Authentication Plugins

Stackable Authentication Method(s)

Since many institutions and organizations have existing authentication systems, DSpace has been designed to allow these to be easily integrated into an existing authentication infrastructure. It keeps a series, or "stack", of authentication methods, so each one can be tried in turn. This makes it easy to add new authentication methods or rearrange the order without changing any existing code. You can also share authentication code with other sites.

plugin.sequence.org.dspace.authenticate.AuthenticationMethod = \
 org.dspace.authenticate.PasswordAuthentication

The configuration property plugin.sequence.org.dspace.authenticate.AuthenticationMethod defines the authentication stack. It is a comma-separated list of class names. Each of these classes implements a different authentication method, or way of determining the identity of the user. They are invoked in the order specified until one succeeds.

Existing Authentication Methods include

• Authentication by Password (class: org.dspace.authenticate.PasswordAuthentication) (DEFAULT)
• Shibboleth Authentication (class: org.dspace.authenticate.ShibAuthentication)
• LDAP Authentication (class: org.dspace.authenticate.LDAPAuthentication)
• IP Address based Authentication (class: org.dspace.authenticate.IPAuthentication)
• X.509 Certificate Authentication (class: org.dspace.authenticate.X509Authentication)

An authentication method is a class that implements the interface org.dspace.authenticate.AuthenticationMethod. It authenticates a user by evaluating the credentials (e.g. username and password) he or she presents and checking that they are valid.

The basic authentication procedure in the DSpace Web UI is this:

1. A request is received from an end-user's browser that, if fulfilled, would lead to an action requiring authorization taking place.
2. If the end-user is already authenticated:
   • If the end-user is allowed to perform the action, the action proceeds
   • If the end-user is NOT allowed to perform the action, an authorization error is displayed.
   • If the end-user is NOT authenticated, i.e. is accessing DSpace anonymously:
3. The parameters etc. of the request are stored.
4. The Web UI's startAuthentication method is invoked.
5. First it tries all the authentication methods which do implicit authentication (i.e. they work with just the information already in the Web request, such as an X.509 client certificate). If one of these succeeds, it proceeds from Step 2 above.
6. If none of the implicit methods succeed, the UI responds by putting up a "login" page to collect credentials for one of the explicit authentication methods in the stack. The servlet processing that page then gives the proffered credentials to each authentication method in turn until one succeeds, at which point it retries the original operation from Step 2 above.
   Please see the source files AuthenticationManager.java and AuthenticationMethod.java for more details about this mechanism.

Authentication by Password

Enabling Authentication by Password

By default, this authentication method is enabled in DSpace.

However, to enable Authentication by Password, you must ensure the org.dspace.authenticate.PasswordAuthentication class is listed as one of the AuthenticationMethods in the following configuration:

plugin.sequence.org.dspace.authenticate.AuthenticationMethod = \
 org.dspace.authenticate.PasswordAuthentication

Configuring Authentication by Password

The default method org.dspace.authenticate.PasswordAuthentication has the following properties:

• Use of inbuilt e-mail address/password-based log-in. This is achieved by forwarding a request that is attempting an action requiring authorization to the password log-in servlet, /password-login. The password log-in servlet (org.dspace.app.webui.servlet.PasswordServlet) contains code that will resume the original request if authentication is successful, as per step 3. described above.
• Users can register themselves (i.e. add themselves as e-people without needing approval from the administrators), and can set their own passwords when they do this
• Users are not members of any special (dynamic) e-person groups
• You can restrict the domains from which new users are able to register. To enable this feature, uncomment the following line from dspace.cfg: authentication.password.domain.valid = example.com Example options might be '@example.com' to restrict registration to users with addresses ending in @example.com, or '@example.com, .ac.uk' to restrict registration to users with addresses ending in @example.com or with addresses in the .ac.uk domain.

