On Tue, Apr 26, Aaron Toponce wrote:
> On Tue, Apr 26, 2011 at 10:49:36AM -0600, Eric Wald wrote:
>> Over 25,000 reasonable straight-line passwords; double that to include
>> spirals.  Certainly with range for a dictionary attack, but it would
>> take long enough that I could re-print the card, print a new card, and
>> change all of my most important passwords before you're likely to have
>> cracked even one account.
>
> No, not by a long shot. Because the length of your password could be
> infinite, this makes the number of possibilities infinite, even in one
> direction. But even using limited length, say 20, you still have billions
> upons billions of combinations based on starting location, and direction
> traveled. No dictionary attack is feasible with this card.

Perhaps my math is faulty, but I'm having a hard time finding billions
of possible passwords unless you're willing to accept 500-character
passwords with loads of repetition.  I'm counting 29 columns, 8 rows, 8
straight-line directions, and 8 spiral directions.  I could see using
the 8 hippogonal directions, too, but that's a stretch.  For length, I'm
assuming that anything below 8 characters is ridiculously short for
someone security-conscious enough to use such a card, and 30 characters
(wrapping back to the starting column) is a reasonable upper limit.

That gives me 29*8 = 232 starting positions, 8+8+8 = 24 directions, and
len [8..30] = 23 lengths, for 128,064 potential passwords.  That's just
barely larger than my /usr/share/dict/words dictionary.

My 25,000 figure above was assuming only 8 directions, and 13 or 14
reasonable password lengths, which should take care of the most likely
usage scenarios; a cracker could very easily attempt those first, before
expanding to the hundred thousand less-likely candidates, much less the
shorter, longer, or crazy-path candidates.

On the other hand, I can accept the "billions upon billions" figure for
the crazy-path idea.  If the direction of the path is allowed to change
for each character, then that gives you 1.5 billion 8-character
passwords even if the direction always has to be orthogonal and away
from the previous character.  14 characters gets you to a trillion
passwords; 157 quadrillion if you include diagonals.  Granted, these
figures allow overlapping paths and wrapping around the edges, which
would be less likely in a real use case; far more likely is that someone
takes a password in one direction for a while, switching once or twice
before finishing.

At the extreme case, one could claim over a decillion passwords by
allowing any character after any other, but that assumes that each of
the 232 characters are unique.

- Eric

/*
PLUG: http://plug.org, #utah on irc.freenode.net
Unsubscribe: http://plug.org/mailman/options/plug
Don't fear the penguin.
*/

Reply via email to