[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: adding many users

On Thu, Mar 13, 2003 at 06:20:14AM -0500, Anthony E. Greene wrote:
> There are password generators that can generate passwords that
> consist of a combination of dictionary words and special characters,
> eg. quick23walk.

I use a utility called mnencode this way:

  $ head -c 4 /dev/random | mnencode    

And get three word long results like:

iris-farmer-benny or person-london-multi or jumbo-joker-basil

Reasonably easy to type and remember, yet a significant 32-bits of
entropy--far better than most passwords.  (Enough for circumstances
where you don't have a motivated foe with the opportunity to brute
force it--a non-readable /etc/shadow is your friend here.)

To find mnencode see <http://www.tothink.com/mnemonic/>.  It is really
a carefully crafted word list and two complementary programs, mnencode
which turns binary data into words intended to be pronouncable,
spellable, and unambiguous, and mndecode which turns those words back
into that exact binary data.

To move offtopic, for really paranoid security, such as the passphrase
on the private half of a public/private key pair, you can do:

  $ head -c 16 /dev/random | mnencode

And you get 128-bits worth of entropy as, for example:


Take out the new line, put in single dashes throughout, and you have a
long passphrase that is really secure.  But it turns out that a
passphrase with 128-bits of entropy is pretty unwieldy.  It gets hard
to remember (was it jester or joker?, secure or secured?, etc), and it
is suprisingly hard to type blind.  I use exactly one such passphrase
(that I don't type on open wires or keyboards I don't control), but I
do use it to encode my other passwords.  


P.S.  A passphrase with 128-bits of entropy is enough that even a very
powerful and motivated foe will not be able to bruteforce it any time
soon--if ever--and will instead resort to bugging your keyboard,
hiding a camera over your keyboard, sniffing RF-emissions, rubberhose
cryptanalysis, etc.  For example, suppose the NSA really wants your
key and can try a trillion possibilities a second, it would still
take, on average, over 3-months crack a 64-bit passphrase--which is
well within their abilities if they are really interested.  However, a
128-bit passphrase is about 18,442,589,569,024,000,000 times as
difficult as that, something even the NSA can't accomplish.  Note that
this is for a symmetric key, public keys work different and need to be
much longer for equivalent strength.  A 4K bit public key can be
manipulated pretty easily by computers these days and is likely
extremely strong--depending upon possible breakthroughs in factoring
numbers or building quantum computers.

[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]