Layer security

GeoServer allows access to be determined on a per-layer basis.

Note

Layer security and Service Security cannot be combined. For example, it is not possible to specify access to a specific OWS service, only for one specific layer.

Providing access to layers is linked to roles. Layers and roles are linked in a file called layers.properties, which is located in the security directory in your GeoServer data directory. The file contains the rules that control access to workspaces and layers.

Note

The default layers security configuration in GeoServer allows any anonymous user to read data from any layer but only allows admin users to edit data.

Rules

The syntax for a layer security rule can follow three different patterns ([] denotes optional parameters):

globalLayerGroup.permission=role[,role2,...]
workspace.layer.permission=role[,role2,...]
workspace.workspaceLayerGroup.permission=role[,role2,...]

The parameters include:

  • globalLayerGroup— Name of a global layer group (one without workspace associated to it).

  • workspace— Name of the workspace. The wildcard * is used to indicate all workspaces.

  • layer— Name of a resource (featuretype/coverage/etc…). The wildcard * is used to indicate all layers.

  • workspaceLayerGroup— Name of a workspace specific layer group.

  • permission— Type of access permission/mode.

    • r—Read access

    • w—Write access

    • a—Admin access

    See Access modes for more details.

  • role[,role2,...] is the name(s) of predefined roles. The wildcard * is used to indicate the permission is applied to all users, including anonymous users.

Note

If a workspace or layer name is supposed to contain dots, they can be escaped using double backslashes (\\). For example, if a layer is named layer.with.dots the following syntax for a rule may be used:

topp.layer\\.with\\.dots.r=role[,role2,...]

General rules for layer security:

  • Each entry must have a unique combination of workspace, layer, and permission values.

  • If a permission at the global level is not specified, global permissions are assumed to allow read/write access.

  • If a permission for a workspace is not specified, it inherits permissions from the global specification.

  • If a permission for a layer is not specified, it inherits permissions from its workspace specification in all protocols except WMS (where layer groups rules play a role, see below).

  • If a user belongs to multiple roles, the least restrictive permission they inherit will apply.

For WMS, layers will be also secured by considering their containing layer groups. In particular:

  • Rules with Single layer groups only affect the group itself, but not its contents. Single mode is considered just an alias for a list of layers, with no containment.

  • Rules with other types of groups (Named tree, Container tree, Earth Observation tree) also affect contained layers and nested layer groups. If the group is not accessible, the layers and groups contained in that group will not be accessible either.. The only exception is when another layer group which is accessible contains the same layer or nested group, in that case the layers they will show up under the allowed groups.

  • Workspace rules gets precedence over global layer group ones when it comes to allow access to layers.

  • Layer rules get precedence over all layer group rules when it comes to allow access to layers.

The following tables summarizes the layer group behavior depending on whether they are used in a public or secured environment:

Mode

Public behavior

Restricted behavior

Single

Looks like a stand-alone layer, can be requested directly and acts as an alias for a layer list. Layers are also visible at root level

Does not control layer access at all

Opaque container

Looks like a stand-alone layer, can be requested directly and acts as an alias for a layer list. Layers in it are not available for WMS requests

Same as public behavior

Named tree

Contained layers are visible as children in the capabilities document, the name can be used as a shortcut to request all layers.

Restricting access to the group restricts also the contained layers, unless another “tree” group allows access to the same layer

Container tree

Same as “named tree”, but does not have a name published in the capabilities document and thus cannot be requested directly

Same as “named tree”

Catalog Mode

The layers.properties file may contain a further directive that specifies how GeoServer will advertise secured layers and behave when a secured layer is accessed without the necessary privileges. The parameter is mode and is commonly referred to as the “catalog mode”.

The syntax is:

mode=option

option may be one of three values:

Option

Description

hide

(Default) Hides layers that the user does not have read access to, and behaves as if a layer is read only if the user does not have write permissions. The capabilities documents will not contain the layers the current user cannot access. This is the highest security mode. As a result, it may not work very well with clients such as uDig or Google Earth.

challenge

Allows free access to metadata, but any attempt at accessing actual data is met by a HTTP 401 code (which forces most clients to show an authentication dialog). The capabilities documents contain the full list of layers. DescribeFeatureType and DescribeCoverage operations work successfully. This mode works fine with clients such as uDig or Google Earth.

mixed

Hides the layers the user cannot read from the capabilities documents, but triggers authentication for any other attempt to access the data or the metadata. This option is useful if you don’t want the world to see the existence of some of your data, but you still want selected people to who have data access links to get the data after authentication.

Access modes

The access mode defines what level of access should be granted on a specific workspace/layer to a particular role. There are three types of access mode:

  • rRead mode (read data from a workspace/layer)

  • wWrite mode (write data to a workspace/layer)

  • aAdmin mode (access and modify the configuration of a workspace/layer)

Some notes on the above access modes:

  • Write does not imply Read, but Admin implies both Write and Read.

  • Read and Write apply to the data of a layer, while Admin applies to the configuration of a layer.

  • As Admin mode only refers to the configuration of the layer, it is not required for any OGC service request.

Note

Currently, it is possible to assign Admin permission only to an entire workspace, and not to specific layers.

Examples

The following examples illustrate some possible layer restrictions and the corresponding rules.

