SPAKE and Diffie-Hellmann

Simple password-based encrypted key exchange (SPAKE) [4][5] is a fairly recent pre-authentication method supported by MIT Kerberos version 1.17 or newer. It is based on an asymmetric Diffie-Hellmann key exchange [6], but unlike PKINIT, no additional infrastructure in the form of a public key infrastructure (PKI) is required.

Since MIT Kerberos version 1.14, the library has also supported "authentication indicators." You can use these to define requirements for the strength of the initial user login, if needed, before the user is ultimately given a Kerberos service ticket. Therefore, you can meet the security requirements of various systems by requiring more or less strong authentication as a function of the system. In the /var/kerberos/krb5kdc/kdc.conf KDC configuration file you simply define different labels for the pre-authentication methods:

pkinit_indicator = pkinit
spake_preauth_indicator = spake
encrypted_challenge_indicator = fast

You then apply these labels to the different service principals. For example, if you want to issue a Kerberos service ticket for an SSH bastion host on the condition that the user uses the PKINIT pre-authentication method during the initial login, you can set the label for the principal as follows:

kadmin setstr host/ssh-bastion.example.com require_auth pkinit

If you now only use a password for the initial login and then try to get a service ticket for the host with kvno, the attempt will fail:

kinit tscherf
Password for tscherf@EXAMPLE.COM
kvno host/ssh-bastion.example.com
kvno: KDC policy rejects request while getting credentials for host/ssh-bastion.example.com@EXAMPLE.COM

However, if you use PKINIT for the initial login,

kinit -X X509_user_identity='FILE:/tmp/tscherf.pem' tscherf
kvno host/ssh-bastion.example.com
host/ssh-bastion.example.com@EXAMPLE.COM: kvno = 1

you are given the desired service ticket for the SSH host.

Conclusions

Thanks to authentication indicators, Kerberos now offers granular control of the pre-authentication method a user needs to use to gain access to a particular Kerberos service. Depending on how security-critical a service is, it can make sense, for example, only to grant access if FAST or SPAKE were used for the initial login.