex authentication user name is mapped to a simple user name for Trino, and impersonation rules defined above.

These rules serve to enforce a specific matching between a principal and a specified user name. The principal is granted authorization as a user, based on the first matching rule read from top to bottom. If no rules are specified, no checks are performed. If no rule matches, user authorization is denied. Each rule is composed of the following fields:

principal (required): regex to match and group against principal.
user (optional): regex to match against user name. If matched, it grants or denies the authorization based on the value of allow.
principal_to_user (optional): replacement string to substitute against principal. If the result of the substitution is same as the user name, it grants or denies the authorization based on the value of allow.
allow (required): boolean indicating whether a principal can be authorized as a user.

Note

You would at least specify one criterion in a principal rule. If you specify both criteria in a principal rule, it returns the desired conclusion when either of criteria is satisfied.

The following implements an exact matching of the full principal name for LDAP and Kerberos authentication:

{
  "principals": [
    {
      "principal": "(.*)",
      "principal_to_user": "$1",
      "allow": true
    },
    {
      "principal": "([^/]+)(/.*)?@.*",
      "principal_to_user": "$1",
      "allow": true
    }
  ]
}

If you want to allow users to use the exact same name as their Kerberos principal name, and allow alice and bob to use a group principal named as group@example.net, you can use the following rules.

{
  "principals": [
    {
      "principal": "([^/]+)/?.*@example.net",
      "principal_to_user": "$1",
      "allow": true
    },
    {
      "principal": "group@example.net",
      "user": "alice|bob",
      "allow": true
    }
  ]
}

System information rules#

These rules specify which users can access the system information management interface. System information access includes the following aspects:

Read access to details such as Trino version, uptime of the node, and others from the /v1/info and /v1/status REST endpoints.
Read access with the system information functions.
Read access with the System connector.
Write access to trigger Graceful shutdown.

The user is granted or denied access based on the first matching rule read from top to bottom. If no rules are specified, all access to system information is denied. If no rule matches, system access is denied. Each rule is composed of the following fields:

role (optional): regex to match against role. If matched, it grants or denies the authorization based on the value of allow.
user (optional): regex to match against user name. If matched, it grants or denies the authorization based on the value of allow.
allow (required): set of access permissions granted to user. Values: read, write

The following configuration provides and example:

{
  "system_information": [
    {
      "role": "admin",
      "allow": ["read", "write"]
    },
    {
      "user": "alice",
      "allow": ["read"]
    }
  ]
}

All users with the admin role have read and write access to system information. This includes the ability to trigger Graceful shutdown.
The user alice can read system information.
All other users and roles are denied access to system information.

A fixed user can be set for management interfaces using the management.user configuration property. When this is configured, system information rules must still be set to authorize this user to read or write to management information. The fixed management user only applies to HTTP by default. To enable the fixed user over HTTPS, set the management.user.https-enabled configuration property.

Authorization rules#

These rules control the ability of how owner of schema, table or view can be altered. These rules are applicable to commands like:

ALTER SCHEMA name SET AUTHORIZATION ( user | USER user | ROLE role ) ALTER TABLE name SET AUTHORIZATION ( user | USER user | ROLE role ) ALTER VIEW name SET AUTHORIZATION ( user | USER user | ROLE role )

When these rules are present, the authorization is based on the first matching rule, processed from top to bottom. If no rules match, the authorization is denied.

Notice that in order to execute ALTER command on schema, table or view user requires OWNERSHIP privilege.

Each authorization rule is composed of the following fields:

original_user (optional): regex to match against the user requesting the authorization. Defaults to .*.
original_group (optional): regex to match against group names of the requesting authorization. Defaults to .*.
original_role (optional): regex to match against role names of the requesting authorization. Defaults to .*.
new_user (optional): regex to match against the new owner user of the schema, table or view. By default it does not match.
new_role (optional): regex to match against the new owner role of the schema, table or view. By default it does not match.
allow (optional): boolean indicating if the authentication should be allowed. Defaults to true.

Notice that new_user and new_role are optional, however it is required to provide at least one of them.

The following example allows the admin role, to change owner of any schema, table or view to any user, except to``bob``.

{
  "authorization": [
    {
      "original_role": "admin",
      "new_user": "bob",
      "allow": false
    },
    {
      "original_role": "admin",
      "new_user": ".*",
      "new_role": ".*"
    }
  ],
  "schemas": [
    {
      "role": "admin",
      "owner": true
    }
  ],
  "tables": [
    {
      "role": "admin",
      "privileges": ["OWNERSHIP"]
    }
  ]
}

Catalog-level access control files#

You can create JSON files for individual catalogs that define authorization rules specific to that catalog. To enable catalog-level access control files, add a connector-specific catalog configuration property that sets the authorization type to FILE and the security.config-file catalog configuration property that specifies the JSON rules file.

