---------- From: David Honig[SMTP:honig@sprynet.com] Sent: Wednesday, September 13, 2000 11:26 PM To: Trei, Peter; Multiple recipients of list Subject: RE: Is kerberos broken? cpunk

At 11:06 AM 9/13/00 -0400, Trei, Peter wrote:

...

Here's an example of a good passphrase:

"David grossly underestimates the ability of homo sapiens to memorize and exactly reproduce long texts. An examination of American high school students ability to perform the Gettysburg Address is a good counterexample."

222 bytes, more or less. Even if we assume only 1bit of entropy per character (it's ordinary english), that's a pretty tough space to search. It's a safe bet that those two sentences have never been placed together in all of human history before now, so there's no dictionary to check.

The problem is not that passphrases *can't* be made secure - the problem is that most people are unwilling to use good ones.

Peter Trei

Well I'm flattered :-) and impressed. I would be more impressed if e.g., you actually used such an entropic phrase, in real life. Of course, we don't expect you reveal the actual length of your 'phrase.

My passphrases are of substantial length. As for enterprise logins, 'we have a solution to that problem' :-) http://www.rsasecurity.com/products/securid/

I think you have convinced me, reinforcing something I've learned and propogated: convenience over security. You have also reinforced something that fits with what I know of cog sci, and which gets to the limits of H. sapiens: you can only remember large things if they're structured 'meaningfully'. Kasparov can't remember *random* chessboards better than you, only real ones.

DH, CSEE & Cog Sci '86

It's interesting - structure reduces the entropy by making things predictable, but also makes them capable of memorization, despite non-trivial amounts of remnant entropy. Peter