This service selects the words for you guaranteeing the randomness, so the article you reference where people pick their own doesn’t compare.
In this system a dictionary attack can treat each word as a character. So the password strength is the same as a seven character password with a 7000+ character alphabet. This works out to being much stronger than a 20 character password using a 95 character alphabet.
Seven times 12.73 bits of entropy = only ~89 bits of entropy.
I think of ASCII characters having about 5 bits of entropy in practice, even though the 96 characters theoretically could have more than 6 bits; people tend to stick with mostly lowercase letters and numbers.
So your seven word password is about equivalent to a truly random 18 character ASCII string. I question the assertion that this would take years for the NSA to crack if they had access to the encrypted passwords through a tap on the internet (or even intranet) backbone.
(A) salted hashes means the rainbow table is useless since a new table has to be generated for each salt,(b) rainbow tables are impossibly large once you’re dealing with passwords this long. On the order of multiple exabytes.
While everyone was busy memorizing high-entropy passwords, the NSA quietly infected their keyboard firmware.
Since you wrote your firmware, I’m sure it didn’t take you long to spot!
The best you can do for password security is this:
Will this protect you from the NSA? Fuck no. If they want you, they will have you. Will it prevent you from being screwed if you get a keylogger on your computer? Nope. You are screwed, but that was always the case. It will protect you from a solid 99% of all attacks though.
The two three common attacks are:
A password locker and 2-step everything are, by far, the best defenses against all three of those things. Your passwords are all different and random so if someone steals on password they only have that one password, and you can make your password absurdly long and random, making them hard to brute force.
Seriously, use a password locker. If you are not doing this it is truly only a matter of when, not if. A reused password is a worthless password. As a bonus, it makes keeping track of your various accounts vastly easier. A password locker is worth it for the convenience alone.
All fun word Haiku
In the Twenty-First Century
Our Security is fine
Passwords should be passwords rather than pass codes. Leave the codes to the AI and the words to we users. Humans have a grand history of remembering lines and lines of poetry; Here it was not long ago we would all not have to know our own phone numbers. We can get back to using the memory parts of our brains to manage our security.
Useful article, thanks.
For a password manager I like Last Pass. Schneier has an open source manager that seems pretty good: https://www.schneier.com/passsafe.html
PROTIP: when copying and pasting passwords, mind your clipboard.
$1$CfoyagToyhL$ftgwtsiheAsgLitswdLitswdLitswdA
Wheee! Fun game! Make it harder though, that one only took me about 30 seconds.
Or hard drive firmware.
Indeed… I could just imagine that spooks wrote my personality. Then I could just pretend that I don’t exist. It’s probably easier than any of the alternatives.
+1 on the clipboard advice. I have some high value passwords I keep only in my head. For a lot of my website passwords I use a manager. KeePassX has been a good one for me on OsX.
When you’re calculating the entropy of a pass phrase of randomly chosen words, you treat each word as a unit. Four words randomly chosen from a set of 2048 words has the same entropy as four characters chosen from a set of 2048 characters.
The hash of a two-word passphrase will be completely unlike the hash for each of those two words alone. A multi-word passphrase would be no more vulnerable to cracking by means of a rainbow table, than would be a password consisting of a string of randomly generated characters, of an equal entropy.
The opening paragraph of the article says that it proves that passphrases of randomly selected words are not as secure as strings of randomly selected characters. But the research discussed is about how commonly used passphrases are not randomly selected. So the article does not prove what it claims to prove.
I’ve seen this mistake repeated frequently: criticisms of random passphrases, based on the vulnerability of nonrandom passphrases. The randomness is key.
The neat thing about random passphrases is that we’re surprisingly good at coming up with a meaning for passphrases after we generate them, making it relatively easy to remember them. But the critical thing is they’ve actually got to be generated randomly. And that means using dice, or a pseudo-random number generator on a computer. (And the “pseudo” is kinda pedantic, really, as the numbers generated by modern PRNGs are pretty thoroughly unpredictable.)
Diceware remains a good method.
Yep. I’m amazed at the number of sites that still send passwords, by email, plain text.
I find banks to be the worst when it comes to poor password requirements. Most don’t even allow special characters, just letters and numbers. And they usually require you to change the password every three to six months no matter what. Stupid.
The toughest rules I’ve ever seen were on usajobs,com. As if Osama himself is going to hack in there and apply me for some federal jobs. I suppose everybody thinks their site is the most crucial one… but still.
I think part of the problem is that at least when it comes to customers’ data many businesses care more about compliance than about real security. If following best practices from 1978 covers their ass sufficiently, then that will do.
I remember UT Austin being pretty strict, 8 years ago. No dictionary words – not even spelled backwards, nor (IIRC) with numerical substitutions.
I was hoping somebody would get it! Good eye!
