Re: Poor article
Sorry that's a bit garbled, I'm not well at the moment. Say you have a dictionary including common passwords. You then get access to the a set of bcrypt12 hashes and the salt . You can now begin to check for passwords - you add the salt to each password in your dictionary and run it through bcrypt12. Problem - that is a slow algorithm (on purpose). However, AM had also stored the MD5s of some tokens they had foolishly made (I may be simplifying a bit) by concatenating together lowercase usernames, passwords and a salt. "johnhwoods::password123::salt". Now, MD5 is fast, and you know the usernames and the salt, so you can very quickly look for collisions. If you find that password123 gives you a collision, you know that some case variant of it is the answer. So now, you only need to check 256 case variants: you'd probably start "as is" then the 8 combinations with one capital, then the 28 with two etc. Suddenly instead of needing to run your whole dictionary through bcrypt you just have a few variations.
Biting the hand that feeds IT
