Monday, March 29, 2010

Dumb Info Security

A truly great question came up the other day.
"Why change passwords every 90 days? What is the threat scenario countered by that policy?"
Of course strong password policy means constantly changing passwords. Right?

Then I started to think about it. What -- actually -- does a password change protect you against?

The answer, it appears, is nothing. Changing passwords is largely a waste of time and money. I suppose that a password change prevents further abuse of the account. But generally, the abuse is not ongoing. Once in to a system, the trick is to create an additional privileged account that does not belong to any real user; all the password changes in the world have no effect.

This post is spot-on: "Password rules: Change them every 25 years"

In short, there's no threat that's actually countered by changing passwords. However, it's on everyone's checklist.

[Look at http://passcracking.com/hybrid.html for information on rainbow table attacks. The time required is on the order of 10 minutes.]

Since a weak password is broken in well under 90 days, there's no "moving target" to this. A weak password is -- effectively -- broken instantly compared to the 90-day password change. Once broken, the machine's freely available for -- on average -- 45 days.

The comments on this post are helpful also. Most people agree that password changes do not have any possible impact on security. Except that it gives security managers a chance to improve the rules and enforce everyone to change their passwords to meet the new rules.

Missing the Point

One comment that's interesting is this:
You've made two assumptions: 1) all password thieves will give up after a few tries in the case of brute-force attack, and 2) all thieves will give up after a few tries in the case of dictionary attacks.
This misses the point entirely. These two assumptions are not overlooked by this posting. They're not part of it at all. None of this is based on password thieves giving up.

Changing a password does not materially impact the thieves' ability to crack a password. Phishing, and Key Logging always work, no matter how often the password is changed.

A dictionary attack is trivially defeated by disabling the account after a few failures. Changing the password is of no relevance at all.

A rainbow table to undo a hashed password is defeated by using long salt strings with the hash. Changing passwords every 90 days has nothing to do with this, either. There's no "moving target" concept, since a rainbow table attack takes much less than 90 days.