Master_Thesis_2015_by_Sanjeev_Laha_21229267

ISSN 1612-6793
Nr. ZAI-MSC-2015-03
Master’s Thesis
submitted in partial fulﬁlment of the
requirements for the course “Applied Computer Science”
Framework to share group information using
heterogeneous group management systems in
shibboleth
Sanjeev Laha
Gesellschaft für wissenschaftliche
Datenverarbeitung mbH Göttingen (GWDG)
Bachelor’s and Master’s Theses
of the Center for Computational Sciences
at the Georg-August-Universität Göttingen
01. June 2015

Georg-August-Universität Göttingen
Institute of Computer Science
Goldschmidtstraße 7
37077 Göttingen
Germany
+49 (551) 39-172000
+49 (551) 39-14403
ofﬁce@informatik.uni-goettingen.de
www.informatik.uni-goettingen.de
First Supervisor: Prof. Dr. Ramin Yahyapour
Second Supervisor: Dr. Sven Bingert

I hereby declare that I have written this thesis independently without any help from others and
without the use of documents or aids other than those stated. I have mentioned all used sources
and cited them correctly according to established academic citation rules.
Göttingen, 01. June 2015

Masterarbeit
im Studiengang „Angewandte Informatik“
Framework to share group information
using heterogeneous group management
systems in shibboleth
Sanjeev Laha
01 June 2015
Supervisor
Prof. Dr. Ramin Yahyapour
Department for Practical Informatics
Georg-August-Universität Göttingen
Second supervisor
Dr. Sven Bingert
Gesellschaft für wissenschaftliche Datenverarbeitung mbH
Göttingen

Abstract
Federated identity management using shibboleth is quite popular in the ﬁeld of identity
management, especially among federated education and research institutes. Authenti-
cation and authorization are two important aspects of service delegation. While both
authentication and authorization are set by an administrator, there is very less scope
for self authorization of services by users themselves. This thesis provides an archi-
tecture and solution for implementing a system using which group administrators can
authorize group members to use a particular service. The group information system
will work in concurrent with shibboleth. Group information along with identity infor-
mation will be shared by shibboleth identity provider with shibboleth service provider.
Group information will be sourced from multiple heterogeneous group management
systems held at the organization of shibboleth identity provider.

Contents
1. Introduction 1
1.1. Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2. Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2. Fundamentals 5
2.1. Shibboleth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.1.1. Shibboleth Service Provider . . . . . . . . . . . . . . . . . . . . . . 6
2.1.2. Shibboleth Identity Provider . . . . . . . . . . . . . . . . . . . . . 6
2.1.3. WAYF . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.1.4. Flow of data and control in Shibboleth . . . . . . . . . . . . . . . . 8
2.2. Mailman . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.3. UserFrosting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.4. LDAP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.5. Active Directory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3. Heterogeneous group management 14
3.1. Heterogeneous group sharing and advantages . . . . . . . . . . . . . . . 14
3.1.1. Group sharing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.1.2. Heterogeneous group sharing . . . . . . . . . . . . . . . . . . . . . 16
3.1.3. Usage at GWDG . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.2. Requirements Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.2.1. Business processes . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.2.2. IT processes supporting business processes . . . . . . . . . . . . . 19
3.2.3. Activities within each IT process . . . . . . . . . . . . . . . . . . . 19
3.2.4. Functions within each activity . . . . . . . . . . . . . . . . . . . . 19
3.2.5. Use cases for the functions . . . . . . . . . . . . . . . . . . . . . . . 20
4. Implementation 22
4.1. Core elements and architecture . . . . . . . . . . . . . . . . . . . . . . . . 22
4.2. Group information providers UI . . . . . . . . . . . . . . . . . . . . . . . 24
4.2.1. Mailman . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.2.2. UserFrosting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.2.3. Light Weight Directory Access Protocol . . . . . . . . . . . . . . . 25
4.2.4. Active Directory . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
4.3. API development . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
4.3.1. Basics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
iv

Contents
4.3.2. Mailman API . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
4.3.3. UserFrosting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.3.4. LDAP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.3.5. Active Directory . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
4.4. Single logout . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
4.5. Attribute synchronizer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
4.6. Group manager . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
5. Testing group sharing 42
6. Preparations for production system 45
7. Similar work 46
8. Conclusion 48
References 48
A. Installation Guides 52
A.1. AttributeQuery handler plug-in shibboleth SP . . . . . . . . . . . . . . . 52
A.2. Shibboleth 2.x Installation . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
A.2.1. Basic installation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
A.2.1.1. System requirements . . . . . . . . . . . . . . . . . . . . 53
A.2.2. Shibboleth Identity provider Installation & Conﬁguration . . . . 54
A.2.3. Shibboleth Identity provider source code installation: . . . . . . . 54
A.2.4. Problem while building in Shibboleth 2.4.3 . . . . . . . . . . . . . 55
A.2.5. Problem with resulting jar ﬁle in Shibboleth 2.4.3 . . . . . . . . . 55
A.2.6. Analysis for solution of problem: . . . . . . . . . . . . . . . . . . . 55
A.2.7. Service provider installation . . . . . . . . . . . . . . . . . . . . . . 56
A.3. Mailman . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
v

List of Figures
1.1. Federated identity management example . . . . . . . . . . . . . . . . . . 2
1.2. Federated Identity Management with group information sharing . . . . 3
2.1. Shibboleth Architecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2. Shibboleth control ﬂow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.3. Mailman Umbrella list . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.4. LDAP directory tree . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.5. Active directory domain forest . . . . . . . . . . . . . . . . . . . . . . . . 12
3.1. Beneﬁts of sharing group information . . . . . . . . . . . . . . . . . . . . 15
3.2. Heterogeneous group management system . . . . . . . . . . . . . . . . . 16
3.3. Functional Requirement Analysis . . . . . . . . . . . . . . . . . . . . . . . 18
3.4. Use case for functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
4.1. Modules used in group sharing framework . . . . . . . . . . . . . . . . . 22
4.2. Flow chart for group management . . . . . . . . . . . . . . . . . . . . . . 38
4.3. Flow chart for access management . . . . . . . . . . . . . . . . . . . . . . 40
5.1. Shibboleth IdP login response time . . . . . . . . . . . . . . . . . . . . . . 43
5.2. Shibboleth SP attribute synchronizer response time . . . . . . . . . . . . 44
7.1. Grouper based federated group sharing . . . . . . . . . . . . . . . . . . . 46
vi

List of Tables
4.1. Identity provider extension parts . . . . . . . . . . . . . . . . . . . . . . . 27
A.1. System Requirements for Shibboleth Installation . . . . . . . . . . . . . . 53
A.2. Shibboleth IdP conﬁguration changes . . . . . . . . . . . . . . . . . . . . 55
A.3. Shibboleth IdP conﬁguration changes . . . . . . . . . . . . . . . . . . . . 57
vii

List of Code Snippet
4.1. MySQL attribute resolver configuration . . . . . . . . . . . . . . . . . . . 28
4.2. LDAP DataConnector configuration . . . . . . . . . . . . . . . . . . . . . 33
A.1. Shibboleth SP extension configuration . . . . . . . . . . . . . . . . . . . . 52
A.2. AttributeQuery handler invocation . . . . . . . . . . . . . . . . . . . . . . 53
viii

Glossary
ACL Access Control List
AD Active Directory is a directory service provided by Microsoft in its
operating systems.
Ajax Asynchronous JavaScript and XML is a set of techniques used at
client-side to create asynchronous communication with web server.
API Application program interface is used for integrating new features
into existing software.
CGI Common Gateway Interface is a standard used for generating dy-
namic web content and web applications
CIDR Classless Inter-Domain Routing
CSV Comma seprated value file is used for storing delimited data
GWDG Gesellschaft für wissenschaftliche Datenverarbeitung mbH is an IT
service provider based in Goettingen for Georg-August-Universitaet
Goettingen and Max Planck Society.
Handler Shibboleth service provider handler is an extension to provide addi-
tional features to the service provider.
IdP Identity Provider in shibboleth
IT Information Technology
JAR Java Archive is a package file format used for distribution of ag-
gregated Java class files and any other dependent metadata and re-
sources.
JSch Java Secure Channel is a java implementation of SSH2 protocol which
allows to connect securely java programs with SSH based servers.
JSON JavaScript Object Notation
ix

Glossary
LDAP Lightweight Directory Access Protocol is an internet protocol used
by online application and softwares to look for information stored
in a server.
SLO Single Logout is an action to cleanly exit from multiple sessions in a
single sign-on session.
SOAP Simple Object Access Protocol is a messaging protocol which helps
in communication between heterogeneous operating systems using
HTTP and XML protocols
SP Service Provider in shibboleth
SSH Secure Shell (SSH) is a protocol for secure remote login and other
secure network services over an insecure network.
Source: https://tools.ietf.org/html/rfc4253
SVN Sub-Version is a popular version control system used for collabora-
tive project development.
UI User Interface
URI Uniform resource identiﬁer (URI) is a character string used to iden-
tify a resource’s name.
URL Universal Resource Locator
UUID Universally unique identiﬁer is a 128 bit value often displayed as
hexadecimal text with hyphen characters in middle. It is used in
software systems for uniquely identifying information.
XML Extensible Markup Language (rfc4825)
x

1. Introduction
Digital identity management is an important aspect for online services as it provides
authentication and authorization to the users, to be able to use the services provided
by organizations. With growing number of online services, users are expected to main-
tain different digital identiﬁcation information, which means different username and
password for each independent account. Majority of users use the same username and
password for login with every service they are using online[1], which makes the pro-
cess unsafe because a malicious service provider can misuse the stored password of the
user to access other services, or an identity leak will render more number of accounts
vulnerable.
Digital identity management is a very important aspect for both service provider and
the service users. Federated identity management helps both users and service providers
to minimize the risk and burden of maintaining redundant identities[2]. The core func-
tionality of a federated identity management system is to share authentication infor-
mation among federated organizations[3] and allow users from different organizations
to access the services without the need to have dedicated user accounts. Users authen-
ticate themselves with their home organization which in turn sends an authorization
token to the service provider to allow or deny the access to a service. Authorization is
an equally important aspect for federated identity management solution.
“The development of Web-based federated identity solutions has advanced more rapidly
compared to Web-based privilege management mechanisms. The result is a wide gap
between the federated identity and privilege management mechanisms and calls for
an integrated approach to provide a comprehensive access management solution.”[4].
While most of the federated identity management solutions can offer authentication
and authorization by utilizing the same back end provider like LDAP[5] or Active
Directory[6] (AD), but in many cases it could be possible to have a dedicated group
management system in the organization. For instance an organization might have the
requirement to distinguish the systems storing authentication information and autho-
rization information (especially for user groups). The modern federated identity man-
agement systems provide very less or no support to connect any existing group man-
agement system to a federated identity management system.
1

