5 ms·
It's a usability nightmare. I played a bit with password crackers recently and to me it is still often completely counterintuitive which passwords are easy to c
by weinzierl 9y ago
It's a usability nightmare. I played a bit with password crackers recently and to me it is still often completely counterintuitive which passwords are easy to crack and which are not. Now good luck explaining that to your users without frustrating them. Take for example '_=$%=_$ _ !'. Why is that a bad password? Because it's in one of the common wordlists (yes, it has two spaces and no, I didn't make that up).
That's why I think there should be more focus on the point that users should never produce passwords themselves, but leave that job to a good random generator. This is also something I missed from Troy's otherwise excellent article. What is the point of using a password manager if you only use it to store your weak passwords?
I also believe the argument about passwords being memorable is a red herring. It is possible to remember a low number of sufficiently long truly random passwords. It takes effort and it is risky, but it is not impossible. Producing a good password without the help of a random generator on the other hand is simply impossible.
- lostcolony 9y agoI have a few sites where this is effectively what happens. They have really bad security to where triggering a password change has them send me a new -password-, not just a link to reset my password. And then it doesn't force me to change the password when logging in. Well, that actually works out (except that the password is too short to survive brute forcing), because I then log in with that, do my business, and move on. Next time, same thing happens, "forgot password". Which presents a much better authentication flow to my mind. Nevermind passwords at all; just focus on whether they own a particular email account. If the email account gets pwned, the attacker can likely log in with them anyway (and if that's an issue you need 2-factor). So rather than force a password at all, send them a verification email, and have that immediately give them a token. If/when the token expires, have the login window require their email address, and send them another link that, when clicked (in the next X amount of time) will log them in with another token. I know for me that would mean -every- account I have is suddenly 2-factor (since my email account is), they all have strong security (since my email has a long, complicated, hard to brute force password that I keep only in my head; it's the only one like that), and I'd stop having to rely on a password manager. Curious if there's any obvious flaws that I'm missing.
- onli 9y agoNo, you missed nothing, that scheme works. Add a bit of OAuth magic to support direct logins for gmail, and you basically described portier (https://portier.github.io/ https://portier.github.io/), a successor project for Mozillas Persona.
- lostcolony 9y agoThanks, I'll check that out!
- 0xffff2 9y ago>Curious if there's any obvious flaws that I'm missing. The obvious flaw is that email is not instantaneous. https://luxsci.com/blog/why-email-is-not-instantaneous-and-not-supposed-to-be.html https://luxsci.com/blog/why-email-is-not-instantaneous-and-n...
- madeofpalk 9y ago> It's a usability nightmare yeah, and here's the problem. Security is at complete odd with user friendliness. Two-factor auth is amazingly secure, until you've lost your second factor.
- scott00 9y agoIt's even more secure after you've lost your second factor.
- jmulho 9y ago> Producing a good password without the help of a random generator on the other hand is simply impossible You just produced one: pagpwthoargotohisi
- weinzierl 9y agoThis kind of hubris is ecactly the reason why most [1] passwords are cracked so easily[2]. The scheme you applied is well known. Making a wordlists from every sentence of the HN dump should be doable. And why stop there? Project Gutenberg, the whole Wikipedia and other Wikimedia projects like Wikiquote are just a download away. A project I'm currently working on is facilitating the Google Ngram database for password cracking. I already generated almost 500 GiB of wordlists but haven't measured their success rate yet. [1] Good wordlists have a success rate around 70% against the common leaks. Running one of these lists against a set of fast hashes takes less than an hour on a plain average business notebook. [2] Assuming a fast hash, possibly salted. Slow hashes like bcrypt, scrypt, argon2, etc. are another story.
- jmulho 9y agoThe Hubris System(R) provides adequate protection and is highly practical. The scheme is well known, but what attack against it outperforms an exhaustive search? How do you build a database of poetry I haven't written yet and will never publish? Frequency analysis may help, but many of my poems are about a yellow zebra named x-ray who plays the xylophone. If you come up with an attack that outperforms an exhaustive search, I will just write longer poems. You will never catch me. Password cracking is an arms race. The crackers can't win. The math is against them. Every character I add to my password makes their problem exponentially harder. thspapaihp tsiwkbwaaioaes hdybadopihwyawnp famhbmompaaayznxwptx iycuwaatoaesiwjwlpywncm pciaartccwtmiateciatmpmtpeh
- weinzierl 9y ago>The Hubris System(R) provides adequate protection and is highly practical. If the protection is adequate is up to you, but as your phrases get longer they quickly get impractical and memorizing and entering will get error prone. >The scheme is well known, but what attack against it outperforms an exhaustive search? Everything that is not truly random has a pattern that can be exploited to outperform exhaustive search. >How do you build a database of poetry I haven't written yet and will never publish? Frequency analysis may help, but many of my poems are about a yellow zebra named x-ray who plays the xylophone. If you come up with an attack that outperforms an exhaustive search, I will just write longer poems. Per position markov chains are one way. The point is, that as long as it isn't random there is a way to improve on simple brute force. >You will never catch me. Password cracking is an arms race. The crackers can't win. Cracking a single password is always hard but in a spear phishing attack there are often better ways anyway. Cracking passwords en mass is a different story and looking at the success rate for cracking passwords from leaks I'd say the crackers already won. There are also arguments why mass cracking gives an edge to the attacker [1]. >The math is against them. Every character I add to my password makes their problem exponentially harder. Longer passphrases quickly get impractical and as long as the character you add isn't random there is a way to guess it. >thspapaihp tsiwkbwaaioaes hdybadopihwyawnp famhbmompaaayznxwptx iycuwaatoaesiwjwlpywncm pciaartccwtmiateciatmpmtpeh 1. Good luck typing that without mistake. 2. Look closely on your passphrase, don't you see the patterns? [1] http://www.openwall.com/presentations/Passwords12-The-Future-Of-Hashing/ http://www.openwall.com/presentations/Passwords12-The-Future...