A full list of all available Password Authentication Configurations:

Shibboleth Authentication

Enabling Shibboleth Authentication

To enable Shibboleth Authentication, you must ensure the org.dspace.authenticate.ShibAuthentication class is listed as one of the AuthenticationMethods in the following configuration:

plugin.sequence.org.dspace.authenticate.AuthenticationMethod = \
 org.dspace.authenticate.ShibAuthentication

Configuring Shibboleth Authentication

Shibboleth is a distributed authentication system for securely authenticating users and passing attributes about the user from one or more identity providers. In the Shibboleth terminology DSpace is a Service Provider which receives authentication information and then based upon that provides a service to the user. To use Shibboleth, DSpace requires that you use Apache installed with the mod_shib module acting as a proxy for all HTTP requests for your servlet container (typically Tomcat). DSpace will receive authentication information from the mod_shib module through HTTP headers.

Before DSpace will work with Shibboleth, you must have the following:

1. An Apache web server with the "mod_shib" module installed. As mentioned, this mod_shib module acts as a proxy for all HTTP requests for your servlet container (typically Tomcat).  Any requests to DSpace that require authentication via Shibboleth should be redirected to 'shibd' (the shibboleth daemon) by this "mod_shib" module.  Details on installing/configuring mod_shib in Apache are available at: https://wiki.shibboleth.net/confluence/display/SHIB2/NativeSPApacheConfig   We also have a sample Apache + mod_shib configuration provided below.
   1. Please note that in mod_shib the "ShibUseHeaders" must be set to "On", as DSpace currently receives authentication information through HTTP headers.
2. An external Shibboleth Idp (Identity Provider).  Using mod_shib, DSpace will only act as a Shibboleth SP (Service Provider). The actual Shibboleth Authentication & Identity information must be provided by an external IdP.  If you are using Shibboleth at your institution already, then there already should be a Shibboleth IdP available. More information about Shibboleth IdPs versus SPs is available at: https://wiki.shibboleth.net/confluence/display/SHIB2/UnderstandingShibboleth

For more information on installing and configuring a Shibboleth Service Provider see: https://wiki.shibboleth.net/confluence/display/SHIB2/Installation

Note about Shibboleth Active vs Lazy Sessions:

When configuring your Shibboleth Service Provider there are two Shibboleth paradigms you may use: Active or Lazy Sessions. Active sessions is where the mod_shib module is configured to product an entire URL space. No one will be able to access that URL without first authenticating with Shibboleth. Using this method you will need to configure shibboleth to protect the URL: "/shibboleth-login". The alternative, Lazy Session does not protect any specific URL. Instead Apache will allow access to any URL, and when the application wants to it may initiate an authenticated session.

The Lazy Session method is preferable for most DSpace installations, as you generally want to provide public access to (most) DSpace content, while access restricting only particular areas (e.g. administration UI/tools).  When Active Sessions are enabled, this will essentially access restrict your entire DSpace instance.  In other words, when using Active Sesssions, Shibboleth will require everyone to first authenticate before they can access any part of your repository (which essentially results in a "dark archive", since anonymous access will not be allowed).

Apache "mod_shib" Configuration (required)

As mentioned above, you must have Apache with the "mod_shib" module installed in order for DSpace to be able to act as a Shibboleth Service Provider (SP). The mod_shib module acts as a proxy for all HTTP requests for your servlet container (typically Tomcat).  Any requests to DSpace that require authentication via Shibboleth should be redirected to 'shibd' (the shibboleth daemon) by this "mod_shib" module.  Details on installing/configuring mod_shib in Apache are available at: https://wiki.shibboleth.net/confluence/display/SHIB2/NativeSPApacheConfig  General information about installing/configuring Shibboleth Service Providers (SPs) can be found at: https://wiki.shibboleth.net/confluence/display/SHIB2/Installation

A few extra notes/hints when configuring mod_shib & Apache:

• The "ShibUseHeaders" must be set to "On", as DSpace currently receives authentication information through HTTP headers.
• When initially setting up Apache & mod_shib, https://www.testshib.org/ provides a great testing ground for your configurations. This site provides a sample/demo Shibboleth IdP (as well as a sample Shibboleth SP) which you can test against. It acts as a "sandbox" to get your configurations working properly, before you point DSpace at your production Shibboleth IdP.