Chapter 1. Introduction
1.1. Motivation
The main goal of this research paper is to develop a solution which could be utilized
to connect different available group management systems with the federated identity
management system shibboleth[7], in order to be able to share group information. This
section contains discussion about the usage of group information sharing in the fed-
erated organization environment and the beneﬁts which could be leveraged through
such a system.
Figure 1.1 shows an example of two federated organizations connected through shib-
boleth. As shibboleth is mostly used by educational institutes let’s assume organization
A in ﬁg 1.1 is the home university of user A. User A is enrolled in many courses at his
home university. User A is further interested to enrol in a course offered by organiza-
tion B. Normally to get the access to the course at organization B user A will be asked
to make a request for a new account at organization B.
Figure 1.1.: Federated identity management example
Here comes shibboleth into play by replacing the need to create new user account for
user A. Instead, organization A and organization B sign a contract for existence of a
federation between them. With the federation in place the login page of organization B
is simultaneously updated with a link to support login via organization A. The entire
users from organization A including user A can access this link which redirect them
to their home organization for authentication. After successful authentication organi-
zation A sends a valid authentication token with additional attributes understandable
by organization B’s web service. In the next step organization B creates an account for
user A by processing the authentication token and attributes returned by organization
A. This step helps the service provider identify returning users and also stores their
preferences and other information related to the service offered. In this example sce-
nario, user A can now ask the course coordinator at organization B to give him access
2

to the course as his account is active in organization B. Due to federation in place user
A is able to access the service and the course within minutes which might have taken a
couple of days to be addressed earlier.
Organization A also contains group information of the users (fig 1.1). Let’s assume user
A is the administrator of a course group in his university and this information is stored
in the group information storage system at organization A. Further, user A has been
privileged to administer a course in organization B. User A wants to provide access
of this new course to the group members of user A from organization A. To achieve
this scenario, user A will request all the group members to login into organization B’s
service in order to create the initial account with the help of shibboleth. In the next step
user A will search and add individual users to the new course group in organization
B. This process of group creation could be very tedious if there are 100’s of users who
belong to a group and given the fact that the group already exist in another federated
organization within the federation. Moreover, user A has to wait until a user creates his
account at organization B with the help of shibboleth to add the new user to the course
group.
Figure 1.2.: Federated Identity Management with group information sharing
The main goal of this research paper is to identify a solution for sharing group in-
formation using shibboleth as the mediator and many different group management
systems as the source of group information. Also, the goal is to identify the key use
cases in which a group information sharing might be useful specially in a federation
of universities, scientific institutes and IT service providers like Gesellschaft für wis-
senschaftliche Datenverarbeitung mbH Göttingen (GWDG). As described in fig 1.2 the
motivation of this research paper is to connect the heterogeneous group management
systems between different federated organizations which will create a link between the
systems to allow exchange of group information. Users will be able to import already
existing groups from their home organization to minimize redundant group creation
3

and provide new users belonging to the group with instant access to shared services.
Further, the solution will help in real time synchronization of current group members,
providing or revoking proper access rights to a group user. There will be discussion on
the best practices for single log out procedure in shibboleth’s environment to safely end
multiple signed in sessions by the user created by accessing the shared service.
1.2. Overview
The following chapters will discuss about the path towards identiﬁcation of key re-
quirements for development and implementation of the solution for group information
sharing. Chapter 2 contains information related to shibboleth, its overall architecture
and information about plug-in development for shibboleth. Also, there will be discus-
sion about some group management systems as use cases which will act as a source of
group information.
Chapter 3 will concentrate on the requirements for devising the solution, the architec-
ture of the solution and what steps will be required to achieve the ﬁnal solution. It
will also provide the advantage of a group management system in a federated environ-
ment. Further, I will describe the use cases for testing the developed solution. Chapter
4 will describe the detailed implementation process, general steps for connecting any
group management system with shibboleth, group attribute sharing, group informa-
tion synchronization and group information storage at service provider’s unit. Chapter
5 provides some test results on the performance of the designed modules. Chapter 6
describes the time frame and steps towards the organizational deployment of the de-
signed solution. Chapter 7 contains information about similar projects and solutions.
Finally, chapter 8 contains the conclusion of the thesis work.
4

2. Fundamentals
2.1. Shibboleth
Shibboleth is an open-source implementation developed by internet21 which provides
authentication and authorization services to federated organizations using Security As-
sertion Markup Language[8] (SAML) protocol. This section is about the architecture
of shibboleth. The two main components of shibboleth involved in web based single
sign-on and attribute exchange are the identity provider (IdP) and the service provider
(SP). Both components comprise of many different core sub-components as described
in ﬁg 2.1 and further explained in following sub sections. The architecture explained
in following sub-sections has been derived from the ofﬁcial shibboleth architecture
documentation[9].
Figure 2.1.: Shibboleth Architecture
1http://www.internet2.edu/
5

Chapter 2. Fundamentals
2.1.1. Shibboleth Service Provider
The shibboleth service provider (SP) resides at the organization hosting the service.
Service provider’s primary task is to protect unauthorized access to a service, speciﬁ-
cally an URL and its contents in the web server. Service provider resides in the same
server as the HTTP server (preferably Apache) hosting the service. All provided tech-
nical details, discussions and development work has been performed using shibboleth
SP version 2.5.4.
Service Provider consists of the following main components:
1. Assertion Consumer Service2: It is an URL/URLs (HTTP resource) hosted by
the Apache server which acts as receptor of HTML form submissions and even-
tually redirects the user to a requested resource. The secure context is established
through new user session which is further handled by shibboleth service provider
software.
2. Session Initiator: Used for generating authentication request between a user’s
browser connecting to shibboleth service provider and extending the request to
shibboleth identity provider.
3. DS (Discovery Service): Primary task of discovery service is to facilitate auto-
matic/manual selection of shibboleth identity provider which is serving a single
shibboleth service provider.
4. SAML POST: Part of assertion consumer service that handles the received SAML
assertion from shibboleth identity provider.
5. Shib Module: It is a module build to facilitate shibboleth service provider inte-
gration with Apache or any other HTTP server.
6. Shib Daemon: Dedicated service running in the background in an operating sys-
tem to invoke shibboleth service provider related calls and services.
7. Web Resource: The web resource which resides in HTTP server and protected by
2.1.2. Shibboleth Identity Provider
The shibboleth identity provider is the entity responsible for authenticating users and
release of attributes. The authentication information and attributes are together bun-
dled in a SAML assertion which is sent to the service provider. All provided technical
2https://webauth.service.ohio-state.edu/~shibboleth/shire.html
6

details, discussions and development work has been performed using shibboleth IdP
version 2.4.3. The components involved in the authentication and authorization pro-
cess by shibboleth identity provider are:
1. Authentication Engine/Authority: It is a SAML based service responsible for
authenticating the user’s credentials and issuing an authentication assertion for
users to be able to access the respective shibboleth service provider. Authentica-
tion engine does not directly authenticate the user but uses the defined authenti-
cation handler for the given user. The authentication handler information is set
by an identity provider administrator during installation. For more information
please refer to installation guide in appendix A.
2. Attribute Authority: SAML based service used for processing attribute query re-
quests and issuing attribute assertions to service providers with mutual authenti-
cation. Further task include, filtering of attributes before release based on service
provider specific filtering policy within the federation.
3. Attribute DB: It is the database or application used in the backend for storing
attributes belonging to users. Multiple backend storage systems are already sup-
ported by shibboleth identity provider. The most popular are LDAP and Win-
dows Active Directory. Other extension based backend support could be pro-
vided for MySQL, MongoDB[10], etc. Generally speaking, any attribute provid-
ing system can be ported into shibboleth identity provider provided that a java
connector API exists for the specific application.
4. Authentication DB: It is the database used for storage of authentication informa-
tion of users. Shibboleth identity provider provides active support for authentica-
tion via LDAP server for username/password based authentication. Third party
extensions are required to provide authentication support for MySQL. Some other
third party authentication extension could be obtained from contribution page of
shibboleth.
5. Single Sign-On (SSO) service/profile: It is an HTTP resource contained within
shibboleth identity provider that receives and processes authentication requests
sent from user’s browser via shibboleth service provider. Authentication process
is initiated by SSO service, further, it helps in issuing proper HTTP responses to
user’s browser after interacting with authentication authority.
6. Username Password Authentication: It is one of the many authentication han-
dler types available in shibboleth which displays the authentication form after
redirecting to identity provider from service provider, takes the user’s credentials
for authentication and sends the data to the authentication authority via HTTP
POST. Other available authentication handlers are IP address based authentica-
tion, Remote authentication, etc. Any other authentication handler other then
Username Password authentication is out of scope of this thesis paper. Interested
7

people can refer more at https://wiki.shibboleth.net/confluence/display/
SHIB2/IdPUserAuthn.
2.1.3. WAYF
“Where Are You From” is an optional service in shibboleth which sits in between
shibboleth service provider and identity provider. Its task is to discover correct iden-
tity provider for incoming requests from service provider. Mostly, shibboleth service
provider contains an in-built WAYF service. In our case WAYF is already implemented
in shibboleth service provider and a more extended version of it is available as shib-
boleth discovery service. Going deeper into this topic is not required as it is not rele-
vant for this thesis work. Interested people can read more about it at https://wiki.
shibboleth.net/confluence/display/SHIB2/DiscoveryService, the wiki at ofﬁcial
page for shibboleth documentation.
2.1.4. Flow of data and control in Shibboleth
Figure 2.2.: Shibboleth control ﬂow
8

Figure 2.2 shows the complete flow of data and control between the various parts of
shibboleth as discussed in previous sections. Following is the explanation of shibboleth
control and flow where the numbers correspond to the process number in fig 2.2.
1. User accesses the service through a web browser, the service which is protected
by shibboleth service provider checks if the user is authenticated to access the
service.
2. Session initiator creates authentication request and sends it to the browser with
information about how to reach shibboleth identity provider (URL).
3. Browser sends the request to shibboleth identity provider using HTTP POST.
Identity provider checks if user is authenticated or not, SSO service will take ap-
propriate steps for selection of proper authentication handler.
4. Web browser gets redirected to username/password login handler (set by admin-
istrator). The selected authentication handler is username/password authentica-
tion handler. Appropriate web form is displayed to the user for authentication.
5. User enters authentication information (username/password) and presses sub-
mit. The browser sends HTTP POST with the authentication data to the authenti-
cation engine. After successful authentication, attribute authority proceeds with
attribute resolution and filtering. A SAML assertion is prepared containing an
authentication statement and an attribute statement containing user’s attributes.
6. The SAML assertion is sent as HTTP POST from shibboleth identity provider to
shibboleth service provider. Service provider processes the SAML assertion, re-
leases attributes after service provider side attribute filtering and redirects the
control towards the requested resource by the user at the beginning.
7. After redirect, shibboleth service provider checks again whether the user is au-
thenticated, and finds the user as authenticated this time. Finally, user is provided
access to the resource. The web page related to the service requested is displayed
in user’s browser.
2.2. Mailman
This section contains brief overview about GNU mailman as it is being used as one of
the use case for providing group information. GNU mailman[11] is a free software used
for electronically managing mailing lists. Generally it is used for managing electronic
mail discussions and e-newsletter lists. Mailman provides ease of manageability due
to its integration with web. Users can easily manage their online subscription accounts
and list administrators can manage their lists through online portal.
9

Mailman can be used for membership management, a list administrator creates a pub-
lic or private list and either adds users manually to the list or users can request for
membership into public list. These group of list act as a well managed user pool. Mail-
man also supports hierarchical list creation. The parent list comprises sub list which in
succession contains more sub lists or list members. These type of hierarchical lists are
known as Umbrella list in mailman.
Umbrella lists are quite important within organizations as well as educational institu-
tions as most of the departments form a hierarchy of sub-departments and members.
Figure 2.3.: Mailman Umbrella list
Figure 2.3 displays a simple example of mailman umbrella list. Any email sent to the
parent list “University of Göttingen” will eventually be distributed to every subsequent
list and its members. In the same way any email sent to the list “Practical Informat-
ics” will be delivered to all of its subsequent lists and its members. This hierarchy of
mailing list comprises group membership information which will be leveraged while
including mailman as an use case for group information provider with shibboleth. Cur-
rently mailman could be installed only on Unix operating system and doesn’t supports
windows operating system.
10

