🚀 go-pugleaf

dhesi@cirrus.com (Rahul Dhesi) writes:
[t0 is the time at which the intruder acquires the encrypted password,
T1 is the time to crack it,
T2 is the time to use it (login)]

>It makes sense to set the password expiration period to be less than T1 +
>T2.  Then, by the time the intruder is ready to use the password, it has
>been changed.
...
>For well-chosen Unix passwords T1 is at least decades, if not several
>centuries...
>It should be reasonable to expire a password in about a year.
...
>CONCLUSION 1:  If passwords are well-chosen, password expiration adds
>nothing to security.

This conclusion does not follow from your above discussion.  It looked like
you were leading up to saying that expiring passwords annually will add to
security.  I thought that the exabyte tape method decreases T1 to be less
than decades even for reasonable passwords.  This is not including the possible
availability of the plaintext of passwords (e.g. ethernet snooping, looking
over people's shoulder).

>If the intruder can
>immediately get access to the encrypted form a password when it is
>changed, then password expiration becomes irrelevant.  Mathematically
>speaking, if the intruder is always trying the *current* encrypted
>password, his chances of guessing it neither increase nor decrease if the
>encrypted password changes every 6 hours (or however often).

This conclusion does not follow from your above discussion.  How frequently the
intruder gets access to the encrypted password is your "t0", which does not
appear after your original definitions.  Access to the current encrypted
password has no effect given your analysis on the effectiveness of password
expiry.