SHIBBOLETH WEB-PROXY FOR SINGLE SIGN-ON OF CLOUD
SERVICES
Thomas R
¨
ubsamen and Christoph Reich
Cloud Resarch Lab, Hochschule Furtwangen University, Furtwangen, Germany
Keywords:
Cloud Computing, Shibboleth, Single Sign-on, Reverse Web Proxy.
Abstract:
Single Sign-On (SSO) allows users to access services, for which they possess sufficient access rights, without
re-authentication once they are authenticated successfully. Shibboleth supports SSO of web services and
allows building federations.
In this paper a Shibboleth web proxy is described, which integrates a Shibboleth service provider to manage
authentication and extends Cloud management systems by enabling SSO of multiple cloud services. It is
shown how this approach can be used for highly dynamic Cloud environments, where services are often added
and removed. The Shibboleth web proxy implementation has been contributed to the Open Source Community
and is made available in the OpenNebula EcoSystem.
1 INTRODUCTION
Single sign-on solutions are more and more widely
deployed in today’s Internet environment. The main
reason for this is the rapidly growing number of cre-
dentials (e.g. usernames and passwords) a user has to
manage. Single sign-on allows the reduction of user
credentials while offering a uniform access mecha-
nism to web services.
Shibboleth(shi, 2011) is an open source single
sign-on system which is commonly used in academic
service infrastructures like the ”Authentication and
Authorization Infrastructure” provided by the Ger-
man Research Network (”Deutsches Forschungsnetz,
DFN-AAI”)(DFN-AAI, 2011). This infrastructure
forms a federation of research facilities like uni-
versities or institutes but also commercial and non-
commercial entities in Germany. So far there are 68
identity providers organized in the DFN-AAI, includ-
ing the identity provider of the University of Applied
Science Furtwangen. In a federation, Shibboleth-
enabled web services can be used by all members of a
federation according to the access policies of the ser-
vices, based on common agreed Shibboleth attributes.
Another important technology in today’s IT in-
frastructures is cloud computing. One of its main
characteristics is a highly dynamic service environ-
ment. The University of Applied Science Furtwan-
gen is actively developing a private cloud infrastruc-
ture, which offers services to its employees and stu-
dents (Sulistio et al., 2009) and is based on Open-
Nebula (OpenNebula, 2011). The integration of this
private cloud infrastructure and the services hosted
in this environment into the university’s user man-
agement and especially into the existing Shibboleth-
based single sign-on system is the focus of this paper.
To keep an overview of existing bindings between
services (and their service providers) and the identity
provider it is quite common to hard-code this rela-
tionship. This leads to manageability problems when
Shibboleth needs to be used in very dynamic environ-
ments where services are often added and removed.
Static configuration also means a lot of administrative
overhead. In this scenario the identity providers need
to be reconfigured every time a service (and its service
provider entity) gets added or removed. The result-
ing downtimes are unacceptable because of the major
importance of this system component for other, non-
cloud-based services. In cloud-based environments
the integration of services into a shibboleth-based au-
thentication and authorization infrastructure need to
be automatic and as dynamical as possible.
In this paper a system is described, which was
developed to allow usage of Shibboleth-based Sin-
gle Sign-On for cloud-based services in a dynamic
way. With this system it is also possible to host
web services in a private network, as defined by RFC
1918 (rfc, a) to overcome the limited availability of
public IPv4 addresses.
This paper is structured in the following way. In
89
Reich C. and Rübsamen T..
SHIBBOLETH WEB-PROXY FOR SINGLE SIGN-ON OF CLOUD SERVICES.
DOI: 10.5220/0003926000890095
In Proceedings of the 2nd International Conference on Cloud Computing and Services Science (CLOSER-2012), pages 89-95
ISBN: 978-989-8565-05-1
Copyright
c
2012 SCITEPRESS (Science and Technology Publications, Lda.)
this section the necessity and the goals of the devel-
oped system were described. In section 2 follows
a short introduction to the Shibboleth single sign-on
system and a classification of different web proxy sce-
narios which are relevant for the developed system. In
section 3 follows a description of the proposed solu-
tion with its associated advantages and disadvantages.
Section 4 describes the OpenNebula integration, fol-
lowed by section 5, which presents the results of an
internal evaluation of the performance of the Shibbo-
leth web proxy system. Related work is presented in
section 6 followed by a conclusion and possible future
development tasks in section 7.