2.3. UserFrosting
UserFrosting3 is an opensource user management system developed by Alexander Weiss-
man4. The web based software is written in PHP5 and uses MySQL6 database for stor-
age of data. It is a freely available online for download and use, it is based on open-
source licensing, installation and download instructions are available at the following
link https://github.com/alexweissman/UserFrosting/. UserFrosting will support the
concept of group information sharing from MySQL database. MySQL’s user interfaces
is designed mostly for running sql queries. Therefore, a web based or standalone soft-
ware with MySQL database in the backend is required to be used as a group informa-
tion provider. The opensource licensing of UserFrosting gives the right to modify and
customize the code as per requirement to fetch group information.
UserFrosting provides a backend Ajax[12] API for querying data from the server. Ajax
API allows a pathway to contact web server for retrieving user data. A simple HTTP
GET or HTTP POST request is initiated from client’s system at the backend to contact a
predeﬁned location at the web server. This predeﬁned location is served by a PHP script
which checks the parameters received via HTTP request and processes them to send a
respond to the client. The data contained in the response is shown to the user. This
method will help shibboleth IdP to retrieve data from UserFrosting’s MySQL database.
2.4. LDAP
Figure 2.4.: LDAP directory tree
3http://www.userfrosting.com/http://www.userfrosting.com/
4http://alexanderweissman.com
5http://php.net/
6https://www.mysql.com/
11

LDAP is a popular directory service protocol used for information sharing in intranet
and internet applications. LDAP could be used for sharing users, systems, services
and applications information. Data inside LDAP is arranged in a hierarchical format as
shown in the example in ﬁg 2.4.
From the root node, the directory is separated among different internet domains. In-
dividual domains can contain different organizations as child element in the hierarchy.
Organizational nodes can have different organizational unit like people, servers, etc.
as child nodes. Organizational units contains individual node for individual entity, for
instance people organizational unit can have persons. Each person possesses all the
relevant attributes like, uid for identiﬁcation, name, address, etc. A classic example
is a hierarchy of email list in an organization. Identity validation through LDAP is
the default process used in shibboleth IdP to authenticate users. Moreover LDAP can
provide additional attributes related to users like name, email etc. to shibboleth service
provider. LDAP can be used for creating group of users as well. This group information
can be shared in the form of attributes with shibboleth service provider.
2.5. Active Directory
Active directory is a directory service developed solely for Windows operating sys-
tem. Active directory stores the data in a similar fashion as it is stored in LDAP. There
exists a hierarchical data structure storing domain, organization, organizational units
and users. Users information is accompanied by different attributes related to users
information. It is also possible to store group information in active directory like email
hierarchy. Users can create new groups in active directory and add or remove mem-
bers to the group. This group information and membership related information could
be later shared with shibboleth service provider. Figure 2.5 contains a typical example
from LDAP showing a domain forest.
Figure 2.5.: Active directory domain forest
12

A domain forest is a hierarchy of domain names. In the above example gwdg.de is the
parent domain and other domains follow with added preﬁx to parent domain. Let’s
take an example of a session cookie. A cookie made for *.gwdg.de will give a user access
to all the sub-domains. If a user is given cookie for domain email.gwdg.de then access is
given to only sub-domains of email.gwdg.de. This is an example of authorization. Re-
placing the domains with groups will help in group based authorization. This autho-
rization information could be shared with shibboleth service provider for authorization
of services.
13

3. Heterogeneous group management
This chapter covers the importance of heterogeneous group management system in an
organization. Further, it covers the requirement analysis towards planning and devel-
opment of such a system.
3.1. Heterogeneous group sharing and advantages
3.1.1. Group sharing
Federated identity management and privilege management are supported by most of
the modern federated management system. Shibboleth can use LDAP server in the
backend for providing authentication and authorization information simultaneously.
The main task of shibboleth is to perform authentication while the authorization in-
formation is shared with the service provider in the form of attributes[13]. Service
provider is responsible for further authorization of services based on the attributes re-
ceived. As stated earlier, most organizations have multiple systems which could con-
tain valuable group related information, but these systems are not directly connected
to some already supported backend data provider in shibboleth like LDAP server for
authorization purpose. When federated organizations need to share the group infor-
mation retrieved from these systems they will be forced to synchronize the group in-
formation into an existing shibboleth attribute provider service like LDAP or MySQL
server. Or else the group information provider system needs to establish a direct con-
tact with the service provider to provide data which will impose added policy making,
planning and development. Sharing the attributes through shibboleth is easier and rec-
ommended as, shibboleth already takes care of authentication, and its also possible to
release attributes in the same channel.
Ability to share group information using heterogeneous group information provider
systems is advantageous in a multifaceted organization. Before going deeper into the
advantages of heterogeneous group management system lets discuss the advantages
of group sharing in a federated environment. At the beginning of this thesis report
there has been discussion on the possibility to provide access to a group of users from
external organization. This task involves importing group information from external
federated organization for providing access instead of re-creating the group of users.
14

Chapter 3. Heterogeneous group management
The biggest benefit leveraged by importing group information is mitigation of redun-
dant group creation. This is quite identical to federated identity management where
redundant user account creation is minimized. The advantage of importing group in-
formation instead of recreation can be easily realized in a situation where the group
contains hundreds of members. Additionally, users are authorized to existing groups
immediately after there first login based on the shared group information attributes
coming from their home organization.
Figure 3.1.: Benefits of sharing group information
Importing group information in a federated environment adds benefit for the adminis-
trators as they are not burdened to create individual groups and add members into it.
Time is saved in group creation and users receive access in a more efficient and faster
way. Additional information could be carried with the general group information for
more varied services. Group information could carry attributes related to preference
of users within the group. Considering the use case for sharing group information for
courses, roles of users within the group could be shared, for instance if a user is an
teaching assistant, a tutor or a student, thus organization B (fig 3.1) will have a course
with users from organization A and having similar roles for user A (tutor) and user B
(student).
There is an obvious benefit of importing group information in scenarios where frequent
group creation takes place and the groups being created are either derived or are du-
plicate of existing groups. If a group tag is attached to the group information attribute
then a search facility could be availed to search similar or like minded groups. For
instance a group of physicists interested in astronomy could have the following tags,
“physics”, ”astrology”, ”extra terrestrial objects”, etc. Searching on this tags can fetch
groups and there working.
15

3.1.2. Heterogeneous group sharing
Figure 3.2.: Heterogeneous group management system
Heterogeneous group management systems inherits the beneﬁts of group sharing in
federated environment as discussed above. Additionally, heterogeneous group man-
agement systems minimize the boundary of including any available group provider in
the market, for using it to provide valuable information regarding group members and
additional attributes to support the group functionality. Users are not bounded to use
only supported (by shibboleth) group management system like LDAP. There is possi-
bility to leverage group information from enterprise applications like human resource
management application or membership management application and more.
Without the presence of a heterogeneous group sharing framework users will be forced
to synchronize group information to the only available group management application
supported by the federated identity management solution. In case of shibboleth it could
be LDAP or Active Directory.
As shown in ﬁg 3.2, organization A consists of 3 separate group information storage
system and all can contribute individually the group information for a user. Later part
of this research paper shows that it’s possible to fetch information from some of the
group management systems using shibboleth’s attribute manager in identity provider.
Also group information from different systems could be either merged together and
sent as single identity, or it could be sent as individual attribute for each group man-
agement system.
16

3.1.3. Usage at GWDG
Heterogeneous group management system could help institutions to import groups
while using services provided by GWDG1. GWDG is the IT service provider for Georg-
August-Universität Göttingen and the Max Planck Society in Germany. Following are
some use cases where group creation through importing might be required to share a
service among a group of users:
1. File sharing: It is quite common to share files with a group of users for academic
work or organizational work. Online storage and file sharing mediums allow you
to create groups of users collaborating on a single folder. Users belonging to such
a group could be imported from the home organization of the users.
2. Virtual machine sharing: Group of users working on a project might require to
use the same Unix system. Or they might want to share the available comput-
ing power, memory and storage area. In such a situation group access could be
provided by importing the group related information to create automatic Unix
accounts for the group members.
3. Access to managed services: GWDG provides many managed services to the
users like Wiki[14], Git[15] or SVN[16]. These services could involve collabo-
rative work among a group of users. Access to the managed services could be
authorized using group information imported from shibboleth identity provider.
A read access to these services is usually provided to all users (except for private
service groups). Administrative rights could be delegated using imported group
information from identity provider.
4. Access to paid software: GWDG hosts many paid software for which special
access permission is required. A group of users requiring access to a software
like Microsoft Office 3652 could be delegated using imported group information
to external users within federated organizations.
3.2. Requirements Analysis
This section contains discussions regarding the requirements analysis for designing the
heterogeneous group management system. The steps are part of IBM’s suggested pat-
tern for requirements analysis[17].
1Gesellschaft für wissenschaftliche Datenverarbeitung mbH Göttingen : http://www.gwdg.de/
2https://products.office.com/en-us/business/explore-office-365-for-business
17

Figure 3.3.: Functional Requirement Analysis
Figure 3.3 shows a chart of the functional requirement analysis for the development of
the synchronization of heterogeneous group management system in shibboleth. The
core functionalities of the requirement analysis have been discussed in following sec-
tions.
3.2.1. Business processes
The major requirement of the business process is to allow sharing and synchronization
of group information from heterogeneous group management system in a federated en-
vironment using shibboleth. It should provide the ability to connect shibboleth identity
provider with different heterogeneous group information storage systems being used
by an organization. The connection will allow sharing group information and related
attributes with service providers from multiple group management systems to avoid
redundant group creation and for extra services based on group related attributes. The
solution will allow users to self authorize other users in the form of groups to access
dedicated services.
18

3.2.2. IT processes supporting business processes
The core IT processes supporting the business process are:
1. User management: User management in the context of this thesis work is re-
stricted to managing existing users within a group and doesn’t involve new user
creation or deletion.
2. Access management: Access management is a core functionality as the groups
are being shared for authorization of dedicated services. Only members who are
part of a predeﬁned group will get access to the service. Access management will
be performed by group administrator and supported by additional modules.
3.2.3. Activities within each IT process
The IT processes comprises the following activities:
1. Group related activity: The primary activity of user management process is to
provide group management related activities. This activities could belong to ei-
ther shibboleth identity provider or shibboleth service provider. All the functions
belonging to this activity will support group administration and group sharing in
shibboleth.
2. Access regulation activity: Access regulation activity supports the access man-
agement IT process. The core task of this activity lies in authorization mechanisms
and periodically update group members information for real time authorization.
3.2.4. Functions within each activity
Group related activity can be subdivided for identity provider and service provider.
The core functions of identity provider are create, update or delete a group, view mem-
bers of a group and change the privacy setting of the group. Changing the privacy
setting will control sharing of group information outside the organization.
Further group related activity in identity provider has functions to add or delete users
from a group. This task is mostly handled by the group administrator and this task also
acts as a building block of authorizing services.
In service provider, group related activities requires functions to import or delete exist-
ing group and also allow administrator to view which members of the group are active
in service provider. The active members information is stored by service provider dur-
ing the ﬁrst login of the user to access the service.
19

The third functionality in the group related activity is the aggregation of group infor-
mation data. Multiple systems will connect to shibboleth identity provider, the group
aggregator function pulls relevant group information related data and attributes and
packs them for sharing via shibboleth identity provider.
Access regulation activity requires two functions. One of the core function will grant or
revoke access to a service based on the membership information of a user to a prede-
ﬁned set of group deﬁned by group administrator. Another function will synchronize
the group information for logged in users to support real time authorization of services.
3.2.5. Use cases for the functions
Use cases surrounds the two prominent users of the system i.e. group administrator
and group members. Figure 3.4 shows the functions discussed above and the support-
ing users who can access these functions.
Figure 3.4.: Use case for functions
20