Below, we have provided a sample Apache configuration.  However, as every institution has their own specific Apache setup/configuration, it is highly likely that you will need to tweak this configuration in order to get it working properly. Again, see the official mod_shib documentation for much more detail about each of these settings: https://wiki.shibboleth.net/confluence/display/SHIB2/NativeSPApacheConfig  These configurations are meant to be added to an Apache <VirtualHost> which acts as a proxy to your Tomcat (or other servlet container) running DSpace.  More information on Apache VirtualHost settings can be found at: https://httpd.apache.org/docs/2.2/vhosts/

# While this sample VirtualHost is for HTTPS requests (recommended for Shibboleth, obviously), you may also need to create one for HTTP (*:80)
<VirtualHost *:443>
   ...
   # PLEASE NOTE: We have omitted many settings (ServerName, LogLevel, SSLCertificateFile, etc) 
   # which you may need/want to add to your VirtualHost
   
   # As long as Shibboleth module is installed, Enable all Shibboleth related settings
   <IfModule mod_shib>
       # Shibboleth recommends turning on UseCanonicalName
       # See "Prepping Apache" in https://wiki.shibboleth.net/confluence/display/SHIB2/NativeSPApacheConfig
       UseCanonicalName On

       # Most DSpace instances will want to use Shibboleth "Lazy Session", which ensures that users 
       # can access DSpace content without first authenticating via Shib. 
       # This section turns on Shibboleth "Lazy Session". Also ensures that once they have authenticated
       # (by accessing /Shibboleth.sso/Login path), then their Shib session is kept alive
       <Location />
         AuthType shibboleth
         ShibRequireSession Off
         require shibboleth
         # If your "shibboleth2.xml" file specifies an <ApplicationOverride> setting for your DSpace, 
         # then you may need to tell Apache which "id" to redirect Shib requests to. 
         # Just uncomment this and change the value "my-dspace-id" to the associated @id attribute value.
         #ShibRequestSetting applicationId my-dspace-id
       </Location>

       # If a user attempts to access the DSpace shibboleth login page, force them to authenticate via Shib
       <Location "/shibboleth-login">
         AuthType shibboleth
         ShibRequireSession On
         # DSpace requires using Shibboleth Headers. So this MUST be "On"
         ShibUseHeaders On
         require valid-user
       </Location>
         
       # Ensure /Shibboleth.sso path (in Apache) can be accessed
       # By default it may be inaccessible if your Apache security is tight.
       <Location "/Shibboleth.sso">
         Order deny,allow
         Allow from all
         # Also ensure Shibboleth/mod_shib responds to this path
         SetHandler shib
       </Location>
 
       # Finally, optionally ensure that requests to /Shibboleth.sso are NOT redirected 
       # to Tomcat (as they will be handled by mod_shib instead).
       # NOTE: THIS SETTING IS LIKELY ONLY NEEDED IF YOU ARE USING PROXYPASS TO REDIRECT
       # ALL REQUESTS TO TOMCAT (e.g. ProxyPass / ajp://localhost:8080/)
       # ProxyPass /Shibboleth.sso !
   </IfModule>
 
   ...
 
</VirtualHost>

DSpace Configuration Options

Authentication Methods:

DSpace supports authentication using NetID, or email address. A user's NetID is a unique identifier from the IdP that identifies a particular user. The NetID can be of almost any form such as a unique integer, string, or with Shibboleth 2.0 you can use "targeted ids". You will need to coordinate with your shibboleth federation or identity provider. There are three ways to supply identity information to DSpace:

1) NetID from Shibboleth Header (best)

The NetID-based method is superior because users may change their email address with the identity provider. When this happens DSpace will not be able to associate their new address with their old account.
 
2) Email address from Shibboleth Header (okay)

In the case where a NetID header is not available or not found DSpace will fall back to identifying a user based-upon their email address.
 
3) Tomcat's Remote User (worst)
 
In the event that neither Shibboleth headers are found then as a last resort DSpace will look at Tomcat's remote user field. This is the least attractive option because Tomcat has no way to supply additional attributes about a user. Because of this the autoregister option is not supported if this method is used.