2 SHIBBOLETH AND HTTP
PROXY
First, in this section, an overview about the Shibboleth
SSO systems (shi, 2011) is given. Second the differ-
ences between the used reverse proxy concept and the
commonly used forward proxy concept is described.
2.1 Shibboleth
The two main components in a Shibboleth infrastruc-
ture are the Identity Provider (IdP) and the Service
Provider (SP). The Identity Provider is usually con-
nected to an organization’s user management system
and takes control over user authentication. Addition-
ally the IdP provides information about users in form
of attributes which are stored in the user management
backend (e.g. a LDAP backend). These attributes
(e.g. username, role etc.) can be used to make autho-
rization decisions whether a user can access certain
resources or not. The Service Provider (SP) is the sec-
ond important component of Shibboleth. It is usually
run as a plugin on the web server hosting the resource
which needs to be protected. The SP authorizes re-
source access requests and also takes control of the
user authentication process if no authentication token
exists in the session context. These processes are usu-
ally completely independent of the resource which is
to be protected. This means a resource (e.g. a web ap-
plication) does not necessarily have to be customized
for use in combination with Shibboleth.
If multiple resources are located on one web
server it is also possible to protect those with one SP
running on that host. One important thing to con-
sider is that there always has to be a binding between
a SP and its IdPs. This can be done by adding the
SP’s metadata to an IdP’s configuration, which can
either be done manually or automatically by request-
ing current metadata information from the SP (Shib-
boleth 2 Documentation, 2011a). Another way to cre-
ate this binding is to register the SP in a federation
which propagates the metadata information to the IdP.
A Shibboleth Federation is an agreement between ser-
vice providers and IdPs wishing to access those re-
sources. All parties need to agree on a common set of
acceptable authorization attributes for their users, and
a schema to describe them.
2.2 HTTP Proxy
The classic web proxying differentiates two distinct
communication directions: Forward Proxy, a clas-
sic web cache is also called a forward proxy and is
usually located near the client. Reverse Proxy, a re-
verse proxy is near the requested resource and used
for server performance optimization. Many of today’s
HTTP daemons provide reverse proxying functional-
ity, such as the Apache HTTP Daemon which has been
used for developing our approach described in this pa-
per.
3 SHIBBOLETH WEB-PROXY
In this section the concept of a Shibboleth web proxy
and the developed prototype will be described. A re-
verse proxy in combination with a Shibboleth service
provider (SP) on the same host will be used to provide
authentication and authorization to resources hosted
in the cloud (PaaS and SaaS services). Any HTTP
requests to resources or services will be forwarded
through the proxy server. Figure 1 displays the archi-
tecture of the developed system. The client is a classic
browser session where a protected resource has been
requested. The IdP acts like in the classic scenario.
One major difference to the classic Shibboleth sce-
nario is that the web servers hosting the resource do
not act as SPs anymore. The SP functionality has been
moved to a single host, the reverse proxy, in the mid-
dle of figure 1. There is only one SP which is bound
to the home organization’s IdP. Obviously in the de-
scribed architecture the reverse proxy is a single point
of failure and could be a single point of attack also as
all traffic must be routed over it. We are aware of this,
but see solving this problem as part of future research
as it is not in the scope of this paper.
The reverse proxy is responsible for forwarding
requests to the web servers. One problem that arises
is how to identify which resource has to be contacted.
To solve this, there are a couple of different possi-
ble identification schemes which can used. These
schemes will be discussed in the following. One thing
to keep in mind is that all requests are sent to the re-
CLOSER2012-2ndInternationalConferenceonCloudComputingandServicesScience
90
verse proxy. Information about the actual resource
that is to be contacted must be encoded into the re-
quest URL in some way. There are two possible ways
to do this. First a part of the the request URL path
can be used. Second the DNS domain name part of
the request URL, especially the hostname part, can
be used. This way Apache’s virtual host functionality
can be used to identify which resource to contact.
Figure 1: Shibboleth web proxy scenario.
3.1 Resource Identification using URL
Path
Request are always sent directly to the proxy (e.g.
”http://proxy.example.com”). To identify the ac-
tual requested resource, the path portion of the re-
quest URL is used (e.g. ”http://proxy.example.org/
servicea”). The proxy analyzes the path component
and forwards requests to the according web server.