Protecting a single workspace and a single layer

The following example demonstrates how to configure GeoServer as a primarily a read-only server:

*.*.r=*
*.*.w=NO_ONE
private.*.r=TRUSTED_ROLE
private.*.w=TRUSTED_ROLE
topp.congress_district.w=STATE_LEGISLATORS

The mapping of roles to permissions is as follows:

Role

private.*

topp.*

topp.congress_district

(all other workspaces)

NO_ONE

(none)

w

(none)

w

TRUSTED_ROLE

r/w

r

r

r

STATE_LEGISLATURES

(none)

r

r/w

r

(All other users)

r

r

r

r

Locking down GeoServer

The following example demonstrates how to lock down GeoServer:

*.*.r=TRUSTED_ROLE
*.*.w=TRUSTED_ROLE
topp.*.r=*
army.*.r=MILITARY_ROLE,TRUSTED_ROLE
army.*.w=MILITARY_ROLE,TRUSTED_ROLE

The mapping of roles to permissions is as follows:

Role

topp.*

army.*

(All other workspaces)

TRUSTED_ROLE

r/w

r/w

r/w

MILITARY_ROLE

r

r/w

(none)

(All other users)

r

(none)

(none)

Providing restricted administrative access

The following provides administrative access on a single workspace to a specific role, in additional to the full administrator role:

*.*.a=ROLE_ADMINISTRATOR
topp.*.a=ROLE_TOPP_ADMIN,ROLE_ADMINISTRATOR

Managing multi-level permissions

The following example demonstrates how to configure GeoServer with global-, workspace–, and layer-level permissions:

*.*.r=TRUSTED_ROLE
*.*.w=NO_ONE
topp.*.r=*
topp.states.r=USA_CITIZEN_ROLE,LAND_MANAGER_ROLE,TRUSTED_ROLE
topp.states.w=NO_ONE
topp.poly_landmarks.w=LAND_MANAGER_ROLE
topp.military_bases.r=MILITARY_ROLE
topp.military_bases.w=MILITARY_ROLE

The mapping of roles to permissions is as follows:

Role

topp.states

topp.poly_landmarks

topp.military_bases

topp.(all other layers)

(All other workspaces)

NO_ONE

w

r

(none)

w

w

TRUSTED_ROLE

r

r

(none)

r

r

MILITARY_ROLE

(none)

r

r/w

r

(none)

USA_CITIZEN_ROLE

r

r

(none)

r

(none)

LAND_MANAGER_ROLE

r

r/w

(none)

r

(none)

(All other users)

(none)

r

(none)

r

(none)

Note

The entry topp.states.w=NO_ONE is not required because this permission would be inherited from the global level (the entry *.*.w=NO_ONE).

Invalid configuration

The following examples are invalid because the workspace, layer, and permission combinations are not unique:

topp.state.rw=ROLE1
topp.state.rw=ROLE2,ROLE3

Security by layer group in WMS

To clarify, lets assume the following starting situation, in which all layers and groups are visible:

root
+- namedTreeGroupA
|   |   ws1:layerA
|   └   ws2:layerB
+- namedTreeGroupB
|   |   ws2:layerB
|   └   ws1:layerC
+- layerD
+- singleGroupC (contains ws1:layerA and layerD when requested)

Here are a few examples of how the structure changes based on different security rules.

  • Denying access to namedTreeGroupA by:

    namedTreeGroupA.r=ROLE_PRIVATE
    

    Will give the following capabilities tree to anonymous users:

    root
    +- namedTreeGroupB
    |   |   ws2:layerB
    |   └   ws1:layerC
    +- layerD
    +- singleGroupC (contains only layerD when requested)
    
  • Denying access to namedTreeGroupB by

    namedTreeGroupB.r=ROLE_PRIVATE
    

    Will give the following capabilities tree to anonymous users:

    root
    +- namedTreeGroupA
    |   |   ws1:layerA
    |   └   ws2:layerB
    +- layerD
    +- singleGroupC (contains ws1:layerA and layerD when requested)
    
  • Denying access to singleGroupC by:

    singleGroupC.r=ROLE_PRIVATE
    

    Will give the following capabilities tree to anonymous users:

    root
    +- namedTreeGroupA
    |   |   ws1:layerA
    |   └   ws2:layerB
    +- namedTreeGroupB
    |   |   ws2:layerB
    |   └   ws1:layerC
    +- layerD
    
  • Denying access to everything, but allowing explicit access to namedTreeGroupA by:

    nameTreeGroupA.r=*
    *.*.r=PRIVATE
    *.*.w=PRIVATE
    

    Will give the following capabilities tree to anonymous users:

    root
    +- namedTreeGroupA
        |   ws1:layerA
        └   ws2:layerB
    
  • Denying access to nameTreeA and namedTreeGroupB but explicitly allowing access to ws1:layerA:

    namedTreeGroupA.r=ROLE_PRIVATE
    namedTreeGroupB.r=ROLE_PRIVATE
    ws1.layerA.r=*
    

    Will give the following capabilities tree to anonymous users (notice how ws1:layerA popped up to the root):

    root
    +- ws1:layerA
    +- layerD
    
  • Denying access to nameTreeA and namedTreeGroupB but explicitly allowing all layers in ws2 (a workspace rules overrides global groups ones):

    namedTreeGroupA.r=ROLE_PRIVATE
    namedTreeGroupB.r=ROLE_PRIVATE
    ws2.*.r=*
    

    Will give the following capabilities tree to anonymous users (notice how ws1:layerB popped up to the root):

    root
    +- ws2:layerB
    +- layerD
    +- singleGroupC