As shown in fig 3.4, group administrator has access to all the functions except the sup-
porting functions. At service provider, group administrator can import a group, delete
an existing group or view the active users of the group who have made the first login
with the service provider. Before first login, shibboleth service provider is completely
unaware of the existence of a user. Users data is generally stored in service provider for
saving user specific preferences and to identify returning users to the service. A group
user might be allowed to view the members of the group.
An organization hosting identity provider, group administrator can add or remove
users from existing groups, create, modify or delete groups and view existing mem-
bers of the group. A group user might be allowed to view the members of the group.
Supporting functions is the group of functions which are not directly controlled by the
group admin or group members but it provides aid in the functioning of the overall
system group management system. Group data aggregator works at the organization
hosting identity provider. Grant or revoke access as well as function to sync user’s
group information are part of organization providing the service to users.
21

4. Implementation
In this chapter I have discussed he planning and development involved in building
individual modules of the architecture of group information sharing system.
4.1. Core elements and architecture
For allowing shibboleth to share group information in an federated environment, some
additional modules are required. These modules support various tasks of group infor-
mation sharing as well as group member’s authorization.
Figure 4.1.: Modules used in group sharing framework
Figure 4.1 shows the shibboleth architecture improvised with the additional modules
(highlighted in bold) to support group sharing in federated environment using different
types of heterogeneous group information storage systems. The ﬁrst task is to prepare
22

Chapter 4. Implementation
the group management systems and the user interface accessible by group administra-
tors and group members. User interface will allow the group administrator to create,
update, delete, hide or view a group as well as add or remove new members from the
group. Additionally user interface will allow group members and group administrator
to view the current members in the group.
Next requirement is to develop APIs for connecting shibboleth IdP with individual
group information provider application. This needs development of individual API
for every group management system due to their heterogeneous nature. The task of the
API is to read the parameters and requirements from shibboleth IdP and query the same
from group information provider and return the results to shibboleth IdP. Shibboleth
IdP forwards the group information as attributes to the SP. Details for development has
been discussed in upcoming sections.
The above steps will complete the process of group information sharing in the identity
provider’s organization. The next step is to setup the modules at the service provider’s
organization. The user manager module in figure 4.1 is not part of the development in
this thesis work, but normally user manager exists in the service provider’s organiza-
tion to store the attributes and preferences of logged in user. User manager stores the
data in a dedicated database associated with the service for storage of user information.
Even for services allowing anonymous logins, existence of a user manager is possible
which stores the IP address of the user for the duration of the session to save prefer-
ences. Its even possible to store the anonymous users information longer for auditing
purpose.
Group manager module stores and processes group related data received from identity
provider. A database is associated with the group manager to store group related data.
Group manager will additionally provide group administrator with the interface to
import or delete groups and will also allow administrators to check active members
of the group. To check active members information, support of both user manager
module and group manager module will be required. User manager acts as the supplier
of active members information which is matched against group information by group
manager to find out active members.
Single Logout (SLO) module supports the access revocation of a user from active ser-
vices. Its main purpose is to remove all existing session of the user as he might have
logged in into multiple services through single sign-on. During the user’s first login,
sessions are created at both IdP and SP in shibboleth as well as a session might be cre-
ated in the service itself. Afterwards, additional session could be created when user
accesses additional services through single sign-on. The core purpose of single sign-on
is to allow users to login once and access multiple independent services without the
need to login again. Logout from services is an important aspect from the point of view
of access revocation. Therefore, services residing on sub-domains of a single parent
domain can achieve SLO by destroying the cookie to its parent domain. SSO sessions
23

could exist within the same domain or the services could reside on different unrelated
domains, in this case its not possible to destroy all the active session by just destroy-
ing the session for one service. A more sophisticated logout mechanism is required to
achieve true single logout in systems supporting SSO. Shibboleth’s in-built SLO mech-
anism is not very efficient1 and requires custom development to support true SLO from
all the services.
The last module is the attribute synchronizer. The core task of this module is to period-
ically query updated attributes from IdP for a specific user. Updated attributes reflects
changes in group membership for a user. This will help in access grant or access revo-
cation of a service for the user. Both SLO and attribute synchronizer will be invoked by
group manager when required. Group manager will periodically invoke attribute syn-
chronizer for updated attributes and will invoke SLO when a user has been removed
from a group and his active sessions needs to be removed. The SLO will logout the user
from all the existing sessions of a service for which access have been given due to the
users membership in the respective group linked with authorization of a service.
Following sections describes the implementation method for individual module. Addi-
tionally it contains information on various available solution for each module and valid
arguments on the selected solution.
4.2. Group information providers UI
Group information management system requires a graphical user interface to support
group managers with a easy to use mechanism to maintain their groups. In this section,
I have discussed the graphical user interface support for the four selected group infor-
mation provider application I am going to consider, namely Mailman, UserFrosting,
LDAP and Active Directory.
4.2.1. Mailman
Mailman provides a built-in user interface in the form of a web portal accessible with
the help of CGI scripts. The user interface is ready to use for new group creation
in the form of mailing lists. Each mailing list contains an administrator and many
users. Users identification is there email address. The portal is accessible in the path
http://[hostname]/mailman/admin/[list name]/. The portal allows users with sufficient priv-
ileges to create new mailing list. This option is accessible from the admin panel in the
path http://[hostname]/mailman/admin/. New members can be added or removed from
1https://wiki.shibboleth.net/confluence/display/CONCEPT/SLOIssues
24

the membership management panel in the administration page. Membership manage-
ment displays the current members in the list and also provides panels to bulk insert
many users simultaneously or select and delete multiple users at a time. The member-
ship panel is accessible at http://[hostname]/mailman/admin/[list nam]/member/. Users can
control the privacy of the mailing list through the control panel. First open the particu-
lar mailing list’s administration panel and go to privacy options, under privacy option
there is “subscription rules” with option to advertise the mailing list. Advertised mail-
ing lists are returned in result set when a user requests all the mailing list in the server
through shell command. This feature of advertising mailing list could be used later
during API development for the purpose of hiding the list from sharing among other
federated organizations.
Mailman supports all the functions listed in the use case for identity provider’s organi-
zation in fig 3.4. All the functions are supported in the default functionality of mailman
and there doesn’t exists any requirement to modify or extend any feature of mailman.
4.2.2. UserFrosting
UserFrosting supports membership management and group management. The user in-
terface supports user creation, user deletion, group creation and group deletion. Group
membership is provided to all new users being added to the system. The current user
interface lacks vivid functionalities to manage group information and membership ef-
ficiently, but it could be easily extended and customized as per requirement due to the
opensource licensing of the software. In conclusion, the user interface partially sup-
ports group management as it lacks the ability to remove users from groups using the
user interface. But the software supports the sole purpose of group management sys-
tem backed by a database system (MySQL), which could be used as group information
provider among federated organizations.
The user interface of UserFrosting will not be modified in this thesis work because for
testing perspective it is sufficient to only receive group attributes from the system.
4.2.3. Light Weight Directory Access Protocol
Light Weight Directory Access Protocol (LDAP) is a popular open source directory ac-
cess protocol which is often used for user management as well as authentication. Infact
shibboleth by default authenticates users through LDAP or Active Directory. There
are many LDAP servers freely available in the market for installation. OpenLDAP2
is the preferred LDAP server for installation as it supports memberof attribute, I will
later show the requirement for this attribute in group sharing in federated environment.
2http://www.openldap.org/
25

LDAP installation doesn’t comes with a pre-installed user interface. But there are many
freely available and highly capable user interfaces that can interact with LDAP and
helps in performing administrative work on LDAP server. Apache directory studio3
and phpldapadmin4 are known and stable for using as an user interface with LDAP
server.
After installing an user interface its very easy to control and administer user groups
in LDAP server. The user interface supports all the functions in the use case for iden-
tity provider’s organization in fig 3.4. Group administrators can create new groups or
delete existing ones and add or remove members from the group. The memberof at-
tribute is useful for returning the list of groups in which the user is a member, without
the need to search and loop through every group. Although this attribute has the draw-
back that it doesn’t returns results from nested group structure, in that case recursive
search is an better option to find group membership. The other requirement of the
search filter is to return the owner of the group in response to the query. The managedby
attribute in LDAP could be leveraged here to find out the prospective members with
administrative privileges. The owner will be given the privilege to import the group in
4.2.4. Active Directory
Active directory (AD) is a windows based directory service, ready to use for authentica-
tion and authorization services in windows operating system. AD can be used for user
management as well as group management. Active directory’s user interface could be
used for supporting the functions in the use case for identity provider’s organization
in fig 3.4. Active Directory Administrative Center5 is the supported active directory
GUI for Windows based systems. The user interface allows group creation, deletion,
modification as well as provides the ability to view current group members. New users
could be added by modifying the members tab in properties for a group. The memberof
attribute in active directory is also beneficial as it eliminates the need to loop through
every group of the active directory system to search membership information for a spe-
cific user. Though the memberof attribute doesn’t reports nested groups in the result. A
carefully developed query is required which can search all the relevant groups in which
the user is a member. Search filters might be helpful in hiding groups from search re-
sults for private groups that are restricted for sharing in a federated environment. The
search filter should also be able to return the owner or user with administrative power
for the group. There exists the managedby attribute in AD groups which could be help-
ful in finding the list group administrators for the consecutive group. This member will
be given the right to import the group at shibboleth service provider.
3http://directory.apache.org/studio/
4http://phpldapadmin.sourceforge.net/wiki/index.php/Main_Page
5https://technet.microsoft.com/en-us/library/dd560651(v=ws.10).aspx
26

4.3. API development
This section explains the process of extension development to build API’s for individual
group management systems and query group information.
4.3.1. Basics
API development for shibboleth IdP is the process of developing extensions or plug-ins
in java which can be embedded with shibboleth IdP. The documentation given at the
shibboleth wiki page provide highly technical information on extension development
and contains adequate detail for an experinced developer[18]. This section contains
general overview of extension development for shibboleth IdP, the expected knowledge
of a developer and the steps needed for planning an extension development.
There are many types of possible extensions which could be developed for shibboleth
IdP. The most interesting for this research topic is development of a data connector[19].
Shibboleth wiki strongly recommends the need to have good understanding of Spring
framework6 as well as the knowledge of XML7 and XML schema8 to proceed with de-
velopment of extensions. Additionally a thorough understanding of java web develop-
ment is expected from the reader as well as the developer.
Extension development for shibboleth IdP can be divided into three parts as described
in tab 4.1.
Part Description
Operation Lists the set of actions the extension can perform on the
connected application.
Input attributes List of attributes necessary for the extension to work properly.
Output attributes Predeﬁned or dynamic output variables could be set in the
extension.
Table 4.1.: Identity provider extension parts
Operations are a set of actions the extension can perform on the connected application.
For instance, a database related extension can establish connection to the database, exe-
cute some queries to search data and close the active connection to the database. Mostly,
data connector extensions are used for data retrieval from an existing database. So it
6http://www.springframework.org/
7http://www.w3.org/XML/
8http://www.w3.org/XML/Schema
27

