Messages must pass between three parties: the supplicant, authenticator, and authentication server. IEEE 802.1X uses EAP, or more specifically, EAPOL, to pass these messages between the supplicant and the authenticator. Let's start by following through the sequence of events when a new supplicant arrives.
An outline of the authentication sequence is shown in Figure 8.10. When a supplicant wants to connect, it must first attract the attention of the authenticator. In most cases the authenticator is alerted by the connection process. It might be that the act of plugging in the cable or, in the case of wireless, associating with the access point is enough. Otherwise, the EAPOL-Start message can be used.
Figure 8.10. EAP Message Flow
The authenticator first responds with an EAP-Request/Identity message. This is a standard EAP message that is equivalent to shouting, "Who's there?" The authenticator is allowed to skip this step if it knows the identity of the supplicant by some other method. The supplicant must respond with an EAP-Response/Identity message. This raises an interesting issue because so far the supplicant can't be certain whether the authenticator can be trusted, especially in a wireless network. Suppose the authenticator is a rogue access point set up by an attacker. The supplicant might not want to reveal its identity at that time and uses a pseudonym instead. Some schemes support the use of pseudonyms ; but, for the moment, let's assume the supplicant is not shy and is prepared to send its identity to the authenticator.
So far all the messages we have discussed have gone between the supplicant and the authenticator (in IEEE 802.11, this would be the mobile device and the access point). The authentication server has not been involved until now. It is important not to waste the authentication server's time until the supplicant has shown that it actually speaks IEEE 802.1X by responding to the first EAP-Request. Having obtained the identity of the supplicant, the authenticator needs to contact the authentication server to find out whether this supplicant is to be allowed in. The authentication server can't make this decision until it has verified that the supplicant really corresponds to the identity it has given. This is the whole point of authentication. To avoid the need for the authenticator (in the access point) having to know all the authentication methods, the EAP messages used for authentication are passed directly to the authentication server.
In effect, during this phase the supplicant and the server are talking directly. In our earlier office building analogy, the security guard has opened the door and asked the person's name, but not let him in yet. Then the guard calls his boss and says, "Can we let Harry Acker in?" The boss runs through a set of questions, which the security guard asks Harry one by one. The guard passes the answers back to the boss. The guard just relays the questions and answers and, in the end, the boss makes a decision on entry. Note that during this phase, the guard might not understand the purpose of the questions as in the following scenario:
Note that the questions and answers relayed by the guard made no sense to him but clearly enabled the boss to make a decision that the guard then put into effect.
During the authentication process, the authenticator takes a quick look at each EAP message that is passed between the supplicant and authentication server. It is watching for certain messages that it understands. In particular, it is looking for an EAP-Success or an EAP-Failure. It must wait until the authentication server indicates whether the supplicant has been accepted or rejected. A RADIUS message provides the indication when RADIUS is being used.