Identity Scheme Migration Strategies:

If you are currently using Email based authentication (either 1 or 2) and want to upgrade to NetID based authentication then there is an easy path. Simply enable shibboleth to pass the NetID attribute and set the netid-header below to the correct value. When a user attempts to log in to DSpace first DSpace will look for an EPerson with the passed NetID, however when this fails DSpace will fall back to email based authentication. Then DSpace will update the user's EPerson account record to set their netted so all future authentications for this user will be based upon netted. One thing to note is that DSpace will prevent an account from switching NetIDs. If an account all ready has a NetID set and then they try and authenticate with a different NetID the authentication will fail.

EPerson Metadata:

One of the primary benefits of using Shibboleth based authentication is receiving additional attributes about users such as their names, telephone numbers, and possibly their academic department or graduation semester if desired. DSpace treats the first and last name attributes differently because they (along with email address) are the three pieces of minimal information required to create a new user account. For both first and last name supply direct mappings to the Shibboleth headers. In additional to the first and last name DSpace supports other metadata fields such as phone, or really anything you want to store on an eperson object. Beyond the phone field, which is accessible in the user's profile screen, none of these additional metadata fields will be used by DSpace out-of-the box. However if you develop any local modification you may access these attributes from the EPerson object. The Vireo ETD workflow system utilizes this to aid students when submitting an ETD.

Role-based Groups:

DSpace is able to place users into pre-defined groups based upon values received from Shibboleth. Using this option you can place all faculty members into a DSpace group when the correct affiliation's attribute is provided. When DSpace does this they are considered 'special groups', these are really groups but the user's membership within these groups is not recorded in the database. Each time a user authenticates they are automatically placed within the pre-defined DSpace group, so if the user loses their affiliation then the next time they login they will no longer be in the group.
 
Depending upon the shibboleth attributed use in the role-header it may be scoped. Scoped is shibboleth terminology for identifying where an attribute originated from. For example a students affiliation may be encoded as "student@tamu.edu". The part after the @ sign is the scope, and the preceding value is the value. You may use the whole value or only the value or scope. Using this you could generate a role for students and one institution different than students at another institution. Or if you turn on ignore-scope you could ignore the institution and place all students into one group.

The values extracted (a user may have multiple roles) will be used to look up which groups to place the user into. The groups are defined as "role.<role-name>" which is a comma separated list of  DSpace groups.

eperson.metadata = \
 SHIB-telephone => phone, \
 SHIB-cn => cn

role.faculty = Faculty, Member \
 role.staff = Staff, Member \
 role.student = Students, Member

LDAP Authentication

Enabling LDAP Authentication

To enable LDAP Authentication, you must ensure the org.dspace.authenticate.LDAPAuthentication class is listed as one of the AuthenticationMethods in the following configuration:

plugin.sequence.org.dspace.authenticate.AuthenticationMethod = \
 org.dspace.authenticate.LDAPAuthentication

Configuring LDAP Authentication

If LDAP is enabled, then new users will be able to register by entering their username and password without being sent the registration token. If users do not have a username and password, then they can still register and login with just their email address the same way they do now.

If you want to give any special privileges to LDAP users, create a stackable authentication method to automatically put people who have a netid into a special group. You might also want to give certain email addresses special privileges. Refer to the Custom Authentication Code section below for more information about how to do this.

Here is an explanation of each of the different LDAP configuration parameters:

Enabling Hierarchical LDAP Authentication

If your users are spread out across a hierarchical tree on your LDAP server, you may wish to have DSpace search for the user name in your tree. Here's how it works:

1. DSpace gets the user name from the login form
2. DSpace binds to LDAP as an administrative user with right to search in DNs (LDAP may be configured to allow anonymous users to search)
3. DSpace searches for the user name as within DNs (username is a part of full DN)
4. DSpace binds with the found full DN and password from login form
5. DSpace logs user in if LDAP reports successful authentication; refuses login otherwise

Configuring Hierarchical LDAP Authentication

Hierarchical LDAP Authentication shares all the above standard LDAP configurations, but has some additional settings.