will include connection establishment and connection removal from the external appli-
cation. Further, it will allow to run a custom search query provided by shibboleth IdP
administrator or developer. The custom search query will return a response with some
result.
Input set of attributes prepares the extension with all the required values useful while
a users data is retrieved by attribute authority. This attributes could be connection pa-
rameters like host, username and password for establishing connection. Further a pre-
defined search query could be provided by administrators which runs during attribute
authority call, users identification attribute could be supplied additionally to the search
query.
Output variables are either predefined within the extension or dynamically defined by
administrators. Data type of the output attribute is defined during configuration along
with column names for identification.
<resolver:DataConnector id=" DBRef"
xsi:type =" RelationalDatabase "
xmlns ="urn:mace:shibboleth :2.0: resolver:dc">
<ApplicationManagedConnection
jdbcDriver ="com.mysql.jdbc.Driver"
jdbcURL =" jdbc:mysql :// localhost :3306/ shibdb"
jdbcUserName =" shibadmin"
jdbcPassword =" abc123 "/>
<QueryTemplate >
<![ CDATA[
SELECT groupname FROM user_groups WHERE
username = ’$requestContext .principalName ’
AND groupname <> ’role ’
]]>
</QueryTemplate >
<Column columnName =" groupname" attributeID =" groups "/>
</resolver:DataConnector >
Code Snippet 4.1: MySQL attribute resolver configuration
.
Code snippet 4.1 shows configuration of MySQL with input and output attributes and
search query. Input attributes provides the connection parameters (jdbcdriver, jdbcURL,
jdbcusername, jdbcpassword) and output attribute (<column columnName=”” attributeID
=”” />) defines the column name mapping between the attribute and search queries re-
sponse. The QueryTemplate takes a string which is a select query, username for the user
currently being authenticated is provide by the $requestContext.principalName. All the
28

operations takes place using the MySQL JDBC9 connector plug-in which creates the
bridge between shibboleth IdP and MySQL server.
Once all the three requirements are prepared, extension development can be initiated.
Following is a flow of steps required for developing the extension based on shibboleth
wiki[18]:
Step 1: First step is to setup the extension template provided at the wiki page of shib-
boleth IdP extension development[18]. The imported project is Maven10 based, so any
dependencies of the extension development can be imported through adding depen-
dencies in pom.xml metadata file. Final result of the extension development is a JAR
file. Shibboleth IdP needs to be built/rebuilt with this JAR file placed in lib directory to
be able to use the extension.
Step 2: Plug-in’s schema definition describes the properties needed by the extension.
Properties corresponds to input and output attributes. Schema file allows the IdP to
perform validation of input attributes based on constraint like mandatory field. The
schema definition is an XML schema file, it contains many different elements describing
the properties of the attributes like data type, requirement, etc. The schema definition
file is used to declare the input and output arguments supported by the extension. The
schema definition file is placed at “project_directory/main/resources/schema” location from
project directories perspective.
Step 3: In this step I create the Spring (framework) configuration files i.e., spring.schemas
and spring.handlers. Both files are placed in the path “project_directory/main/resources/
META-INF”. The spring.schemas file informs Spring the location of the schema file based
on a particular namespace. Below shown is an example of a schema file which contains
the XML namespace URI followed by the location of the schema file from projects per-
spective.
urn: example.org: shibboleth :2.0: resolver =
schema/myConnectors.xsd
The spring.handlers file tells Spring the Namespacehandler used for identifying a particu-
lar namespace. A typical example is shown below with XML namespace URI followed
by fully qualified class name for the given Namespacehandler.
urn: example.org: shibboleth :2.0: resolver =
com.example.shibboleth. MyDataConnectorNamespaceHandler
9http://dev.mysql.com/downloads/connector/j/
10https://maven.apache.org/
29

Backslash () are used as escape character due to special meaning of colon (:) character
in property files.
Step 4: In this step I create a Spring BeanDefinitionParser and a BeanFactory method. The
work of this methods are, to parse XML configuration for the extension and set the re-
quired bean properties and/or constructor arguments for the new java class developed
for the extension in step 4. In simple terms this class will get the list of attributes re-
quired by the extension, gets the corresponding values of the attributes and sets them
for use by the extension’s main java class file from step 5. The SCHEMA_TYPE param-
eter is set in this class which provides the QName11 for the extension type.
Step 5: Create a new java class file by extending BaseDataConnector class. Return state-
ment consists the list of attributes which will act as output attribute’s value. All the
required input arguments are passed as properties and accessed through constructor
or by means of a setter method. The set of operations the extension is expected to per-
form are written in this new java class file. Supplementary java class files could exist
supporting the functionalities of the new java class file. The default function resolve is
used to return the output attributes received after completing operations. The attributes
are returned in Map<String, BaseAttribute> format to attribute handler.
Step 6: Here I implement a NamespaceHandler class by extending BaseSpringNamespace-
Handler. Its work is to instruct Spring to invoke BeanDefinitionParser for the given ex-
tension using an init() method. init() method calls the function registerBeanDefinition-
Parser() with two arguments. First argument is SCHEMA_TYPE field defined in Bean-
DefinitionParser at Step 4 and the second argument is an instance of BeanDefinitionParser.
Step 7: Installation of the extension is straightforward. The JAR file produced after
extension development is placed in lib folder of shibboleth IdP source folder and af-
terwards the IdP is rebuilt/installed. After rebuilt/installation is completed the new
namespace is added in atribute-resolver.xml for using the extension. The namespace
and XML schema file location are added to the root element as shown below:
attribute -resolver.xml File
<resolver: AttributeResolver
...
...
xmlns:uid="urn:example.org:shibboleth :2.0: resolver"
xsi:schemaLocation ="
...
...
urn:example.org:shibboleth :2.0: resolver
classpath :/ schema/myConnectors.xsd">
11http://docs.oracle.com/javase/7/docs/api/javax/xml/namespace/QName.html
30

Next requirement is to add a data connector whose work is to establish connection to
the external application being used for data extraction. DataConnector element takes the
input and output arguments additionally. Finally an AttributeDefinition element which
defines the new argument going to be shared with shibboleth SP. DataConnector acts as
a dependency to the AttributeDefinition element.
4.3.2. Mailman API
The first step towards API development for mailman is to identify the input and output
attributes as well as the operation supported by the API. Group information from mail-
man will be extracted using Unix shell[20] commands, as mailman already provides
built-in commands to extract membership information as well as group administrators
information. Special measures needs to be taken for distinguishing groups which could
be shared among all federated organizations and those which are hidden from external
organizations. First way of hiding the groups has been mentioned in section 4.2.1 where
privacy options from mailing list administration panel could be used to mark whether
the group should be advertised or not during a group listing. The API will work by
establishing a secure channel from shibboleth IdP to mailman server and execute shell
commands to retrieve the data. Following are the steps required to get group informa-
tion from mailman server for an user. This steps designates the operations performed
by the API extension.
Step 1: First step is to establish a secure connection to mailman server. JSch12 is a java
based library which is used to establish SSH13 connection to a Unix server. To estab-
lish SSH connection using JSch following parameters are required: hostname, username,
password and port number. All this parameters will be provided as input attributes in
<resolver:DataConnector> element in attribute-resolver.xml file. hostname (IP address of
mailman server), username, password and port number are used to establish SSH connec-
tion with mailman server. Thus an utmost requirement is to setup a SSH server at
mailman server to accept SSH connection.
Step 2: Once the ssh connection is established the first shell command is executed to
fetch the list of mailing lists from mailman after filtering out private mailing lists. The
command list_lists -ab is executed to get the required mailing list names. Here list_lists
by default displays all lists with description. The additional parameters -a and -b dis-
plays only advertised lists (hides private list) and removes mailing list description re-
spectively.
Step 3: The list of mailing list obtained from mailman is further used for querying the
membership information of a particular member. Member’s username or user id is
12http://www.jcraft.com/jsch/
13https://tools.ietf.org/html/rfc4253
31

provided in the input attributes with other attributes as mentioned in step 1. The com-
mand find_member -w -l [groupA] -l [groupB] ... -l [groupZ] [username] is executed to find
out if the username exists in the given list names and whether the user id is an admin
in any of the given lists. The main task of find_member command is to identify a user’s
membership in a list of mailing lists. -w argument is used for identifying the group ad-
ministrator and -l argument suffixed with list name(s) is used to cinduct search only on
given list name(s). The final argument of the command is the username which needs to
be searched for membership. The final output of the query is like following:
find_member -w -l mailman -l gwdg idp@localhost.de
Output:
mailman (as owner)
gwdg
The output shows result of search on lists mailman and gwdg. The user idp@localhost.de
is present in both the lists and the user is the owner or admin of the mailman mailing
list.
Step 4: In the next step the connection from the SSH server is closed and the output
of the Unix command is prepared to be sent to shibboleth IdP in the form of out-
put attribute column name defined in <resolver:DataConnector> element in attribute-
resolver.xml file.
There are two other ways to control the privacy of a group name or mailing list in
mailman. The first alternative way uses the built-in access control in Unix[21] using
which access to a folder or file could be restricted. For hiding a group, the read access
to the group’s location in Unix server could be restricted for the user initiating the
search query. This user is provided in the input attribute as username. General path
to group’s location is /var/lib/mailman/lists/[group’s name]. To make this solution work,
mailman’s admin panel needs to be modified to provide a new option for group privacy
management. Changing the privacy option will run a shell script in the background
to modify read permission of the folder/file for the respective group. So, when the
user executes find_member command read access to the private groups is denied and
by default the search for membership is made only on the public mailing lists. This
solution involves running only one shell command find_member -w [username] to get the
desired output.
The second alternative way uses an additional parameter of the find_member command.
The argument -x excludes the mailing lists from membership search criteria. So, the
final command is find_member -w -x [groupA] -x [groupB] ... -x [groupZ] [username]. All
the list names provided with argument -x will be hidden from membership search. This
method requires shibboleth IdP administrator to provide the exclusion list in the input
attribute list in <resolver:DataConnector> element in attribute-resolver.xml file.
32

The above steps combined with extension development steps will give the mailman
API extension.
4.3.3. UserFrosting
The API for UserFrosting needs the following input attributes: hostname and user_id.
hostname is the URL used for accessing UserFrosting’s index page. user_id is an unique
id generated for each user at the time of new user addition. The extension operates by
accessing user’s group information by calling the backend PHP script stored in User-
Frosting web server for serving Ajax calls. The request is sent in the form of HTTP GET
request to the server and the URL for the request is of the following format:
http ://[ UserFrosting URL]/api/load_users.php?
user_id =[user ’s id]
The output is received in JSON format from the server which contains users informa-
tion in addition to group information. By default UserFrosting releases very few group
information which is generally not sufficient in my case. Therefore, some modifications
where made in the backend PHP scripts to release relevant group information stored
in the MySQL database. The final output with group information is returned in the
output attribute.
The above steps combined with extension development steps will output the UserFrost-
ing API extension.
4.3.4. LDAP
LDAP API is pre-existing in shibboleth IdP. Therefore, I only show the usage of the
LDAP to retrieve group information. The basic setup for <resolver:DataConnector> ele-
ment is provided in code snippet 4.2.
<resolver:DataConnector id=" myLDAP"
xsi:type ="dc:LDAPDirectory"
xmlns ="urn:mace:shibboleth :2.0: resolver:dc"
ldapURL =" ldap :// localhost :10389"
baseDN ="ou=User ,ou=ShibUsers ,ou=system"
principal ="uid=admin ,ou=system"
principalCredential =" secret">
<dc:FilterTemplate >
<![ CDATA[
33

(uid= $requestContext .principalName)
]]>
</dc:FilterTemplate >
</resolver:DataConnector >
Code Snippet 4.2: LDAP DataConnector configuration
The basic input attributes are ldapURL, baseDN, principal and principalCredential. These
are basic parameters required to connect to LDAP server. FilterTemplate is used to pro-
vide LDAP filtering query. This query helps in filtering result sets. Filter query can
fetch relevant group information for a user identified by the parameter $requestCon-
text.principalName. The most important parameter to identify group membership is
memberof parameter in LDAP filter query. The managedby parameter can be used to get
group administrators information.
4.3.5. Active Directory
Shibboleth IdP actively supports AD as an attribute provider. Therefore, API develop-
ment is not required because a carefully constructed filter query can release relevant
group information from AD. Configuration for <resolver:DataConnector> element to re-
lease attributes from AD is similar to code snippet 4.2 provided for LDAP. The parame-
ters to get group membership information and group owner information is also similar
to LDAP, the functionality is supported by memberof and managedby parameters in AD
filter query.
4.4. Single logout
“Web single sign-on (SSO) systems enable users to authenticate themselves to multiple
online services with one authentication credential and mechanism offered by an iden-
tity provider”[22]. According to Suoranta[22] et al. most of the SSO systems gives more
emphasis on SSO and only a few systems gives equal emphasis to the Single Logout phe-
nomenon as well. An SLO is under development for the latest version of shibboleth IdP
(version 3). According to their official documentation single logout is planned as an op-
tional feature in a future release, currently it is not supported due to the lack of space
in cookies for tracking the sessions[23]. When a user logout of a service, the usual ex-
pectation is all other active sessions of the user are destroyed as well which might have
came into existence due to single sign-on[24]. Similar feature is expected by a group ad-
ministrator when the access rights of an user is revoked. That means, the user should
be logged out of all the active sessions which has been a result of SSO and due to the
users group membership.
34

