[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <op.xc5s3ysiyldrnw@laptop-air>
Date: Sat, 22 Mar 2014 22:20:12 -0700
From: "Jeremy Spilman" <jeremy@...link.co>
To: discussions@...sword-hashing.net
Subject: Re: [PHC] Transforming hash to different cost setting
On Sat, 22 Mar 2014 22:06:57 -0700, Jeremy Spilman <jeremy@...link.co>
wrote:
> If you start at 5/5 (meaning a single round at cost '5') and increase
> all hashes offline to be 5/7 (consecutive rounds at 5, 6, 7) then you
> want to find the single round (8/8?) which matches the latency of 5/7 as
> closely as possible.
'5/7' could also mean consecutive rounds just at costs 5 and 7. There's no
point in running more than one additional round for a single offline
upgrade, since it's guaranteed to be less efficient than just picking just
one additional round/cost setting to add.
That is, unless you actually have stacked multiple offline upgrades onto
the same hash while the user still hasn't ever successfully logged in, in
which case you would be forced to keep stacking. But now we're storing
'cost[]' with each hash.
As Krisztian pointed out, at some point you could just clear out JUST the
password to force a reset if they ever do show back up. But then, to avoid
the same attack, you need to actually hide this fact from the user.
Luckily, they haven't logged into your system in so long, they can't
possible remember their old password, so they'll inevitably just blame
themselves for forgetting it when they click 'Reset'. That is, unless
they're using a password manager, in which case they'll probably assume
your site was hacked.
Forget "too many secrets" more like "too many trade-offs"!
Powered by blists - more mailing lists