This way different path values result in requests to
different resources, and usually different web servers,
behind the proxy. Figure 2 illustrates this scenario. A
client requests access to the resource ”servicea”. The
request is sent to the reverse proxy which acts as a
Shibboleth SP. After authenticating the user and au-
thorizing the access to the resource, the request is for-
warded to web server 1 which hosts the requested re-
source. The authentication and authorization process
works as previously described. Of course it is possible
to use different authentication and authorization con-
figurations for each resource (each configured path on
the proxy) or no Shibboleth at all. If the authenti-
cation or authorization process fails, requests are not
forwarded to the web server and access is denied.
This architecture allows relatively easy integration
of a single SP into a classic reverse proxy. However
there are some major downsides to this approach. Re-
quest URLs can quickly become overly complex. An-
other downside is that SSL cannot be used effectively.
In fact there can be no end-to-end encryption using
SSL. It is possible to use SSL between the client and
the reverse proxy and it is also possible to use SSL
between the reverse proxy and the web server hosting
the resources, but the proxy server will always (tem-
porarily) terminate the SSL connection and therefore
see unencrypted data.
Figure 2: Proxy scenario with service identification using
paths.
3.1.1 Extension: HTTP-redirect to HTTPS
As previously stated relying on paths for resource
identification can become very complex very quickly
and lead to unhandy URLs. A simple extension to
this approach can reduce complexity by using HTTP
redirects and virtual hosting capabilities on the re-
verse proxy. A name-based virtual host can be cre-
ated on the proxy for every resource hosted behind
it. Clients can now access URLs of a name-based
virtual host. Those virtual hosts are combined with
DNS alias records pointing to the proxy server. The
proxy can now process the virtual hostname and cre-
ate a request to the more complex path-based identi-
fication scheme and redirect the client to this URL.
This method can also be used to force all requests to
use HTTPS. The resulting less complex URLs makes
it more easy for the user to handle them. Neverthe-
less this approach does not solve the SSL end-to-end
encryption problem.
3.2 Service Identification using Virtual
Hosts
Another way to identify resources behind the reverse
proxy is the use of virtual hosting capabilities of the
HTTP daemon (in our case Apache HTTPD). Virtual
hosting was initially developed to allow more web
services to be hosted on a single web server using
multiple DNS aliases. There are two different types
of virtual hosting:
SHIBBOLETHWEB-PROXYFORSINGLESIGN-ONOFCLOUDSERVICES
91
IP-based Virtual Hosting. When IP-based vir-
tual hosting is used the web server listens for
incoming connections on multiple IP addresses.
These addresses can be aliases on the same net-
work interface or addresses on different inter-
faces. For each address the web server is now able
to host different virtual hosts (e.g. different web
sites). To the client this seems like there are differ-
ent web servers hosting each of those sites when
there is in fact only one web server hosting them.
Name-based Virtual Hosting. The second type
is name-based virtual hosting. The web server
listens on only one IP address. Using the fully-
qualified DNS name it is also possible to host mul-
tiple sites on one web server. For this mechanism
to work all DNS names point to one web server.
This can be done using DNS aliases (CNAME re-
source records). When a resource is requested
the web server now analyzes the HTTP request
header. According to the hostname which is found
the request is forwarded to a different virtual host.
Name-based virtual hosting is the more commonly
used variant, when it is required to separate resources
logically and hosting them on the same web server
because it is usually easier to manage DNS aliases
and there is no need for additional IP addresses.
A system which uses a Shibboleth proxy and vir-
tual hosting for resource identification has to do ac-
cess authentication and authorization on the basis of
virtual hosts. Figure 3 displays this scenario. In the
first step a client has to resolve a DNS alias (CNAME)
like ”http://servicea.example.com”. The result will be
the IP address of the proxy server. Now the client will
send the resource request to the proxy server which
extracts the hostname of the target resource from the
request header. According to this hostname a virtual
host is chosen and the request is forwarded to the right
web server. During this process Shibboleth authenti-
cation and authorization is performed.
At first this seems like the more elegant solution to
the problem of resource identification. But there are
also some major downsides to this approach. For in-
stance the usage of SSL or TLS between the client
and the proxy is very complicated. The main rea-
son for this is in the nature of the SSL/TLS proto-
col. The identification of the virtual host to which
the HTTP request has to be sent is included in the
HTTP request header. However, this information is
not available on the proxy right away. In fact the re-
quest has to be decrypted first. The target of the re-
quest is a virtual host (e.g. ”servicea.example.com”)
but the SSL connection was established to the proxy
(e.g. ”proxy.example.com”). This results in a host-
name mismatch error. The most simple solution to
this would be using a wildcard certificate (e.g. for
*.example.com) on the proxy, but this is a generally
discouraged approach. A better solution would be us-
ing a TLS protocol extension which tries to solve this
problem.