In general three types of sessions are created when user is provided access to a service
in a shibboleth based federated environment. The first session is created at shibbo-
leth IdP after the user authenticates, post that shibboleth SP creates a session based
on the SAML assertion received from shibboleth IdP as well as the IDP session cookie
and finally a session might be created at service level. A full fledged SLO mechanism
should be able to destroy all the three sessions as well as the session created with other
service provider by utilizing SSO. Currently logout is invoked using /Logout handler
configured in shibboleth SP. Initial tests revealed that using the /Logout handler is able
to delete the IdP session and the SP session of the logout initiator. Additionally the /Lo-
gout handler also tries to remove the session for all the other sessions created by means
of SSO but it fails in removing other sessions. Please note the above result belongs
to shibboleth IdP version 2.4.4. With shibboleth IdP version 2.4.3 /Logout is only able
to delete the session of IdP and may or may not delete the SP session from the user’s
browser.
Different possible solutions for SLO has been discussed by Suoranta et al.[22] and some
effective solution has been proposed by them. Two solutions can help in implementing
SLO in our case. First solution uses polling mechanism to check for active IdP sessions.
This solution works for sessions created between multiple SP’s but authenticated by
same IdP. The polling mechanism works in SP by periodically sending a message to IdP
to check if its session still exists, in which case it renews the session within SP and of the
service. If it fails to receive positive response from IdP then the session at SP and the
service is destroyed. A negative response from IdP might come in two scenarios, first
if IdP’s session has expired and second, if user has already initiated a logout request
from another SP’s service. The drawback of this mechanism is added network load
generated by the polling mechanism which periodically sends and receive messages
from IdP.
Second solution which could be implemented is through session tracking by IdP. In this
method IdP stores information of all the SSO sessions for every user. In case of SLO IdP
could initiate a SAML logout on each SP based on collected session information. This
method has its own drawbacks though, first it breaks users privacy as IdP is tracking
each user’s service access. Moreover, IdP initiated logout is more efficient in SOAP
based logout requests, this generates added burden for each service provider to support
SOAP based logout requests. Moreover SOAP based logout work in the background,
therefore it doesn’t provides any information to the user regarding the reason of the
logout.
A third solution has been developed by Hungarian NIIF institute as an extension for
shibboleth IdP. The NIIF SLO extension[25] can initiate logout from IdP as well as from
SP. Moreover, it supports SOAP based logout as well as URL based logout. SOAP based
logout works silently whereas URL based logout actually shows the logout process to
the user in the form of redirection of pages and also displays proper messaging on
success or error. NIIF SLO extension also returns proper response in JSON format after
35

the logout is successful. The extension works in both systems supporting JavaScript or
those not supporting JavaScript. The extension might fail in some situations due to lack
of support for session cache by shibboleth IdP, though very less information is provided
about the failure situations. Logout could fail definitely during IdP session timeout.
A fourth solution is based on the new feature of shibboleth IdP version 2.4.4. The fea-
ture provides the user with information about all the SSO sessions the user has initi-
ated during his service access. When the user executes logout using /Logout handle
of shibboleth, the user is logged out of the current session and IdP. Logout response
contains the list of all the services the user is logged in into and an attempt is made to
remove those sessions. Though in reality the session is not removed. As a solution, a
background script could execute logout on individual sessions based on the response
received from first logout request. This logouts could run on a hidden iframe and exe-
cute /Logout on every service. In this way all the users sessions will be deleted and true
SLO could be achieved.
The only possible way to attain complete SLO is through a customized logout mecha-
nism. A future solution by shibboleth developers for IdP version 3 might be a perma-
nent solution.
4.5. Attribute synchronizer
Attribute synchronizer will work in the background specifically at service provider’s
server and periodically query IdP for updated attributes. The main motive to use the
module is to receive updated group related attributes. Any change in the group infor-
mation will help in access management for a given user. A logged in user will be forced
to quit the service in case the updated attributes doesn’t contains the group informa-
tion required for using the respective service. Similarly, a user might be given access to
a service based on updated group information returned by the attribute synchronizer
post login.
The attribute synchronizer uses an extension developed for shibboleth service provider
which can be invoked to get attributes for an user. This extension has been developed
by Academic Access Management Federation in Japan[26]. The extension uses the built-in
function of shibboleth to request attributes if it was not delivered during the first login.
This was the default method to get attributes until Shibboleth IdP version 1. A SOAP
call is placed in the backend directing to IdP to release the attributes. The installation
details for the plug-in is available in appendix A.1.
For using the extension an URL is formed which accesses the extension and passes on
additional parameters. Following is the list of acceptable runtime parameters14:
14https://bitbucket.org/PEOFIAMP/shibsp-plugin-attributequery-handler#rst-header-runtime-
parameters
36

• protocol : a protocolSupportEnumeration value to be used for querying IdP, or more
simply it designates the SAML message encoding format which is acceptable to
IdP. If omitted, urn:oasis:names:tc:SAML:2.0:protocol is used.
You can also use the following aliases.
o SAML2.0 alias of urn:oasis:names:tc:SAML:2.0:protocol
• entityID: entityID of an IdP, as mentioned in the metadata of IdP. It is the URL
where IdP is reachable.
• format: a SAML name identifier format. If omitted, the following value is used
urn:oasis:names:tc:SAML:2.0:nameid-format:persistent
• nameId: a SAML name identifier value, precisely the unique key to identify the
user.
Mandatory parameters are entityID and nameId, all other parameters are optional and
could be used as per requirement. A typical attribute request looks like following:
https ://X.X.X.X/Shibboleth.sso/ AttributeQuery?entityID=https
://X.X.X.X/idp/shibboleth&nameId= user1@localhost .de
The first part of the attribute query format contains the shibboleth service provider’s
path with an additional invoke to the handler AttributeQuery, followed by the entityID
of shibboleth IdP and the nameId for each attributes are being queried. The response
comes in JSON[27] format as shown below:
"sn" : "One",
"cn" : "User One"
The response will contain all the information attributes for the requested user released
from IdP. The attribute query is limited to users whose IP address has been provided in
the access control list in the handler invocation in shibboleth SP’s configuration file. In
general the access should be provided only to the service provider’s server. Presently
its not possible to release the attributes after some kind of authentication, that means if
all the parameters are correct then IdP will response with the attributes even if the user
has never log-in to access any service via the service provider. Attributes are released
even if there is no active session of the user with the service provider. This is a drawback
of the extension from user’s privacy perspective.
37

4.6. Group manager
Group manager is a PHP script assigned with the task to handle incoming group in-
formation and authorizing services to users. I have developed the group manager in
PHP but there is no restriction on the chosen language for implementation. This section
describes the main design concepts and criteria of a group manager plug-in, which is
hosted by the service provider’s organization. Group manager doesn’t integrates with
the shibboleth SP but works as a separate entity in collaboration with shibboleth SP.
Following are the use cases being handled by the group manager plug-in:
Group management: Group management is the process of adding or removing groups
to facilitate shared service authorization for group members. Figure 4.2 shows the ﬂow
of processes involved in group management. After a user logins he/she can try to
access the panel for group management. If the user is not an admin then he/she can
only see the list of group names already imported and the user is a member of.
Figure 4.2.: Flow chart for group management
38

If the user is an admin then access is granted to the group management panel. This
panel could be used to add or delete groups into the database stores by the service. The
added groups will allow members to access a shared service. The panel displays a list
of groups available for addition and a list of groups already added. The lists already
added could be deleted. Only those group names can be added for which the user is an
administrator. The group name is accompanied by an additional attribute which iden-
tifies the group administrator. This parameter is used to displaying the group names
in import list. Further attributes could be added into group information to identify the
source of the group name, e.g. mailman, LDAP, etc.
Service provider receives group information on a per user basis and not on a per group
basis. This means the group information will come as a binding for respective users. My
solution doesn’t shares individual groups and the list of its members. Group sharing
on a per user basis is faster as whenever group members are updated within a group,
there is no need to send the updated member list to shibboleth SP. Moreover, its not
mandatory to store the group names the user is a member of. We need to store only
those group names which have been imported and access can be granted if there is a
match between users group information attribute and stored group names. The access
management has been discussed in next use case.
Its very much possible that the imported group names are not unique among inter or
intra organization level. Therefore, some mechanism needs to be in place to distin-
guish the group names. Several mechanisms are possible to resolve this issue, I have
listed two possibilities to resolve the issue, one at shibboleth IdP and the second at
the service. The first method involves modification of attribute definition at IdP in
attribute-resolver.xml file. Additional unique parameter is prefixed or suffixed to the
group name using a ResolverScopedAttributeDefinition15 element. A scoped attribute def-
inition creates an attribute from the values of another attribute definition or data con-
nector combined with a specified scope value. The prefix or suffix added to the group
information could be a UUID. The second method works at service where it prefix/suf-
fix the group name with the domain name of the user. This method expects the group
names to be unique within an organization.
Access management: Access management handles the access grant/revoke for shared
services. Figure 4.3 shows the flow of processes involved in access management. Once
the user logins he/she can proceed to access a shared service available to the members
of a specific group. When the user tries to access the service the group manager checks
if the available group list in the local database has a match with the user’s group in-
formation attribute provided by shibboleth IdP. If there is a match user can proceed to
access the service.
15https://wiki.shibboleth.net/confluence/display/SHIB2/ResolverScopedAttributeDefinition
39

Figure 4.3.: Flow chart for access management
If there is no match between local group list stored in database and user’s group infor-
mation attribute then the user is denied access to the service and he/she is prompted
to contact the group admin responsible for the respective service’s authorization. At
the same time a time-based PHP scheduler is running in the background which checks
for updated attributes from shibboleth IdP using the attribute synchronizer plug-in for
respective user. The time period for the scheduler can be set by an administrator and it
could vary for individual services. A normal service will have a longer cycle (24 hours)
while a more important service will have a shorter cycle (1 hour) between attributes are
synchronized.
After the receiving updated attributes the group manager again checks for matching
group names between local group list stored in database and user’s group information
attribute. If there is a match then a user will be provided access to the service or an
existing user’s access will be continued. If the access is denied then SLO mechanism of
shibboleth designed in section 4.4 is initiated to revoke access from any service autho-
rized by membership to a group.
Finally, a user can logout voluntarily or a session might terminate at service/SP/IdP
level which will logout the user from the service and force him to login again to access
40

either the group management panel or the shared service controlled by group member-
ship. Local storage of group information could be realized using MySQL database.
41

