Re:Moore’s Law won’t kill passwords
by acqq on Friday January 18, 2013 @03:11AM
(#5077681)
It's not just a computational expense of one password hashing that has to be considered, the possibility of having the hashes precomputed can't be ignored. In such a light, we definitely have to start using longer passwords or passphrases regardless of the expense of computing one hash. Otherwise the search space is simply too small.
[ Parent
| on HN ]
Pre-computed? You mean rainbow tables right? They have gone the way of the dinosaur with salts. Bcrypt has a big fat salt making it mightily impractical to use rainbow tables as the space required is so large.
Sadly, many systems for many years will still use unsalted MD5s. Rainbow tables will still have a place in the cracker's toolkit for some time.
Another problem with rainbow tables is that their size depends on the password length, and with current technology they become impractical almost as soon as brute forcing, and much sooner than smart dictionary attacks (e.g. hybrids where you take a dictionary and for each entry try every possible way to add 3 random characters).
I'm talking about what's on the systems we use, not about what's the theoretical best case for us. I do envy those living in their ivory towers, downvoting for a laugh.
The systems we use now use salts. If they don't they are doing it seriously wrong.