Achieving Nomadicity: Accessing the Internet Anytime, Anywhere

Federated Identity

When you cross the border into another country or when you enter a shop that offers Wi-Fi, you really don't want to have to sign up for a contract each time to connect to a network, not to mention the burden of remembering all the different usernames and passwords.

Here is where federated identity comes in. In the federated model, a user has a contract with only one (or a few) operators—the "home operator" that establishes the identity of the user. That one identity is then used to gain access to networks or applications managed by other operators. To make this work, the operators of the different networks need to establish a roaming or federation agreement. Such an agreement specifies under what conditions a visited network accepts an authentication statement ("this is a valid user") from the home operator, how the authentication credentials and accounting data are exchanged, and what financial arrangement is in place for visiting users.

So, in this model, as shown in Figure 3-5, the home operator (in identity lingo called the Identity Provider [IdP]) acts as a trusted third party for the serving operator, called the Service Provider (SP) or Relying Party (RP). There is no direct contractual or trust relationship between the user and the visited network, only between the user and the IdP and between the IdP and the RP. Because there is a trust relationship between the user and IdP and between the IdP and RP, the RP "trusts" the user.

Figure 3-5 Federated Identity

The RADIUS¹⁰ protocol allows forwarding of authentication requests to another RADIUS server; this is why RADIUS is widely used for network roaming. The authentication requests here are forwarded by the RP (usually also a RADIUS server) to the RADIUS server of the home operator (the IdP), and the outcome of the authentication is sent back. By combining RADIUS with EAP, the confidentiality of the user credentials can be preserved.

For access to web-based applications, a number of different protocols are used, such as Security Assertion Markup Language (SAML¹¹), OpenID¹², or Open Authentication (OAuth)¹³.

When users roam between networks using the same technology (that is, from one Long Term Evolution (LTE) network to another LTE or similar network), this is called horizontal roaming. Roaming between different types of networks, such as roaming from an LTE network to a Wi-Fi network, is called vertical roaming.

3GPP standardizes access to non-3GPP networks but places the LTE core network firmly in control; authentication is always performed in the LTE EPC. IEEE (the standardization body for the Wi-Fi standards) in its 802.21 standard¹⁴ addresses vertical roaming with a more equal role for the different access technologies but does not address federated network authentication.

Federated Access in LTE

3GPP distinguishes two types of federated access:

• 3GPP access¹⁵: Describes horizontal roaming.
• Non-3GPP access¹⁶: Describes vertical roaming.

In both cases, the home network needs to establish a roaming agreement beforehand (and the user's subscription should allow roaming access).

3GPP Access

3GPP access is access to an LTE network of another operator or to a UMTS or GPRS network. UMTS networks (and GPRS networks that support interworking with LTE) support AKA authentication.

Based on the IMSI (that contains a country and an operator code), the MME can ask the home HSS of the user to verify the user rather than the HSS in the serving network. The home HSS must check whether the subscription agreement with the user allows roaming, but apart from that, the authentication process is the same as for the nonroaming case.

Non-3GPP Access

Examples of non-3GPP access are CDMA-2000, WiMAX, and Wi-Fi. These networks don't use the same authentication methods, and the elements in these serving networks don't understand how to deal with an AKA authentication. So, rather than involving a network element in the serving network directly in the authentication flow with the home network, EAP is used instead. Here EAP provides the necessary abstraction from the actual authentication using AKA. For this purpose, EAP-AKA and its more secure successor EAP-AKA' (EAP-AKA Prime) have been created. The EAP identity contains the IMSI or a pseudonymous identifier that was established in a prior authentication to locate the authentication server for the user.

Figure 3-6 shows how AKA authentication can be encapsulated in EAP.

Figure 3-6 EAP-AKA

Although you can use the IP connectivity of the serving network, the typical use is to tunnel all traffic back to the home network using IPsec.

Federated Access to Wi-Fi Networks

Originally, Wi-Fi was intended to be used as a local-area network technology, typically covering an area with a radius of some 30–50 meters. Nowadays, sometimes hundreds to thousands of Wi-Fi access points together form hotspots that provide coverage to complete campuses or even cities. Still, the majority of the hotspot operators provide access to an area with a limited geographical scope (unlike the nationwide coverage that cellular operators provide). To provide coverage beyond the geographical region, hotspot operators need to collaborate so that subscribers of one operator can gain access to the network of another operator.

Roaming to Other Wi-Fi Networks

The main challenges in roaming access for Wi-Fi networks are setting up the roaming agreements and verifying the user credentials at the home network.

Because, unlike cellular networks, Wi-Fi hotspots are by virtue of the local-area character of the technology relatively small, setting up roaming agreements with a large number of Wi-Fi operators is hard to scale. To solve the scaling problem, three different types of organizational models emerge:

• The first model mimics the cellular model. A large operator acquires or leases a large number of hotspots and unifies the authentication across these hotspots. AT&T hotspots and T-Mobile hotspots are examples of this arrangement.
• In the second model, a third party acts as a broker for a large number of hotspot operators. The users have a contract with the broker and authenticate and pay for access to the broker. The broker in turn pays the hotspot operator. Examples of this arrangement are Boingo and iPass.
• The last model has individual hotspot operators join forces and agree on roaming conditions and credential verification methods. Examples of the latter are FON and the Wi-Fi roaming infrastructures that many schools worldwide participate in—eduroam. (This is further explained later in the section "Example of Wi-Fi Roaming: eduroam.")

Verification of the credentials of the users at the home network requires transporting the credentials to the home network and sending the outcome of the authentication back to the visited network. The dominant transport protocol for transporting the credentials is RADIUS.