For example, the following Iceberg catalog configuration properties use the rules.json file for catalog-level access control:

iceberg.security=FILE
security.config-file=etc/catalog/rules.json

Catalog-level access control files are supported on a per-connector basis, refer to the connector documentation for more information.

Note

These rules do not apply to system-defined tables in the information_schema schema.

Configure a catalog rules file#

The configuration file is specified in JSON format. This file is composed of the following sections, each of which is a list of rules that are processed in order from top to bottom:

schemas
tables
session_properties

The user is granted the privileges from the first matching rule. All regexes default to .* if not specified.

Schema rules#

These rules govern who is considered an owner of a schema.

user (optional): regex to match against user name.
group (optional): regex to match against every user group the user belongs to.
schema (optional): regex to match against schema name.
owner (required): boolean indicating ownership.

Table rules#

These rules govern the privileges granted on specific tables.

user (optional): regex to match against user name.
group (optional): regex to match against every user group the user belongs to.
schema (optional): regex to match against schema name.
table (optional): regex to match against table name.
privileges (required): zero or more of SELECT, INSERT, DELETE, UPDATE, OWNERSHIP, GRANT_SELECT.
columns (optional): list of column constraints.
filter (optional): boolean filter expression for the table.
filter_environment (optional): environment used during filter evaluation.

Column constraints#

These constraints can be used to restrict access to column data.

name: name of the column.
allow (optional): if false, column can not be accessed.
mask (optional): mask expression applied to column.
mask_environment (optional): environment use during mask evaluation.

Filter environment and mask environment#

These rules apply to filter_environment and mask_environment.

user (optional): username for checking permission of subqueries in a mask.

Note

mask can contain conditional expressions such as IF or CASE, which achieves conditional masking.

Function rules#

These rules control the ability of a user to create, drop, and execute functions.

When these rules are present, the authorization is based on the first matching rule, processed from top to bottom. If no rules match, the authorization is denied. If function rules are not present, access is not allowed.

user (optional): regular expression to match against user name. Defaults to .*.
group (optional): regular expression to match against group names. Defaults to .*.
schema (optional): regular expression to match against schema name. Defaults to .*.
function (optional): regular expression to match against function names. Defaults to .*.
privileges (required): zero or more of EXECUTE, GRANT_EXECUTE, OWNERSHIP.

Care should be taken when granting permission to the system schema of a catalog, as this is the schema Trino uses for table function such as query. These table functions can be used to access or modify the underlying data of the catalog.

The following example allows the admin user to execute system.query table function from any catalog, and all users to create, drop, and execute functions (including from views) in the function schema of this catalog:

{
  "functions": [
    {
      "user": "admin",
      "schema": "system",
      "function": "query",
      "privileges": [
        "EXECUTE"
      ]
    },
    {
      "schema": "function",
      "privileges": [
        "EXECUTE", "GRANT_EXECUTE", "OWNERSHIP"
      ]
    }
  ]
}

Session property rules#

These rules govern who may set session properties.

user (optional): regex to match against user name.
group (optional): regex to match against every user group the user belongs to.
property (optional): regex to match against session property name.
allow (required): boolean indicating whether this session property may be set.

Example#

{
  "schemas": [
    {
      "user": "admin",
      "schema": ".*",
      "owner": true
    },
    {
      "group": "finance|human_resources",
      "schema": "employees",
      "owner": true
    },
    {
      "user": "guest",
      "owner": false
    },
    {
      "schema": "default",
      "owner": true
    }
  ],
  "tables": [
    {
      "user": "admin",
      "privileges": ["SELECT", "INSERT", "DELETE", "UPDATE", "OWNERSHIP"]
    },
    {
      "user": "banned_user",
      "privileges": []
    },
    {
      "schema": "hr",
      "table": "employee",
      "privileges": ["SELECT"],
      "filter": "user = current_user"
    },
    {
      "schema": "default",
      "table": ".*",
      "privileges": ["SELECT"],
      "columns" : [
         {
            "name": "address",
            "allow": false
         },
         {
            "name": "ssn",
            "mask": "'XXX-XX-' + substring(credit_card, -4)",
            "mask_environment": {
              "user": "admin"
            }
         }
      ]
    }
  ],
  "session_properties": [
    {
      "property": "force_local_scheduling",
      "allow": true
    },
    {
      "user": "admin",
      "property": "max_split_size",
      "allow": true
    }
  ]
}