One caveat: Don’t do this if your phone is not set up to be able to receive texts. Yes, I know, that sounds too obvious to mention, but it happened to my uncle this past weekend.
Again, a caveat: Use random words. A sentence is not random words. Even a nonsense sentence is not random words. Try flipping open a dictionary to a random page, then close your eyes and poke at the page, and pick whichever word you’re touching (and not the nicer-sounding word right next to it, because that’s also not random), and repeat about five times. Actually, that’s still not as random as it should be, because you’re more likely to flip to a page in the middle of the dictionary.
The former, by a very large margin. The only practical way to make something that can’t be cracked is to prevent it from communicating with the outside world at all, and that’s useless for a password server.
I keep a password manager on my phone, synced (for backup purposes) to my computer at home. I have my various work accounts in there, as well as my various personal ones. So I need to
have my phone with me
remember the phone unlock password
remember the password manager password
It works quite well, though it is a bit tedious typing some of the more arcane passwords by hand.
I use Keepass which is the only common open source password manager. You’re right that I have to implicitly trust it since I don’t have the knowledge or ability to review the code myself.
Still, given the number of developers who (1) use it themselves (2) make the many plugins, extensions, and ports, and (3) the tin-foil hat brigade who has been using KeePass for 10 years, I feel reasonably confident that it doesn’t have any blatant security risks. At least, none of the ExportPw(evil.mountain.fortress1.ru) variety.
However, even if KeePass is compromised, not all of my eggs are in that basket. For my most critical accounts (email, banking, dropbox, etc) I have two factor authentication set up. To log in from a computer or IP address that’s not familiar to Google or whoever else, I need to enter a code sent by text to my phone. If my passwords are compromised, I am still secure as long as I have my phone in my possession.
I’ve got my password database stored on Dropbox, synced to all of my computers and my phone. I can open it on my phone and, if necessary, use it to manually enter passwords.
I’ve got a strong, memorized password for Dropbox, and there is strong encryption on the KeePass database. I am as confident as anyone can be that there is negligible risk of someone obtaining my password database and success brute-forcing its encryption.
ETA: I see most of my points were already made! That’s what I get for leaving to shovel the snow halfway through posting.
Also, set up those backup codes, especially if you’re going to leave the country and are planning on using wifi and email to keep in touch with everyone. :smack:
I suspect that if someone, say, obtained every document that I have written and spent a lot of effort analyzing my personal dictionary and grammar structures, they would have a better chance of brute forcing some of my passphrases. Even though I picked random words, I did it until I had a string that was vaguely meaningful and memorable to me.
But that kind of attack would require physical access to my computer, and if that is possible there are all sorts of ways that I could be completely screwed.
At some point in the paranoia spiral, you have to trust that your wife, house guests, and coworkers aren’t putting a keylogger on every device you own. And if the NSA is after me personally, I’m sure they can figure out a way to get my [del]precious bodily fluids[/del] personal data.
But yes, normally accounts should be locked, IP’s banned, and requests throttled after several wrong guesses. Not everyone does this right though, especially smaller companies.
If someone’s put a keylogger on your computer, then you’re screwed no matter how good your password is. There’s no sense in worrying about something that you can’t do anything about. But there’s a lot of room for someone to get ahold of your KeyPass database without being able to install a keylogger. So, yes, it really is a very good and prudent idea to make your KeyPass password as secure as possible.
If someone is targeting you, you have a greater chance of getting screwed. But for the most part when people steal the hash list they are going to hash the most common passwords and see what matches. If the hackers get 10,000 hits on “password” they aren’t going to bother with your reasonably strong one.
If I were a celebrity or something I’d choose a real strong one - and my real passwords are a lot stronger than that example, which was a spinoff from Terr’s story.
When I was at Bell Labs the performance review packet for managers came with a note from Brian Kernighan warning everyone that your stuff isn’t as secure as you think it is. Though he was talking about hacked C compilers.
The crucial factor here is what is the incremental benefit to the hacker of getting into an account with a reasonably strong password? If the account is owned by a star with dirty pictures, pretty high. If it is just a credit card number, pretty low, since these are sold in bulk.
If hashing the standard list of bad passwords plus the dictionary with perhaps some simple substitutions will give you 60% of the passwords, you’re probably going to stop. There is also an incentive for speed, since if the break-in is detected the cards might go on a watch list which will reduce their value. (And increase the risk to those using them, but the hacker probably won’t care about that.)
Picking two unrelated words (and pretty rose isn’t good) makes the password a lot more secure. And easier to remember.
Yeah. The XKCD approach is really about having a provable amount of entropy in the result. Rolling dice absolutely guarantees this (when done correctly).
Any kind of cleverness with substitutions does not make the same guarantee. An O<=>0 substitution contains less than a bit of entropy, at least if you haven’t done it with a coin flip.
In fact, you can’t really guarantee that substitutions have any finite amount of entropy at all. What do people do if they try “password” at a site and get the response “password must contain at least one capital and one number”? Guaranteed that almost all of them will be “Password1”. I’ve done it myself (with a slightly better base password, mind you)! They didn’t get several additional bits of entropy as one might naively expect; they got 0.1 bits or less.
So yeah, the only way to get guaranteed entropy is via a truly random process. I rolled a die 25 times to get my last long passphrase.