The main advantage of the captive portal method for Wi-Fi is that it only requires a web browser on the user device. This is also why it is the most commonly used access method at public hotspots. The main downside is that because the Wi-Fi link is unprotected, simple MAC spoofing can be used by an attacker to piggyback an authenticated user's connection. Additionally, the user credentials are visible to every hotspot operator (they have to be entered in the web page that the captive portal shows) and can be observed by every RADIUS server in the path to the home RADIUS server. When 802.1X is used, the combination of 802.1X, EAP, and RADIUS allows user credential privacy. This means that users don't have to worry about giving their password to potentially thousands of hotspot operators, let alone rogue hotspot operators. An added benefit of using 802.1X is that all user traffic is encrypted over the Wi-Fi radio link, allowing the operator to be sure that every packet sent into the network originated from an authentic Wi-Fi user.

The added security features of 802.1X and better support in the most common operating systems have resulted in a slow but steady increase in use, especially in corporate environments.

802.11u

Two issues that are particularly important for Wi-Fi access are the fact that most Wi-Fi hotspots are relatively small and that there are thousands of them. In a densely populated area, a user easily often "sees" 30 or 40 different Wi-Fi networks, without knowing which of those will have a roaming agreement with the home operator and, if so, under what conditions.

This is the problem space that the upcoming IEEE 802.11u¹⁷ standard addresses. Hotspots that are 802.11u enabled can broadcast information about the roaming consortia they belong to and under what conditions they can be used.

Example of Wi-Fi Roaming: eduroam

An example of a Wi-Fi roaming service is eduroam¹⁸. This service is limited to educational institutions. However, its technical setup and broad uptake (more than 500 universities in some 50 countries with over 10 million users) warrant attention.

eduroam started out in the Netherlands in 2003 and gained fast popularity in most European countries and later in Australia, Japan, Hong Kong, and Canada. Lately, U.S. schools are joining eduroam and Internet2 is supporting the initiative.

Figure 3-7 shows the European national research and education networks that participate in eduroam. (For an up-to-date overview of all participating institutions in Europe and elsewhere, refer to the eduroam website.¹⁹)

Figure 3-7 European National Research and Education Networks Participating in eduroam as of May 2010 (courtesy of TERENA)

eduroam consists of a few basic elements, described in the following paragraphs.

A RADIUS hierarchy is set up consisting of a set of institutional (redundant, for failover purposes), national, and continental RADIUS servers. All institutional RADIUS servers connect to the national servers in their country. All national servers connect to the top-level servers for their continent, and the continental servers (Europe, America, and Asia-Pacific) connect to each other.

Figure 3-8 shows the RADIUS hierarchy that constitutes eduroam. The top-level servers that are fully meshed know which top-level servers serve what national domains. The national servers are connected to all institutional servers in their country and to the top-level servers in their continent. The institutional servers are connected to their national servers.

Figure 3-8 The eduroam Hierarchy

802.1X is used for secure access to the institutional Wi-Fi networks.

EAP is used to protect user credentials. EAP identities are of the form anonymous@domain-name-of-institution or, instead of anonymous, a pseudonymous identifier. Users' authentication requests are forwarded through the RADIUS hierarchy based on the domain name of the institution to which the user belongs.

In other words, the home institution authenticates the user and the serving institution authorizes the user for access. The home institution of the user can decide which authentication method and what EAP method to use.

Figure 3-9 shows a typical eduroam authentication, which is described in further detail in the list that follows.

Figure 3-9 The eduroam Basic Operation (Courtesy of SURFnet)

1. A user from University B in the Netherlands tries to gain access to the network at University A, also in the Netherlands.
2. The authenticator asks the user (or rather the supplicant) to authenticate.
3. The user sends the authentication credentials encapsulated in EAP with an EAP identity of anonymous@university_b.nl to the authenticator.
4. The authenticator at University A forwards the EAP message to the RADIUS server of University A.
5. The University A RADIUS server observes that the EAP identity does not belong to University A and forwards the EAP message to the national RADIUS server for the Netherlands operated by SURFnet, the Dutch research and education network.
6. The SURFnet RADIUS server for the .nl domain sees that the EAP identity belongs to University B and forwards the EAP message to the University B RADIUS server. (If the EAP identity were not for the .nl domain, the EAP message would be forwarded to the European top-level server.)
7. The University B RADIUS server deencapsulates the EAP message and verifies the credentials.
8. University B sends the result of the authentication back along the same route.
9. The RADIUS server at University A instructs the authenticator to allow access to the user (and possibly to assign the user to a specific VLAN for guests).

Federated Access to Applications with SAML

When you assume that more and more applications will be offered "in the cloud," it is imperative that scalable mechanisms exist for federated identity. The most widespread systems for federated identity to (mainly) web-based application make use of the Security Assertion Markup Language (SAML) protocol suite. SAML is an XML-based markup language for transporting authorization assertions between IdPs and RPs.

Figure 3-10 shows a typical SAML (version 2.0) flow, which is further described in the list that follows:

1. The user uses his browser to try to access a resource under control by the RP.
2. The RP issues an authentication request to the browser (plus a redirect to the IdP).
3. The browser sends the authentication request to the IdP and asks for an authentication statement.
4. The user (if not already authenticated) authenticates at the IdP.
5. The IdP issues an authentication statement to the browser stating that the user is successfully authenticated (plus a redirect back to the RP).
6. The browser presents the authentication statement to the RP.
7. The RP gives the user access to the resource (assuming that the user satisfies the RP's policies and a roaming arrangement exists between the IdP and RP).

Figure 3-10 SAML 2.0 Authentication Flow