3 ms·
I'm of two minds about password stuff like that - while requiring long passwords "adds friction", it's also clear that sub-eight-char passwords are trivially re
by bigiain 13y ago
I'm of two minds about password stuff like that - while requiring long passwords "adds friction", it's also clear that sub-eight-char passwords are trivially reversible in the face of many common password hashing schemes, and even you write your own or use a framework with bcrypt/scrypt/PBKDF2 password storage, short or dictionary passwords are easily vulnerable if the password hashes are exposed.
As a site owner I'd like to be able to look people (or a judge) in the eye and say "I followed realistic best practices in protecting your password - I put technical measures in place to ensure the password storage wasn't so weak they could be expected to fall in under an hour to a curious journalist running HashCat on has laptop as part of a story he's writing."
(Story time: I once signed up for this interesting sounding new website/service "just to see what it was". I used my usual "throw away" password. The same email and password I'd signed up to PerlMonks with. Several years later, that new website had become a regular part of my online work and social life, and a place where I cared about my reputation. I hadn't - back then - had a process in place to remind myself where I was using weak or reused passwords, so when the PerlMonks site got hacked (with their plain-text password storage!), all of a sudden my friends/colleagues/clients started getting Acai Berry Spam from my Twitter account…)
- rcfox 13y agoI don't know a whole lot about this area, but why would dictionary passwords be more vulnerable when considering a password hash? Couldn't you just salt the hashes with enough characters to defeat rainbow tables?
- dylangs1030 13y agoYou're right - dictionary attacks aren't more vulnerable using a password hash. As a rule, a hash will obviously only add complexity. Salting the hash in addition to that will prevent a list of common words, and corresponding hashes ("dictionary attack") from working because of the nonsense salt value. However, the real value of salting a hashed password is in defeating multiple password cracks. The salt is like a nonce added to a each individual password so that if you crack one, you can't formulaically crack them all. While en masse this is tedious, applied to a single password it's meaningless because the processing time for brute-forcing a salted hash and a non-salted hash will be trivial. It's mainly useful because you can't "recycle" computational power applied to one cracked password for another, assuming you have e.g. a database in front of you. To give a quick explanation of why a salt won't protect a single password, consider it this way - salting passwords allows each hash to come out differently, such that if you try to crack a list of passwords using a single algorithm it won't work because the salt randomizes each one. Repeating the computational steps for cracking one password will be null and void for finding additional passwords if they're salted - you'll have to start over.
- bigiain 13y agoThere's a lot of modern password hash attacks like HashCat that don't use rainbow tables. There just aren't enough words in most people's vocabulary for a "dictionary word" to provide enough bits. Salts help, but HashCat on a decent GPU will test over 2million password/sec for salted MD5 hashes using phpass (the current WordPress password storage). That'll rip through the whole RockYou password list (32m words) in 15 seconds. Anecdotal evidence suggests something like 25% of typical passwords from publicly available dumped hashes fall to the RockYou list. Salts mean I ned to run each hash individually, but a 25% chance of revealing each hash's password in ~15secs isn't much of a challenge. (Oh, and the "journalist with a laptop" comment was about this: http://arstechnica.com/security/2013/03/how-i-became-a-password-cracker/ http://arstechnica.com/security/2013/03/how-i-became-a-passw... not a bad read if you're curious. See page 3, almost 5000 passwords out of a list of 17000 unsalted MD5 hashes in _one_ minute - on a laptop without using a GPU…)
- anigbrowl 13y agoI totally get where you're coming from and applaud your responsibility, but as a user I'd kinda prefer to asked 'are you sure about this laughably insecure password?' and click a buton saying 'yes, and it's my problem,' for most things short of email, banking and suchlike. Basically, if you don't collect my credit card details, then chances are I don't care about the security - if I do, I'll switch it later. That said, I signed up anyway and hope the service does well. I just think this is an ideal candidate for OAuth or some other low-friction thing, because it's supposed to be fun and enjoyable - and I'm a musician, so I'm more professionally interested than most users would be.
- bigiain 13y agoI've had this slowly turning over in my head all day – how about a process that notices "insecure" passwords, and responds with "Hey, looks like you don't _really_ want to secure this right now – how about we just set a cookie so you can come back and see 'your' stuff, and if you want a proper account later you can set a strong password then?"? (Oh, and on the "I'll switch it later" comment – do you have a plan in place to enforce that? For a while I'd register throwaway accounts with a universal-low-grade-reused password, then search for that password every month in 1Password and decide to change/keep the password, or to delete the login. But it's less effort to just let 1Password generate good passwords when I register rather than deal with bad passwords on a recurring monthly basis... (And now I just get to rant about sites that won't accept random 25 char upper/lower/digit/symbol passwords…))
- anigbrowl 13y agoI think that'd be a great idea - you could even use it to flag the user as 'uncommited' and get better analytics, then when they did something more involved you could pop up and suggest some security is needed.