Figure 3: Proxy scenario with service identification using
virtual hosts.
3.2.1 Extension: Server Name Indication
Server Name Indication (SNI) is an extension for the
Transport Layer Security (TLS) (rfc, b) protocol. It
allows the client to send the server name of the request
URL during establishment of the the encrypted con-
nection. The server name is usually the fully qualified
DNS hostname of the requested resource. This way it
is possible for the web server to choose the right cer-
tificate (e.g. the certificate matching the hostname of
the virtual host).
A major problem in current browser implementa-
tions of TLS is the support for SNI. Despite the wide
support of SNI on all major platforms (Microsoft
Windows, Linux, Mac OS) and Browsers in their cur-
rent releases it gets problematic if not the most cur-
rent browser versions are used. For best compatibility
it is still best to not rely on SNI support (TLS SNI
Test Site: *.sni.velox.ch, 2011). The most problem-
atic combination would be Windows XP, which is still
widely used, and Internet Explorer.
4 IMPLEMENTATION OF
SHIBBOLETH WEB-PROXY IN
THE CLOUDIA PROJECT
The University of Applied Science Furtwangen is ac-
tively researching Cloud Computing technology. The
Cloud Research Lab’s project ”Cloud Infrastructure
CLOSER2012-2ndInternationalConferenceonCloudComputingandServicesScience
92
Figure 4: Cloud management system of CloudIA.
and Applications” (CloudIA) (Cloud Research Lab -
Department of Computer Science, 2011) has devel-
oped a private cloud infrastructure based on Open-
Nebula (OpenNebula, 2011) and cloud based services
to enhance this platform.
CloudIA leverages various virtualization tech-
nologies, such as Xen and KVM, and supports
Service-Level Agreements (SLAs) in IaaS and SaaS
models, as shown in Figure 4. In this Figure, the
Cloud Management System (CMS) of CloudIA is di-
vided into several layers (User Interface, Business,
System, Resource Interface) as well as two compo-
nents (Monitoring & Management, Security) for ex-
tensibility and maintainability.
Services in this cloud system are hosted on dy-
namically created virtual machines. That means the
fluctuation of VMs (creation and deletion) is very
high. As previously stated it is one of the major goals
to integrate the cloud infrastructure with the user man-
agement system of the university. Therefore it was
necessary to find a solution which integrates services
hosted in the cloud with our Shibboleth infrastructure
while maintaining the dynamic character of cloud en-
vironments. As a result of the research conducted by
the Cloud Research Team, a Shibboleth web proxy
was developed as shown in Figure 1. The proxy itself
is running on a virtual machine hosted in the cloud.
Because of the large number of VMs which can
be hosted in the cloud it became a requirement to be
able to use private IP addresses. By passing all traffic
through the publicly available reverse proxy it is very
simple to run services in a private subnet while still
being able to access them publicly through the proxy.
Resource identification in our private cloud is
done using the previously described path-based iden-
tification mechanism. The main reason for this was
the better support for SSL without relying on SNI.
Although, it would also be possible to use the vir-
tual host mechanism in our architecture. In fact, both
the path- and virtual host based approaches are imple-
mented in our private cloud system.
Users who run services in the cloud can activate
Shibboleth for protecting those services via a web-
based management console. It is also possible to con-
figure authorization rules using this console. This
concept of a Shibboleth web proxy has been imple-
mented and provided as Open Source under the name
StudiCloud (Cloud Research Lab HFU, ) as an exten-
tion of Opennebula (OpenNebula, 2011).
5 EVALUATION
To demonstrate whether the reverse proxy based ap-
proach to shibboleth authentication is feasible in prac-
tice or not, an evaluation was performed on the HFU
Cloud Research infrastructure.
The evaluation scenario is composed of one re-
verse proxy and one target web server which hosts
a simple HTML page approximately 10KB in size.
Both servers are using the Apache 2.2 Web Server
on Debian 5.0 and are hosted as virtual machines on
the HFU cloud. To eliminate any interference be-
tween virtual machines, these two have been deployed
on separate cloud hosts, interconnected with GBit-
Ethernet.
