Re: ArcotSign (was Re: Does security depend on hardware?)

[Mike Stay <staym@accessdata.com>]

>>In that case please allow me to go back to a point raised by me
>>previously. The user uses his 'remembered secret' (of fewer bits) 
>>through a public algorithm (including protocol) to retrieve from a 
>>pool the password (of more bits). If the attacker doesn't have the 
>>pool then everything looks fine. But if he manages to get the pool
>>(a case someone mentioned in this thread) then he can obviously
>>brute force offline, I believe, since he possesses now everything
>>the legitimate user has, excepting the 'remembered secret'. Or is
>>there anything wrong with my logic?
>
>Yes.  There is something wrong with you logic.
>
>Bruce
>**********************************************************************
>Bruce Schneier, President, Counterpane Systems     Phone: 612-823-1098
>101 E Minnehaha Parkway, Minneapolis, MN  55419      Fax: 612-823-1590
>           Free crypto newsletter.  See:  http://www.counterpane.com

According to the website, there's no pool of passwords.  There's a
truncated hash that will catch most mistakes, but is useless as a test
criterion in a dictionary attack.  If you get the user's "public" key,
then you can do a dictionary attack.  The user's "public" key isn't
public, however; not even the user knows it.  If I'm understanding it
right, it's stored encrypted and the key is only given to a set of
predefined servers.  Prior relationship must exist; they admit that.
-- 
Mike Stay
Cryptographer / Programmer
AccessData Corp.
mailto:staym@accessdata.com