This section describes the first layer of security that remote users must pass through when they attempt to access a network. The first step in the Kerberos authentication process is for users to authenticate themselves to the boundary router. The following process describes how users authenticate to a boundary router:

1. The remote user opens a PPP connection to the corporate site router.
2. The router prompts the user for a username and password.
3. The router requests a TGT from the KDC for this particular user.
4. The KDC sends an encrypted TGT to the router that includes (among other things) the user's identity.
5. The router attempts to decrypt the TGT using the password the user entered. If the decryption is successful, the remote user is authenticated to the router.

A remote user who successfully initiates a PPP session and authenticates to the boundary router is inside the firewall but still must authenticate to the KDC directly before being allowed to access network services. This is because the TGT issued by the KDC is stored on the router and is not useful for additional authentication unless the user physically logs on to the router.

This section describes how remote users who are authenticated to the boundary router authenticate themselves to a KDC.

When a remote user authenticates to a boundary router, that user technically becomes part of the network; that is, the network is extended to include the remote user and the user's machine or network. To gain access to network services, however, the remote user must obtain a TGT from the KDC. The following process describes how remote users authenticate to the KDC:

1. The remote user, at a workstation on a remote site, launches the KINIT program (part of the client software provided with the Kerberos protocol).
2. The KINIT program finds the user's identity and requests a TGT from the KDC.
3. The KDC creates a TGT, which contains the identity of the user, the identity of the KDC, and the expiration time of the TGT.
4. Using the user's password as a key, the KDC encrypts the TGT and sends the TGT to the workstation.
5. When the KINIT program receives the encrypted TGT, it prompts the user for a password (this is the password that is defined for the user in the KDC).
6. If the KINIT program can decrypt the TGT with the password the user enters, the user is authenticated to the KDC, and the KINIT program stores the TGT in the user's credential cache.

At this point, the user has a TGT and can communicate securely with the KDC.In turn, the TGT allows the user to authenticate to other network services.

The following process describes how a remote user with a TGT authenticates to network services within a given Kerberos realm. Assume the user is on a remote workstation (Host A) and wants to log in to Host B.

1. The user on Host A initiates a Kerberized application (such as Telnet) to Host B.
2. The Kerberized application builds a service credential request and sends it to the KDC. The service credential request includes (among other things) the user's identity and the identity of the desired network service. The TGT is used to encrypt the service credential request.
3. The KDC tries to decrypt the service credential request with the TGT it issued to the user on Host A. If the KDC can decrypt the packet, it is assured that the authenticated user on Host A sent the request.
4. The KDC notes the network service identity in the service credential request.
5. The KDC builds a service credential for the appropriate network service on Host B on behalf of the user on Host A. The service credential contains the client's identity and the desired network service's identity.
6. The KDC then encrypts the service credential twice. It first encrypts the credential with the SRVTAB that it shares with the network service identified in the credential. It then encrypts the resulting packet with the TGT of the user (who, in this case, is on Host A).
7. The KDC sends the twice-encrypted credential to Host A.
8. Host A attempts to decrypt the service credential with the user's TGT. If Host A can decrypt the service credential, it is assured the credential came from the real KDC.
9. Host A sends the service credential to the desired network service. Note that the credential is still encrypted with the SRVTAB shared by the KDC and the network service.
10. The network service attempts to decrypt the service credential using its SRVTAB.
11. If the network service can decrypt the credential, it is assured the credential was in fact issued from the KDC. Note that the network service trusts anything it can decrypt from the KDC, even if it receives it indirectly from a user. This is because the user first authenticated with the KDC.

At this point, the user is authenticated to the network service on Host B. This process is repeated each time a user wants to access a network service in the Kerberos realm.

1.    Router# ankusername@REALM

2.    Router# ankusername/instance@REALM

For example, to add user loki of Kerberos realm CISCO.COM, enter the following Kerberos command:

ank loki@CISCO.COM

Note	The Kerberos realm name must be in uppercase characters.

You might want to create privileged instances to allow network administrators to connect to the router at the enable level, for example, so that they need not enter a clear text password (and compromise security) to enter enable mode.

To add an instance of loki with additional privileges (in this case, enable, although it could be anything) enter the following Kerberos command:

ank loki/enable@CISCO.COM

In each of these examples, you are prompted to enter a password, which you must give to user loki to use at login.