Three different evaluation scenarios for accessing
the target web server via the proxy have been devel-
oped:
Shibboleth Authentication. In this scenario a shib-
boleth authentication workflow using the reverse
proxy is performed before access to the target web
server is granted.
Basic Authentication. Apache supports a simple
way of authenticating users. This mechanism is
used to compare the complexity against the shib-
boleth scenario.
No Authentication. The direct access scenario
without any authentication needed is used as a ref-
erence to compare to.
JMeter is a well-known and well established tool
for stress testing and performance testing of web sites
and web applications. It provides enough flexibility
and customizability options to perform benchmarks
on the above mentioned scenarios.
During the evaluation, test plans for shibboleth,
basic and no authentication workflows have been de-
veloped using JMeter. The primary focus of the evalu-
ation is to show whether shibboleth in a reverse proxy
scenario performs well enough to be used in high
load, high traffic setups, as using it as a primary gate-
way in PaaS and SaaS scenarios would be. To elim-
inate any interferences, like load fluctuations on the
cloud hosts, KVM process scheduling, these separate
test plans have been merged into a single test plan and
run in parallel. For each of the scenarios a setup of
10 threads requesting the same page sequentially 500
SHIBBOLETHWEB-PROXYFORSINGLESIGN-ONOFCLOUDSERVICES
93
Table 1: JMeter evaluation.
Average request time (ms)
No authentication 51
Basic authentication 52
Shibboleth authentication 54
times, which results in 5000 requests per scenario and
15000 requests overall, has been implemented.
Table 1 shows average request/response times of
the tested scenarios. Obviously and expectedly using
a reverse proxy which is forwarding requests to other
web servers without any authentication performs best.
Also using basic authentication which does no more
than sending additional login credentials in the HTTP
request header is only insignificantly slower. In the
shibboleth scenario a slightly worse result could have
been expected because of the complexity of shibbo-
leth authentication. For instance there is additional
communication between the apache mod shib mod-
ule and the service provider daemon shibd, which
holds session information. But obviously this over-
head does not pose much of a negative impact on the
overall performance of this setup.
These results make the aforementioned integra-
tion of Shibboleth authentication into cloud comput-
ing infrastructures using a reverse proxy, an attractive
way to maintain manageability and security.
Things to consider when integrating a Shibboleth
reverse proxy are proxy resources, bandwidth and
failover. Scaling up is the most common way to en-
hance the performance of a reverse proxy, when de-
mands are rising. Scaling out (e.g. distributing load
on multiple reverse proxies) can be a bit problem-
atic because Shibboleth SPs store session informa-
tion, which in this case need to be distributed among
multiple reverse proxies. Shibboleth supports some
mechanisms to cluster SPs. One way is to choose a
master SP which stores session information and let
other SPs communicate via TCP to it in case they
need additional information. Another way would be
to store session data in a central database and connect
SPs to it. Either way those approaches add a lot of
complexity and it needs to be seen, whether or not
they are feasible performancewise. Another approach
is to only protect the web application’s login script
and let the application itself handle authentication and
authorization in its session management. This way
the SP is only needed for initial authentication and
attribute exchange at the very beginning of a session
with the web application. Adding session stickiness
to this timeframe solves a lot of problems, but requires
the application to have session management and a lo-
gin script has to be developed, which handles session
creation. (see (Shibboleth 2 Documentation, 2011b)
for further information on SP clustering)
The approach described in this paper suffers from
one major problem. Requiring all requests to be han-
dled by the reverse proxy makes it a single point of
failure. This can be solved by applying common fail-
over strategies. In fact deploying the proxy as a vir-
tual machine has the advantage of hot migratability of
the VM in case the hosting cloud node fails. Further
research has to be conducted in this area.
6 RELATED WORK
Guanxi (Jie et al., 2008) is an open source implemen-
tation of the Shibboleth protocol and architecture for
e-social science. Guanxi Shibboleth is integrated into
the Sakai collaborative and learning environment and
PERMIS technology is enabling a policy-driven, role-
based, fine-grained access control. Our approach es-
pecially enables the OpenNebula Cloud Management
System with Shibboleth but enhances the access to
Cloud services by policies.
The national grid service (NGS) provides access
to compute and data resources for UK academics. The
work from Xiao Dong Wang et.al. (Wang et al., 2009)
describes an architecture by which users are authen-
ticated by the UK access management federation to
acquire low assurance credentials to access Grid re-
sources on the NGS. Technically it is integrated into a
portal, whereas our approach is a proxy which can be
integrated in other portals.