5. Testing group sharing
This section contains information related to performance testing of the group sharing
architecture. Precisely, it contains information related to response time for two mod-
ules. For login test, I have used the IdPUnsolicitedSSO1 mechanism. This is an IdP
originated login process, which means there is no redirection from SP to IdP, instead a
dedicated URL belonging to IdP is called to initiate login process. Below I have shown
the URL pattern for unsolicited login process:
https ://{ IdP Host }/idp/profile/SAML2/Unsolicited/SSO?
providerId ={SP ’s entity ID}
IdPUnsolicitedSSO call takes the providerId as input parameter. providerId is the name
of the service provider entitled to work with shibboleth IdP. The test file is executed on
Apache JMeter2 and is available at shibboleth IdP wiki for version 33. The test takes
username/password from a list of 1000 users stored in a CSV file and runs 100 parallel
threads looping with a delay of 500 milliseconds, trying to login into shibboleth IdP.
A test is considered successful after receiving the SAMLResponse from IdP. Login test
is sub-divided into two tests. The first part of the test logins without any group man-
agement plug-in and the second part of the test logins with mailman and UserFrosting
plug-in returning group information. Delay in response time can be seen at the graphs
below in fig 5.1. The complete test duration is of 60 seconds.
1https://wiki.shibboleth.net/confluence/display/SHIB2/IdPUnsolicitedSSO
2http://jmeter.apache.org/
3https://wiki.shibboleth.net/confluence/display/IDP30/Load+Testing+Contributed+Results
42

Chapter 5. Testing group sharing
Figure 5.1.: Shibboleth IdP login response time
The output shows a increase in response time due to the additional querying of group
data by IdP to two different group management systems.
Second test result is related to the standalone response time of attribute synchronizer
extension as shown in ﬁg 5.2. Response time has been measured for 100 concurrent
threads trying to access the attribute synchronizer extension to get updated attributes
with a delay of 5 seconds between each call. The test duration is of 60 seconds. There is
a constant delay of 500 milliseconds between for an individual thread. The test has two
parts, one for attribute query without any group API’s and one test for attribute query
with group API’s in shibboleth IdP. The results shows expected increase in response
time due to increased operations to fetch group information at IdP.
43

Chapter 5. Testing group sharing
Figure 5.2.: Shibboleth SP attribute synchronizer response time
Due to lack of resources the tests where performed using a single IP address. That
means all the requests have originated from same IP address. Both the test scripts
could be found at the Appendix B.
44

6. Preparations for production system
This section contains step by step instructions to implement the group sharing system
in a production environment. Majority of time is spent in planning. Initial expectation
is to have the required group management systems and shibboleth installed in the or-
ganization. The installation and planning could be divided between IdP’s organization
and SP’s organization. All the necessary installation guide has been provided in the
Appendix A to aid a developer.
The IdP only requires API development. Foremost requirement is to plan the input and
output arguments as well as the operation the API must be able to perform. Afterwards
API development could be initiated. Each API development including planning could
vary from 1- 4 weeks depending on the complexity of the operations it could perform.
Rest of the modules are required in the SP’s organization. Three solutions has been
provided for implementing SLO. The best one needs to be selected, customized and de-
veloped in accordance to organizational needs. Using SLO could be avoided in case all
the services are under a single parent domain, in which case session cookie will handle
access revocation. The SLO implementation will require 2-4 weeks for implementation
and testing. The attribute synchronizer extension requires re-installation of shibboleth
SP. Post installation configuration are straight forwards and requires very less time and
effort. The JSON based response from the extension is supplied to the next and final
module of SP, the group manager. The attribute synchronizer extension will require 1
week of time but could extend for another week in case of any issues faced during shib-
boleth SP source build installation. Implementing group manager will vary between
different service providers as not every organization will handle the attributes in a sim-
ilar fashion. Planning and implementation of group manager will require 3-6 weeks
depending on the complexity of the tasks to be handled by group manager.
Approximately 3 months time should be sufficient to bring the system into production.
The time from development to production will also depend on the experience level of
the user with development for shibboleth. Some organizations might already have a
SLO in place working in production. Such an organization would be able to save a
couple of weeks in bringing the system to production.
45

7. Similar work
Grouper1 is a tool offered by Internet22, the organization behind the development of
shibboleth. Core functionality of grouper is to facilitate collaboration within an orga-
nization by allowing users to create groups themselves and administrate them. This
group information is stored within Grouper and can be used to authorize multiple sys-
tems like email list, calendar group, web site, and so on. An application needs to have
an API connecting with grouper to fetch the relevant group information for user au-
thorization. Grouper behaves as a centralized group information storage system. Any
modiﬁcation in group information will be propagated to all other dependent systems.
This functionality is different from my work in which each system acts as a standalone
group information provider. Figure 7.1 shows a system based on grouper and shibbo-
leth to provide federated identity management with group sharing. Every application
including shibboleth is dependent on grouper to retrieve group information.
Figure 7.1.: Grouper based federated group sharing
Using grouper delegates group administration to individual group managers, thus re-
1http://www.internet2.edu/products-services/trust-identity-middleware/grouper/
2http://www.internet2.edu/
46

Chapter 7. Similar work
lieves workload from system administrators. Moreover, automatic role management
is addressed in multiple services due to the centrally available group information pro-
vided by grouper. Grouper can be integrated with shibboleth as well. An API could
be designed to retrieve group information from grouper. It is also possible to directly
access group information from LDAP.
Similarity between Grouper and my solution is that, both allows local users to manage
groups and share this information with service providers for authorization purposes.
Difference between Grouper and my work is that, Grouper acts as a centralized group
information provider whereas my solution assembles multiple group management sys-
tems to provide the required data.
47

8. Conclusion
In this master thesis I have proposed a solution using which self created group infor-
mation could be shared among federated organizations for service authorization. Most
of the services control authorization themselves based on parameters set by system
administrators. A solution to minimize the burden on system administrators by al-
lowing users to create group and authorize services based on group membership has
been proposed. The architecture for sharing group information using shibboleth has
been provided with detailed analysis of core modules and requirement analysis. The
core modules like API designing, SLO, group attribute synchronizer and group man-
ager has been discussed thoroughly in order to facilitate designing such a system for
real production environment. Testing has been done to identify the performance of the
group attribute synchronizer as well as to identify the delay in login due to additional
querying of group information by shibboleth during authentication. Requirement anal-
ysis of the system has been provided to ease planning before initiating development.
Further work could be done to migrate the system to support shibboleth IdP version
3.x. Currently due to lack of sufﬁcient documentation for shibboleth IdP version 3.x,
the development and testing work has been restricted to shibboleth IdP version 2.x.
48

References
[1] S. Riley, “Password security: What users know and what they actually do,” Usabil-
ity News, vol. 8, no. 1, pp. 2833–2836, 2006.
[2] D. W. Chadwick, “Federated identity management,” in Foundations of security anal-
ysis and design V. Springer, 2009, pp. 96–120.
[3] R. L. Wexelblat and N. Srinivasan, “Planning for information technology in a fed-
erated organization,” Information & management, vol. 35, no. 5, pp. 265–282, 1999.
[4] R. Bhatti, E. Bertino, and A. Ghafoor, “An integrated approach to federated
identity and privilege management in open systems,” Commun. ACM, vol. 50,
no. 2, pp. 81–87, Feb. 2007. [Online]. Available: http://doi.acm.org/10.1145/
1216016.1216025
[5] M. Wahl, T. Howes, and S. Kille, “Lightweight directory access protocol (v3),”
1997.
[6] A. G. Lowe-Norris and R. Denn, Windows 2000 active directory. O’Reilly & Asso-
ciates, Inc., 2000.
[7] R. Morgan, S. Cantor, S. Carmody, W. Hoehn, and K. Klingenstein, “Federated
security: The shibboleth approach.” Educause Quarterly, vol. 27, no. 4, pp. 12–17,
2004.
[8] P. Hallam-Baker, “Security assertions markup language,” May, vol. 14, pp. 1–24,
2001.
[9] S. Cantor and T. SCAVO, “Shibboleth architecture,” Protocols and Proﬁles, vol. 10,
2005.
[10] I. MongoDB, “Mongodb,” 2012. [Online]. Available: https://www.mongodb.
com/white-papers
[11] J. Viega, B. Warsaw, and K. Manheimer, “Mailman: The gnu mailing list manager.”
in LISA, 1998, pp. 309–316.
[12] J. J. Garrett et al., “Ajax: A new approach to web applications,” 2005.
49

References
[13] H. Gomi, M. Hatakeyama, S. Hosono, and S. Fujita, “A delegation framework
for federated identity management,” in Proceedings of the 2005 workshop on Digital
identity management. ACM, 2005, pp. 94–103.
[14] B. Leuf and W. Cunningham, “The wiki way: collaboration and sharing on the
internet,” 2001.
[15] J. Loeliger, “Collaborating with git,” Linux Magazine, June, 2006.
[16] B. Collins-Sussman, B. Fitzpatrick, and M. Pilato, Version control with subversion. "
O’Reilly Media, Inc.", 2004.
[17] “A simple pattern for requirements analysis,” referred on 01.05.2015. [On-
line]. Available: https://www.ibm.com/developerworks/architecture/library/
ar-analpat/#The%20pattern
[18] “Creating custom idp extensions,” 2013, referred on 18.03.2015.
[Online]. Available: https://wiki.shibboleth.net/confluence/display/SHIB2/
IdPDevCustomExtension
[19] “Data connector extension,” Online, 2013, referred on 18.03.2015.
[Online]. Available: https://wiki.shibboleth.net/confluence/display/SHIB2/
IdPDevExtDataCtr
[20] K. O. Burtch, Linux shell scripting with bash. Pearson Higher Education, 2004.
[21] T. Narten and M. Burgess, “Unix operating system,” 2003.
[22] S. Suoranta, K. Manzoor, A. Tontti, J. Ruuskanen, and T. Aura, “Logout in sin-
gle sign-on systems: Problems and solutions,” Journal of Information Security and
Applications, vol. 19, no. 1, pp. 61–77, 2014.
[23] “Shibboleth idp 3, single logout,” 04 2015, referred on 17.05.2015. [Online].
Available: https://wiki.shibboleth.net/confluence/display/IDP30/Clustering#
Clustering-FeatureLimitations
[24] M. Linden and I. Vilpola, “An empirical study on the usability of logout in a single
sign-on system,” in Information Security Practice and Experience. Springer, 2005,
pp. 243–254.
[25] “Single logout in shibboleth idp,” 2010, referred on 01.04.2015. [Online]. Available:
https://wiki.aai.niif.hu/index.php?title=Single_Logout_in_Shibboleth_IdP
[26] T. Nishimura and H. Ura, “Attributequery handler plugin for shibbo-
leth sp,” May 2014. [Online]. Available: https://bitbucket.org/PEOFIAMP/
shibsp-plugin-attributequery-handler
50

References
[27] “Json,” 2015, referred on 15.05.2015. [Online]. Available: http://json.org/
[28] “Shibboleth sp installation,” referred on 12-12-2014. [Online]. Available: https:
//wiki.shibboleth.net/conﬂuence/display/SHIB2/NativeSPLinuxSourceBuild
51

A. Installation Guides
A.1. AttributeQuery handler plug-in shibboleth SP
The extension comes with precompiled *.so files1, this files could be directly placed in
shibboleth SP for using the extension. But there is chances that the extension won’t
work due to different versions of shibboleth SP and other softwares like Apache,etc.
being used by different organization. The solution here is to build shibboleth SP from
source and then build the extension by pointing to the installation path of shibboleth SP.
Building the extension requires header files from shibboleth SP and its dependencies. A
quick list of acceptable arguments to run configure on the extension can be obtained by
running the command ./configure –help inside the extensions source folder. It will enlist
a long list of arguments of which the most important ones are, –prefix for providing
designated installation directory and –with-shibsp for providing path where shibboleth
SP is installed after source build. A detailed installation step for shibboleth SP source
build in Linux can be found on the wiki page of shibboleth2.
Post installation the *.so files will be available inside shibboleth SP installation at the
path /lib/shibboleth. The next step is to configure shibboleth SP to be able to access the
extension. The code snippet A.1 shows the extension initiation in shibboleth SP. The
OutOfProcess and InProcess blocks are placed at the beginning of the SPConfig in shib-
boleth2.xml file.
<SPConfig >
<OutOfProcess >
<Extensions >
<Library path="attributequery -handler.so"
fatal="true"/>
</Extensions >
</OutOfProcess >
<InProcess >
<Extensions >
<Library path="attributequery -handler -lite.
so" fatal="true"/>
1https://bitbucket.org/PEOFIAMP/shibsp-plugin-attributequery-handler/downloads
2https://wiki.shibboleth.net/confluence/display/SHIB2/NativeSPLinuxSourceBuild
52