The Enabling Kerberos Instance Mapping describes how to map Kerberos instances to various Cisco IOS XE privilege levels.

All routers that you want to authenticate to use the Kerberos protocol must have an SRVTAB. This section and the Extracting SRVTABs describe how to create and extract SRVTABs for a router called router1 . The Copying SRVTAB Files describes how to copy SRVTAB files to the router.

To make SRVTAB entries on the KDC, use the following command in privileged EXEC mode:

Router# ark SERVICE/HOSTNAME@REALM

For example, to add a Kerberized authentication service for a Cisco router called router1 to the Kerberos realm CISCO.COM, enter the following Kerberos command:

ark host/router1.cisco.com@CISCO.COM

Make entries for all network services on all Kerberized hosts that use this KDC for authentication.

SRVTABs contain (among other things) the passwords or randomly generated keys for the service principals you entered into the KDC database. Service principal keys must be shared with the host running that service. To do this, you must save the SRVTAB entries to a file, then copy the file to the router and all hosts in the Kerberos realm. Saving SRVTAB entries to a file is called extracting SRVTABs. To extract SRVTABs, use the following command in privileged EXEC mode:

Router# xst
 router-name host

For example, to write the host/router1.cisco.com@CISCO.COM SRVTAB to a file, enter the following Kerberos command:

xst router1.cisco.com@CISCO.COM host

Use the quit command to exit the kdb5_edit program.

1.    Router(config)# kerberos local-realmkerberos-realm

2.    Router(config)# kerberos serverkerberos-realm {hostname | ip-address } [port-number ]

3.    Router(config)# kerberos realm {dns-domain | host } kerberos-realm

Note	Because the machine running the KDC and all Kerberized hosts must interact within a 5-minute window or authentication fails, all Kerberized machines, and especially the KDC, should be running the Network Time Protocol (NTP).

The kerberos local-realm, kerberos realm, and kerberos server commands are equivalent to the UNIX krb.conf file. The table below identifies mappings from the Cisco IOS XE configuration commands to a Kerberos 5 configuration file (krb5.conf).

default_realm = DOMAIN.COM

kerberos local-realm
 DOMAIN.COM

.domain.com = DOMAIN.COM

domain.com = DOMAIN.COM

kerberos realm
.domain.com
 DOMAIN.COM
kerberos realm
 domain.com DOMAIN.COM

kdc = DOMAIN.PIL.COM:750

admin_server = DOMAIN.PIL.COM

default_domain = DOMAIN.COM

kerberos server
 DOMAIN.COM 172.65.44.2
(172.65.44.2
 is the example IP address for DOMAIN.PIL.COM
)

For an example of defining a Kerberos realm, see the Defining a Kerberos Realm Examples module.

To make it possible for remote users to authenticate to the router using Kerberos credentials, the router must share a secret key with the KDC. To do this, you must give the router a copy of the SRVTAB you extracted on the KDC.

The most secure method to copy SRVTAB files to the hosts in your Kerberos realm is to copy them onto physical media and go to each host in turn and manually copy the files onto the system. To copy SRVTAB files to the router, which does not have a physical media drive, you must transfer them via the network using TFTP.

To remotely copy SRVTAB files to the router from the KDC, use the following command in global configuration mode:

Router(config)# kerberos srvtab remote {hostname
 | ip-address
} {filename
}

When you copy the SRVTAB file from the router to the KDC, the kerberos srvtab remote command parses the information in this file and stores it in the router's running configuration in the kerberos srvtab entryformat. To ensure that the SRVTAB is available (does not need to be acquired from the KDC) when you reboot the router, use the write memory configuration command to write your running configuration (which contains the parsed SRVTAB file) to NVRAM.

For an example of copying SRVTAB files, see the SRVTAB File Copying Example.

You have now configured Kerberos on your router. This makes it possible for the router to authenticate using Kerberos. The next step is to tell it to do so. Because Kerberos authentication is facilitated through AAA, you need to enter the aaa authentication command, specifying Kerberos as the authentication method. For more information, refer to the chapter "Configuring Authentication".

With Kerberos configured thus far, a user authenticated to a Kerberized router has a TGT and can use it to authenticate to a host on the network. However, if the user tries to list credentials after authenticating to a host, the output will show no Kerberos credentials present.