The work from Takaaki, K. et.al. (Takaaki et al.,
2011) developed a web forward proxy server with
authentication method using Shibboleth. This proxy
solves problems in basic access authentication and di-
gest access authentication supported by existing web
forward proxy servers. Here the user already has to
use a proxy and it is not supporting federations.
The approach discussed in this paper is exclu-
sively based on open source technologies which
makes it particularly interesting in an academic en-
vironment. For example TU Munich also uses Shib-
boleth as a single sign-on system (Hommel, 2010).
This proxy solution could also be interesting for such
infrastructures outside of cloud environments if use
cases exist where additional flexibility is needed.
7 CONCLUSIONS AND FUTURE
WORK
In this paper a solution to reduce the number of ser-
CLOSER2012-2ndInternationalConferenceonCloudComputingandServicesScience
94
service provider instances in a dynamic cloud en-
vironment using a Shibboleth reverse proxy is pre-
sented. This reduction results in a more flexible,
easily automatable and simpler administration of ser-
vices while using Shibboleth-based single sign-on,
authentication and authorization.
Two mechanisms for identifying services using
URLs have been presented. The path-based mecha-
nism has proven to be more flexible while providing
best compatibility even to non-current web browsers.
Additionally, an evaluation has been conducted
which has shown, that using a shibboleth service
provider on a reverse proxy is insignificantly slower
than directly accessing web sites and services. Com-
pared to the gains in manageability, flexibility and se-
curity these performance drops are very acceptable.
The presented solution obviously suffers from a
single point of failure, the proxy. Future work will
analyze a clustered architecture to overcome perfor-
mance and single point of failure problems.
REFERENCES
RFC1918 - Address Allocation for Private Internets.
Transport Layer Security (TLS) Extensions.
(2011). Shibboleth. http://shibboleth.internet2.edu/.
Cloud Research Lab - Department of Computer Science
(2011). http://www.wolke.hs-furtwangen.de/.
Cloud Research Lab HFU. StudiCloud. http://
opennebula.org/software:ecosystem:studicloud.
DFN-AAI (2011). https://www.aai.dfn.de/.
Hommel, W. (2010). Campus single sign-on und
hochschul
¨
ubergreifendes identity management. In
Bode, A. and Borgeest, R., editors, Informationsman-
agement in Hochschulen, pages 221–232. Springer
Berlin Heidelberg. 10.1007/978-3-642-04720-6 19.
Jie, W., Young, A., Arshad, J., Finch, J., Procter, R., and
Turner, A. (2008). A guanxi shibboleth based security
infrastructure. In Enterprise Distributed Object Com-
puting Conference Workshops, 2008 12th, pages 151
–158.
OpenNebula (2011). http://www.opennebula.org/.
Shibboleth 2 Documentation (2011a). Communicating
with a Service Provider. https://spaces.internet2.edu/
display/SHIB2/IdPSPCommunicate.
Shibboleth 2 Documentation (2011b). Shibboleth SP
clustering. https://wiki.shibboleth.net/confluence/
display/SHIB2/NativeSPClustering.
Sulistio, A., Reich, C., and Doelitzscher, F. (2009). Cloud
infrastructure & applications cloudia. In Jaatun,
M., Zhao, G., and Rong, C., editors, Cloud Comput-
ing, volume 5931 of Lecture Notes in Computer Sci-
ence, pages 583–588. Springer Berlin / Heidelberg.
10.1007/978-3-642-10665-1 56.
Takaaki, K., Hiroaki, S., Noritoshi, D., and Ken, M. (2011).
Design and implementation of web forward proxy
with shibboleth authentication. In Applications and
the Internet (SAINT), 2011 IEEE/IPSJ 11th Interna-
tional Symposium on, pages 321 –326.
TLS SNI Test Site: *.sni.velox.ch (2011). https://
sni.velox.ch/.
Wang, X. D., Jones, M., Jensen, J., Richards, A., Wallom,
D., Ma, T., Frank, R., Spence, D., Young, S., Dev-
ereux, C., and Geddes, N. (2009). Shibboleth access
for resources on the national grid service (sarongs). In
Information Assurance and Security, 2009. IAS ’09.
Fifth International Conference on, volume 2, pages
338 –341.
SHIBBOLETHWEB-PROXYFORSINGLESIGN-ONOFCLOUDSERVICES
95