3 ms·

Doesn't the length of a salt modify the effective length of the password? If the salt is 16 chars and the password is only 4 but a strong hashing algorithm is u

by nerdy 11y ago

Doesn't the length of a salt modify the effective length of the password? If the salt is 16 chars and the password is only 4 but a strong hashing algorithm is used wouldn't it still effectively be 20 characters? If that's the case (employed properly), doesn't it largely invalidate much of the article, at least the offline fast attack scenario? p.s. I am not advocating 4-character passwords, it was only an example!

bigiain 11y ago
Nope - you can see the salt in the stored hash. If you know what the 16 chars of salt are so you only have to search the 4 character search space to enumerate all possible 4 char passwords. (it makes the number of rainbow tables you would have considered pre-calculating ~10 years ago a much more difficult task, but that's not how passwords are attacked any more)
- nerdy 11y ago
  So the "fast attack" uses the GPUs to compute on-demand, not using precomputed values?
  - bigiain 11y ago
    Yep - plus a specialised password dictionary and bunch of heuristics to permute or mutate that dictionary. OCLHashcat can top a billion hashes a second for poorly chosen password hash algorithms - attackers just keep lists of every password exposure they coma across (starting with the 32 million passwords RockYou exposed to the world) - then combine that and a regular "common word/name/petname" dictionary, and try adding leading or trailing numbers, swapping numbers for letters l337-5p3ak style. One PC with some fairly recent video cards in it could blast through every dictionary word, a few thousand common names - and mutate all of those with common capitalisation and punctuation (you _did_ capitalise the first letter and add a fullstop or exclamation point at the end when the site forced you to use upper/lower and "special characters" right?), then they can add every combo of up to 4 training digits, anything that looks like a date, and they'll probably try using the domain name or bits of it if they know where the list came from (is your email password fluffycat-gmail or GoKitehawks2015-hotmail ?). If you're unlucky enough to be in a password list disclosure from a site still using MD5 (salted or not) that's literally a few _seconds_ work. Salted SHA might up that to minutes per password. Unless the site used bcrypt/scrypt/pbkdf2, if your password was in any way "predictable", someone with a box with 4 modern video cards (or stolen AWS credentials) will have found all the "low hanging fruit" passwords in hours or days.

by nerdy 11y ago

bigiain 11y ago

Nope - you can see the salt in the stored hash. If you know what the 16 chars of salt are so you only have to search the 4 character search space to enumerate all possible 4 char passwords. (it makes the number of rainbow tables you would have considered pre-calculating ~10 years ago a much more difficult task, but that's not how passwords are attacked any more)

nerdy 11y ago
So the "fast attack" uses the GPUs to compute on-demand, not using precomputed values?
- bigiain 11y ago
  Yep - plus a specialised password dictionary and bunch of heuristics to permute or mutate that dictionary. OCLHashcat can top a billion hashes a second for poorly chosen password hash algorithms - attackers just keep lists of every password exposure they coma across (starting with the 32 million passwords RockYou exposed to the world) - then combine that and a regular "common word/name/petname" dictionary, and try adding leading or trailing numbers, swapping numbers for letters l337-5p3ak style. One PC with some fairly recent video cards in it could blast through every dictionary word, a few thousand common names - and mutate all of those with common capitalisation and punctuation (you _did_ capitalise the first letter and add a fullstop or exclamation point at the end when the site forced you to use upper/lower and "special characters" right?), then they can add every combo of up to 4 training digits, anything that looks like a date, and they'll probably try using the domain name or bits of it if they know where the list came from (is your email password fluffycat-gmail or GoKitehawks2015-hotmail ?). If you're unlucky enough to be in a password list disclosure from a site still using MD5 (salted or not) that's literally a few _seconds_ work. Salted SHA might up that to minutes per password. Unless the site used bcrypt/scrypt/pbkdf2, if your password was in any way "predictable", someone with a box with 4 modern video cards (or stolen AWS credentials) will have found all the "low hanging fruit" passwords in hours or days.