Appendix A. Installation Guides
</Extensions >
</InProcess >
</SPConfig >
Code Snippet A.1: Shibboleth SP extension configuration
Next step is to invoke a handler element in shibboleth SP’s configuration file shibbo-
leth2.xml. Code snippet A.2 shows a typical setup for invoking the handler element in
shibboleth configuration file.
<Handler type="AttributeQuery " Location="/AttributeQuery "
acl="127.0.0.1 ::1 x.x.x.x./32" />
Code Snippet A.2: AttributeQuery handler invocation
Typical configuration of the handler element requires the type, Location and acl at-
tributes. While the first two attributes remains the same, acl or access control list con-
tains the IP addresses having the access right to invoke the extension remotely to get
the response. The acl takes a CIDR block for allowing access to an IP address.
A.2. Shibboleth 2.x Installation
A.2.1. Basic installation
The complete Shibboleth installation requires the installation of Shibboleth identity
provider and Shibboleth service provider. Shibboleth installation encapsulates the in-
stallation process followed by configuration. The installation process generates the
IDP’s entity id, IDP’s initial metadata, the key pair with self-signed certificate and
IDP configuration files. Prior to installation the SSL certificate (used to secure IDP’s
browser) and source of SAML metadata for service provider is required. 3
A.2.1.1. System requirements
Product Version
Java 1.5 & Higher
Apache tomcat server 6.0 & Higher
Apache web server 2.2 & Higher
Table A.1.: System Requirements for Shibboleth Installation
3https://wiki.shibboleth.net/confluence/display/SHIB2/IdPInstall
53

The basic requirements of system for IDP installation are:
A.2.2. Shibboleth Identity provider Installation & Configuration
Installation begins by setting up Apache tomcat 8.0 server available in Apache tom-
cat’s official website. I am using the Apache web server installed in WAMP server (2.5).
The Shibboleth IDP is downloaded from Shibboleth website4. Once the IDP zip file is
downloaded the file is unzipped and install.bat file is exectued for windows. The in-
stallation path is set to “path/to/opt/Shibboleth-idp” and domain name is set to “127.0.0.1”
i.e. localhost.
The configuration of Apache webserver requires the addition of “ProxyPass /IDP/ ajp://
localhost:8009/IDP/” line in “HTTPd.conf” file. This is done to pass the IDP requests
from HTTP to tomcat. We are doing username/password based authentication so, the
user’s credential information on LDAP server. The Apache tomcat server configuration
requires to add “<Connector port=’8009’ enableLookups=’false’ redirectPort=’8443’ proto-
col=’AJP/1.3’ request.tomcatAuthentication=’false’ address=’127.0.0.1’ />” to ‘/usr/local/tom-
cat/conf/server.xml’ file. This is done to send usernames to IDP and the stop tomcat
server itself from authenticating the user.
The Shibboleth IDP installation generates the war file which is deployed in Apache
tomcat server once above configuration is done. To deploy war file one can either paste
it in “apache-tomcat-8.0.17webapps” and delete the pre-existing IDP folder and war
file. The other method could be to give the path of the war file to be deployed in
“apache-tomcat-8.0.17confCatalinalocalhostIDP.xml” file.
A.2.3. Shibboleth Identity provider source code installation:
The above installation was done to understand the basic working and ingredients of
Shibboleth. For development of the plug-in, the Shibboleth identity provider source
code installation is required. The Shibboleth source code is downloaded with SVN5.
The source code is the maven project6. To run the IDP in eclipse the installation of check-
style plug-in and TestNG framework is required. The latest Eclipse luna contains maven
m2E.The jdk version 1.7 is used in our project. To import the project the file menu is
opened in eclipse and the following sequence of procedure is done “import->existing
maven project->select the downloaded project->finish”. Once the project is imported it can
be build using maven clean followed by maven install. The maven install for Shibboleth
2.4.3 results in jar file in target folder.
4http://shibboleth.net/downloads/identity-provider/2.4.3/ shibboleth-identityprovider-2.4.3-bin.zip
5https://svn.Shibboleth.net/java-shib-idp2/tags/2.4.3/
6https://svn.shibboleth.net/java-shib-idp2/tags/2.4.3/
54

Post installation the following configuration files needs to be changed present at {path/-
to/IdP/installation}/conf:
File Changes
relying-party.xml Add <metadata:MetadataProvider> with IdP’s and SP’s
metadata.xml files path
login.config Add LDAP connection parameters
logging.xml Change logging level as per requirement
handler.xml Allow only <ph:LoginHandler
xsi:type="ph:UsernamePassword"> and <ph:LoginHandler
xsi:type="ph:PreviousSession"> to provide proper login page
to user
attribute-resolver.xml Add <resolver:DataConnector> and
<resolver:AttributeDefinition> to query attributes and set
them to release
attribute-filter.xml Add <afp:AttributeRule> to filter attributes before release to
SP
Table A.2.: Shibboleth IdP configuration changes
A.2.4. Problem while building in Shibboleth 2.4.3
The missing “tools.jar” file error occurred as the eclipse was pointing to JRE instead of
JDK. To overcome this JDK path was set in “window->preferences->java->Installed Jre’s”.
A.2.5. Problem with resulting jar file in Shibboleth 2.4.3
For the deployment of IDP in tomcat server war file is required. Jar file does not contain
all the dependencies and required files.
A.2.6. Analysis for solution of problem:
The jar file renaming to war file was done. The war file was not able to give the same
results as expected.
The new dynamic web project was made (file->new->other->dynamic web project->project
name->finish). The web project was converted to maven project (right click->configure-
>convert to maven project->packaging (war)->name->finish). Tried to add the entire source
file in newly created project but it was complicated due to the structural differences.
55

The creation of war file requires specific structure of maven project. We tried to recreate
the structure by swapping the files. When project were rebuilt there were numerous
built errors.
We also tried to discover the possibility of conversion of maven project into dynamic
web project and then again reconverting to maven project. This was thought because
conversion of maven project to dynamic web project will restructure the files required
for war build. After reading various documents we found that no proper procedure
was described for this technique.
After reading various documents it was concluded that pom.xml is the file that is used
for jar or build. Hence, we tried restructuring ‘pom.xml’ file. Pom.xml details were
read and structure was understood.
We tried modifying packaging tag to pom.xml file. The jar was replaced by war.
<packaging>war</packaging>
Also plug-in structure was also changed from jar to war
<plug-in> <groupId>org.Apache.maven.plug-ins</groupId>
<artifactId>maven-war-plug-in</artifactId><configuration>
The maven build was done and war file was generated. When this war file was de-
ployed in Apache tomcat server the IDP didn’t got start. When the tomcat logs were
checked “java.io.FileNotFoundException: Could not open ServletContext resource [/$IDP_
HOME$/conf/internal.xml]” error occurred. The digging of error led to information that
there is some issue with web.xml file present in web-INF folder.
When the difference between pre-installed web.xml and presently generated web.xml
was analysed using win merge it was
Pre-installed <param-value> file:/f:/opt/Shibboleth-idp//conf/internal.xml;file:/f:/opt/Shibboleth-
idp//conf/service.xml;</param-value>
present <param-value> $IDP_HOME$/conf/internal.xml; $IDP_HOME$/conf/service.xml;
</param-value>
The value was hardcoded and the war file was again deployed and it runs successfully
in outside the browser but inside eclipse it still give 404 error.
A.2.7. Service provider installation
SP comes in binary format for Windows operating system which could be installed
using windows installer. In Linux systems, SP could be installed using apt-get install
56

libapache2-mod-shib2 or it could be built from source. Following installation steps has
been copied from shibboleth official wiki available at https://wiki.shibboleth.net/
confluence/display/SHIB2/NativeSPLinuxSourceBuild.
The following packages must be built in the following order using the ./configure com-
mands listed. Unless specific version notes are given, any new release is sufficient. Be
sure to make and make install as appropriate for each package, including Shibboleth it-
self. You will also need the Boost7 headers available (but don’t need to actually build
or install the full Boost library set). Due to a bug not yet fixed in a release, only versions
up to 1.52 will work[28].
• log4shib: ./configure –disable-static –disable-doxygen –prefix=/opt/shibboleth-sp
• Xerces-C: ./configure –prefix=/opt/shibboleth-sp –disable-netaccessor-libcurl
• XML-Security-C: ./configure –without-xalan –disable-static –prefix=/opt/shibboleth-sp
• XMLTooling-C: ./configure –with-log4shib=/opt/shibboleth-sp –prefix=/opt/shibboleth-
sp -C
• OpenSAML-C: ./configure –with-log4shib=/opt/shibboleth-sp –prefix=/opt/shibboleth-
sp -C
When building Shibboleth, you can usually rely on the configure script to detect your
Apache version and do the right thing, but in special cases, you can pass various op-
tions to control the Apache module version and locate the apxs script to use.
• Shibboleth: ./configure –with-log4shib= /opt/shibboleth-sp –enable-apache-24
–with-apxs2=/usr/local/apache2/bin/apxs –prefix= /opt/shibboleth-sp
Here /opt/shibboleth-sp could be any path where shibboleth SP should be installed. Post
installation the following configuration files needs to be changed present at {path/-
to/SP/installation}/etc/shibboleth:
File Changes
shibboleth2.xml Add <Host name=""> of server, <ApplicationDefaults
entityID=""> of SP’s URL, <SSO entityID=""> of IdP’s URL
and <MetadataProvider> for IdP
attribute-map.xml Add <Attribute>coming from IdP before passing on to
service
Table A.3.: Shibboleth IdP configuration changes
7http://www.boost.org/
57

Finally generate the SP’s metadata by accessing the URL {SP’s URL}/Shibboleth.SSO
/Metadata. Share this metadata file with shibboleth IdP.
The trouble shooting guide is very handy as it lists the most frequent issues faced
by the users. The guide is available at https://wiki.shibboleth.net/confluence/
display/SHIB2/Troubleshooting.
A.3. Mailman
Mailman can be installed in on Unix based systems. This guide is for installation on
Ubuntu. First task is to install a mailserver. I have installed Postfix8. Installation com-
mand is sudo apt-get install postfix. I have installed postfix with default configuration.
My mailservers host was localhost.
Next step is to install mailman9. The command is sudo-apt get install mailman. During
installation an example Apache configuration file comes with mailman and is placed
in /etc/mailman/apache.conf. In order for Apache to use the config file it needs to be
copied to /etc/apache2/sites-enabled using the command sudo ln -s /etc/mailman/apache.conf
/etc/apache2/sites-enabled/mailman. This will setup a new Apache virtualHost for the
Mailman administration site. Once the symbolic link is created, you’ll need to enable it
and restart Apache sudo /etc/init.d/apache2 restart.
Now we can add new lists into mailman and users to them for sharing their group
information.
8http://www.postfix.org/
9https://help.ubuntu.com/community/Mailman
58

Master_Thesis_2015_by_Sanjeev_Laha_21229267

More Related Content

What's hot (18)

Similar to Master_Thesis_2015_by_Sanjeev_Laha_21229267 (20)

Master_Thesis_2015_by_Sanjeev_Laha_21229267