4 ms·

I'm by no means and expert is this field, but I thought at some point I had heard that using words like the xkcd comic were actually less secure, I thought what

I'm by no means and expert is this field, but I thought at some point I had heard that using words like the xkcd comic were actually less secure, I thought what I heard was that there is a type of dictionary search that can be more efficient in cracking those "all words" passwords (as in, you don't really have 44 bits of entropy). Again, I've got no source, and am not very knowledgeable in this space, so someone correct me.

heinrich5991 8y ago
Assuming you pull these four words randomly (not-human-random, actually random) out of a dictionary of at least 2048 words, and separate them by a space, you have a password of (at least) 44 bits of entropy. There's no way a "dictionary" attack can reduce the possible password space to less than 244 possibilities.
- astura 8y ago
  You can use Dicewear to generate these passwords: http://world.std.com/~reinhold/diceware.html http://world.std.com/~reinhold/diceware.html They recommend a minimum of six words though.
- ImprovedSilence 8y ago
  Some quick googling gave me this:https://paul.reviews/passwords-why-using-3-random-words-is-a-really-bad-idea/ https://paul.reviews/passwords-why-using-3-random-words-is-a... (also good stuff here: https://security.stackexchange.com/questions/151165/is-randomly-generating-passwords-from-an-assortment-of-dictionary-words-cryptogr https://security.stackexchange.com/questions/151165/is-rando...) It seems the consensus is to use 5-6 words, and following the xkcd trick of 4 may not be enough.
  - tashbarg 8y ago
    Your first link uses a few assumptions that are very good security practice but may confuse the unexperienced reader: It assumes that the attacker has complete knowledge of the password generation method. This is good security practice and provides you with a worst case boundary. In reality, though, an attacker seldomly has that advantage. Before an attacker spends x hours/days/weeks to crack pure word-based passwords, they will spend time to crack "passw0rd". If you remove the advantage of password generation method knowledge, all numbers in this article are very different. The reader should know about that! It assumes that whoever is storing the password may do so badly. It even states "assume the site stores our credentials in the weakest possible way". Which is a dangerous assumption since the weakest possible way would be plaintext and then the whole article would be moot. So, obviously we exclude plaintext. The article goes with simple, single md5 hashes instead. While some kind of worst case, it's pretty unrealistic nowadays that someone makes an effort not to store passwords in plaintext and then fails so miserably in googling how to do so. This worst case is probably chosen to have easier and more impressive cracking numbers. The reader should be aware of this. It assumes that the attacker obtains the password database. Again, good security practice and a worst case scenario. But still not exactly 100% realistic. If you argue with this assumption, the reader should be aware of that. In essence, this article proves that the "3 word method" is not secure enough when absolutely everyone uses this exact same method (with knowledge of the exact same words) with a service who incompetently stores passwords and got its password database stolen. While that is true, the advice it gives "Don't use words in passwords. Ever." is just another example of great oversimplification that is harmful in the end. Instead of bashing methods for being not secure enough (whatever that means), we should provide users with practical methods to come up with usable passwords that are reasonably secure for the service in question.

heinrich5991 8y ago

Assuming you pull these four words randomly (not-human-random, actually random) out of a dictionary of at least 2048 words, and separate them by a space, you have a password of (at least) 44 bits of entropy. There's no way a "dictionary" attack can reduce the possible password space to less than 244 possibilities.

astura 8y ago
You can use Dicewear to generate these passwords: http://world.std.com/~reinhold/diceware.html http://world.std.com/~reinhold/diceware.html They recommend a minimum of six words though.
ImprovedSilence 8y ago
Some quick googling gave me this:https://paul.reviews/passwords-why-using-3-random-words-is-a-really-bad-idea/ https://paul.reviews/passwords-why-using-3-random-words-is-a... (also good stuff here: https://security.stackexchange.com/questions/151165/is-randomly-generating-passwords-from-an-assortment-of-dictionary-words-cryptogr https://security.stackexchange.com/questions/151165/is-rando...) It seems the consensus is to use 5-6 words, and following the xkcd trick of 4 may not be enough.
- tashbarg 8y ago
  Your first link uses a few assumptions that are very good security practice but may confuse the unexperienced reader: It assumes that the attacker has complete knowledge of the password generation method. This is good security practice and provides you with a worst case boundary. In reality, though, an attacker seldomly has that advantage. Before an attacker spends x hours/days/weeks to crack pure word-based passwords, they will spend time to crack "passw0rd". If you remove the advantage of password generation method knowledge, all numbers in this article are very different. The reader should know about that! It assumes that whoever is storing the password may do so badly. It even states "assume the site stores our credentials in the weakest possible way". Which is a dangerous assumption since the weakest possible way would be plaintext and then the whole article would be moot. So, obviously we exclude plaintext. The article goes with simple, single md5 hashes instead. While some kind of worst case, it's pretty unrealistic nowadays that someone makes an effort not to store passwords in plaintext and then fails so miserably in googling how to do so. This worst case is probably chosen to have easier and more impressive cracking numbers. The reader should be aware of this. It assumes that the attacker obtains the password database. Again, good security practice and a worst case scenario. But still not exactly 100% realistic. If you argue with this assumption, the reader should be aware of that. In essence, this article proves that the "3 word method" is not secure enough when absolutely everyone uses this exact same method (with knowledge of the exact same words) with a service who incompetently stores passwords and got its password database stolen. While that is true, the advice it gives "Don't use words in passwords. Ever." is just another example of great oversimplification that is harmful in the end. Instead of bashing methods for being not secure enough (whatever that means), we should provide users with practical methods to come up with usable passwords that are reasonably secure for the service in question.