You can optionally configure the router to forward users' TGTs with them as they authenticate from the router to Kerberized remote hosts on the network when using Kerberized Telnet, rcp, rsh, and rlogin (with the appropriate flags).

To force all clients to forward users' credentials as they connect to other hosts in the Kerberos realm, use the following command in global configuration mode:

Router(config)# kerberos credentials forward

With credentials forwarding enabled, users' TGTs are automatically forwarded to the next host they authenticate to. In this way, users can connect to multiple hosts in the Kerberos realm without running the KINIT program each time to get a new TGT.

To use Kerberos to authenticate users opening a Telnet session to the router from within the network, use the following command in global configuration mode:

Router(config)# aaa authentication login {default | list-name
} krb5_telnet

Although Telnet sessions to the router are authenticated, users must still enter a clear text password if they want to enter enable mode. The kerberos instance map command, discussed in a later section, allows them to authenticate to the router at a predefined privilege level.

Another way for users to open a secure Telnet session is to use Encrypted Kerberized Telnet. With Encrypted Kerberized Telnet, users are authenticated by their Kerberos credentials before a Telnet session is established. The Telnet session is encrypted using 56-bit Data Encryption Standard (DES) encryption with 64-bit Cipher Feedback (CFB). Because data sent or received is encrypted, not clear text, the integrity of the dialed router or access server can be more easily controlled.

Note	This feature is available only if you have the 56-bit encryption image. 56-bit DES encryption is subject to U.S. Government export control regulations.

To establish an encrypted Kerberized Telnet session from a router to a remote host, use either of the following commands in EXEC command mode:

Router(config)# connect host
 [port
] /encrypt kerberos

Router(config)# telnet host
 [port
] /encrypt kerberos

When a user opens a Telnet session from a Cisco router to a remote host, the router and remote host negotiate to authenticate the user using Kerberos credentials. If this authentication is successful, the router and remote host then negotiate whether or not to use encryption. If this negotiation is successful, both inbound and outbound traffic is encrypted using 56-bit DES encryption with 64-bit CFB.

When a user dials in from a remote host to a Cisco router configured for Kerberos authentication, the host and router will attempt to negotiate whether or not to use encryption for the Telnet session. If this negotiation is successful, the router will encrypt all outbound data during the Telnet session.

If encryption is not successfully negotiated, the session will be terminated and the user will receive a message stating that the encrypted Telnet session was not successfully established.

For information about enabling bidirectional encryption from a remote host, refer to the documentation specific to the remote host device.

For an example of using encrypted Kerberized Telnet to open a secure Telnet session, see the Encrypted Telnet Session Example.

As an added layer of security, you can optionally configure the router so that, after remote users authenticate to it, these users can authenticate to other services on the network only with Kerberized Telnet, rlogin, rsh, and rcp. If you do not make Kerberos authentication mandatory and Kerberos authentication fails, the application attempts to authenticate users using the default method of authentication for that network service; for example, Telnet and rlogin prompt for a password, and rsh attempts to authenticate using the local rhost file.

To make Kerberos authentication mandatory, use the following command in global configuration mode:

Router(config)# kerberos clients mandatory

As mentioned in the Creating SRVTABs on the KDC, you can create administrative instances of users in the KDC database. The kerberos instance map command allows you to map those instances to Cisco IOS XE privilege levels so that users can open secure Telnet sessions to the router at a predefined privilege level, obviating the need to enter a clear text password to enter enable mode.

To map a Kerberos instance to a Cisco IOS XE privilege level, use the following command in global configuration mode:

Router(config)# kerberos instance map instance
 privilege-level

If there is a Kerberos instance for user loki in the KDC database (for example, loki/admin ), user loki can now open a Telnet session to the router as loki/admin and authenticate automatically at privilege level 15, assuming instance "admin" is mapped to privilege level 15. (See the Opening a Telnet Session to the Router.)

Cisco IOS XE commands can be set to various privilege levels using the privilege levelcommand.

After you map a Kerberos instance to a Cisco IOS XE privilege level, you must configure the router to check for Kerberos instances each time a user logs in. To run authorization to determine if a user is allowed to run an EXEC shell based on a mapped Kerberos instance, use the aaa authorization command with the krb5-instance keyword. For more information, refer to the chapter "Configuring Authorization."

1.    Router# show kerberos creds

2.    Router# clear kerberos creds