Hello,

As I read at http://www.governmentsecurity.org/forum/in...showtopic=15193 ,
it is said that it is not possible cracking salted md5 hashes using rainbowtables.

The way I understand it, is that the salted hashes are md5 hashes of 2 hashes.
So simplified, it would look a bit like this:

salt MD5 hash = aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
pass MD5 hash = bbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbb

final, salted hash = md5(aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaabbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbb)

So in order to crack the pass,
you should first crack the salted hash, then take the last 32 characters of the 'plaintext' and crack this as another hash, right?

So the reason it is 'impossible' to crack these hashes using rainbowtables is because the plaintext is to long (64 characters)?

Am I right, or am I missing something?

The issue boils down to the difficulty of determining what the salt really is. For example, the following psuedocode:

sha1(md5(md5($password)+$registrationdate)+$salt);

Now if you have an md5 of "a8b148f3ed392a" or whatever, you can eventually, after several months, determine what part 1 and part 2 are combined. part 2 can be distinguished as the salt in this case, but part 1 is yet another hash. Without knowing what it is, after another few months can pass and you'll eventually get another part 1 and part 2, part 1 can be run through a rainbow table at this point. But keep in mind it took you months or even years to figure out this one password. If you md5 the registration date/time, it might increase the amount of work for someone to reverse engineer the password out of the situation.

The strength of policies like this is you never know which protection method the administrator is using without access to their source code.