You can optionally specify the search scope. If anonymous access is not enabled on your LDAP server, you will need to specify the full DN and password of a user that is allowed to bind in order to search for the users.

IP Authentication

Enabling IP Authentication

To enable IP Authentication, you must ensure the org.dspace.authenticate.IPAuthentication class is listed as one of the AuthenticationMethods in the following configuration:

plugin.sequence.org.dspace.authenticate.AuthenticationMethod = \
 org.dspace.authenticate.IPAuthentication

Configuring IP Authentication

Once enabled, you are then able to map DSpace groups to IP addresses in authentication-ip.cfg by setting ip.GROUPNAME = iprange[, iprange ...], e.g:

ip.MY_UNIVERSITY = 10.1.2.3, \                  # Full IP
13.5, \                      # Partial IP
11.3.4.5/24, \               # with CIDR
12.7.8.9/255.255.128.0, \    # with netmask
2001:18e8::32                # IPv6 too

Negative matches can be set by prepending the entry with a '-'. For example if you want to include all of a class B network except for users of a contained class c network, you could use: 111.222,-111.222.333.

Notes:

• If the Groupname contains blanks you must escape the spaces, e.g. "Department\ of\ Statistics"
• If your DSpace installation is hidden behind a web proxy, remember to set the useProxies configuration option within the 'Logging' section of dspace.cfg to use the IP address of the user rather than the IP address of the proxy server.

X.509 Certificate Authentication

Enabling X.509 Certificate Authentication

The X.509 authentication method uses an X.509 certificate sent by the client to establish his/her identity. It requires the client to have a personal Web certificate installed on their browser (or other client software) which is issued by a Certifying Authority (CA) recognized by the web server.

1. See the HTTPS installation instructions to configure your Web server. If you are using HTTPS with Tomcat, note that the <Connector> tag must include the attribute clientAuth="true" so the server requests a personal Web certificate from the client.

2. Add the org.dspace.authenticate.X509Authentication plugin first to the list of stackable authentication methods in the value of the configuration key plugin.sequence.org.dspace.authenticate.AuthenticationMethod

   Configuration File:	[dspace]/config/modules/authentication.cfg
   Property:	plugin.sequence.org.dspace.authenticate.AuthenticationMethod
   Example Value:	plugin.sequence.org.dspace.authenticate.AuthenticationMethod = \ org.dspace.authenticate.X509Authentication, \ org.dspace.authenticate.PasswordAuthentication

Configuring X.509 Certificate Authentication

1. You must also configure DSpace with the same CA certificates as the web server, so it can accept and interpret the clients' certificates. It can share the same keystore file as the web server, or a separate one, or a CA certificate in a file by itself. Configure it by oneof these methods, either the Java keystore

   keystore.path =  path to Java keystore file
   keystore.password =  password to access the keystore

   ...or the separate CA certificate file (in PEM or DER format):

   ca.cert =  path to certificate file for CA whose client certs to accept.

2. Choose whether to enable auto-registration: If you want users who authenticate successfully to be automatically registered as new E-Persons if they are not already, set the autoregister configuration property to true. This lets you automatically accept all users with valid personal certificates. The default is false.

Example of a Custom Authentication Method

Also included in the source is an implementation of an authentication method used at MIT, edu.mit.dspace.MITSpecialGroup. This does not actually authenticate a user, it only adds the current user to a special (dynamic) group called 'MIT Users' (which must be present in the system!). This allows us to create authorization policies for MIT users without having to manually maintain membership of the MIT users group.

By keeping this code in a separate method, we can customize the authentication process for MIT by simply adding it to the stack in the DSpace configuration. None of the code has to be touched.

You can create your own custom authentication method and add it to the stack. Use the most similar existing method as a model, e.g. org.dspace.authenticate.PasswordAuthentication for an "explicit" method (with credentials entered interactively) or org.dspace.authenticate.X509Authentication for an implicit method.