I ask, can we do away with passwords entirely? Trying to defeat computers with human memory seems like the wrong approach.

SSH keys all day. They can be troublesome for even an "expert" to look after, so I guess there's no hope for your average idiot. Still, logging in without using a password is the bomb

Keys are great until you find a sudden need to log in from your wife's cellphone or a public terminal..

opsec 101 - don't use other people's devices / machines

Do you never, ever, use devices not in your full possession?

Isn't that what you do?

If you have deals with many strangers do you not expect to get your hands dirty in some way?

I definitely don't store my SSH keys on stranger computers. I however used passwords on my relatives or customers devices on numerous occasions. Don't pretend you never did, I'm not going to believe you.

Sure, I've used other machines in the past. I prefer not to though in most circumstances and assume that any account / password that I use on said machine is compromised from then on.

This was something I changed when I realized that a tax related government agency's machine I was using about 10 years ago man in the middled SSL connections to my online banking. Who knows what was logged and what the retention on that was? I don't have time or the inclination to ensure every relative's / customer's devices are secure, so I use my own or go without.

That should be soluble though. E.g. standardized smartcards that cellphones and terminals know how to communicate with.

In theory yes, although with caveats. You'd need to match the degree of anti-tamper measures currently present in POS terminals to all consumer devices.

In practice, it would mean replacing most of the world's user-facing computer infrastructure.

You can get computers with smartcard readers from virtually every major manufacturer (except Apple, because Apple). The most frequent use case is government users who need to support the DOD Common Access Card (CAC) standard.

The card reader doesn't have to be any more tamper-resistant than the rest of the computer. On a public terminal, it should probably be built like an ATM's card slot (and the rest of the machine should be similarly armored), but on a personal computer it doesn't need to be any more robust than the keyboard a user would otherwise type passwords in with. The card-computer communication is all encrypted anyway, so even if it's built like any old USB peripheral it's still better than typing into a keyboard.

Key disclosure and key loss are things.

The first makes your public key useless for security. The second locks you out of everything.

So you've still got to have patterns-of-fraudulent-use detection, and some sort of token-reset mechanism.

Persistent data encryption is another option, but that compounds the key-loss problem in that you cannot access earlier data.

Yea this seems easy to set up. But, what about <insert generic website>? Like... HN.

I have an encrypted file with website|username|password on my laptop to allow for unique passwords across sites. But it's a pain, and definitely not something everyone will do.

SSL client certificates. If only the browser UX for them didn't suck.

We can certainly do without passwords for every stupid little website. Unless you're rightfully a central force in the user's life - say, their main bank - you really should have users log in through Google, or Facebook, or Twitter, or someone else who almost certainly can guard their passwords better than you can.

It's shameful that it's so much easier to find tutorials on how to store passwords "securely" (including several tutorials that tell you crazy insecure things, like storing with unsalted commodity hashes) than it is to find tutorials on how to integrate your brand new battling-fairies website game with OAuth for authentication.

Yes. Debian uses client-side X.509, OpenPGP and OpenSSH certificates for all service authentication (including websites).

That is just 2fac, isn't it?

Authy codes (as well as Google and Microsoft equivalents) can be the first factor, which would eliminate the need for passwords. Grandparent's question was about usage by groups and teams at the enterprise level, which Authy also supports.

I've often wondered if this is possible. I bet there's an academic paper somewhere about zero-knowledge multi-lookups or something.

This.

I think it's going to have to be legislated:

1. Mandate that all publicly disclosed passwords are collected in a compiled set.

2. Require that any Internet service vendor check against this set on password change revisions, and against different portions of the file, as maintainable, on each authentication.

Known passwords are rejected when set, and mandate changes when detected on login.

Asking a 76 year old mostly-computer-illiterate pereson to come up with a never-been-used-before password simply doesn't work.

This also means mandating password safes on all consumer-grade OSes. Apple already does, Microsoft and Android. For Linux, you might be semi- on your own, though most distros offer several options.

I'd also like to see 2FA via a keyfob generator rather than contact token (e.g., phone, email).

Coming up with appropriate password for ordinary user's is hard enough right now.lets imagine you are not being sarcastic-for the sake of argument-, then I think it is almost impossible to create good enough password, yet rememberable.

Lastpass does this. It will warn you when saving a password, and it also warns you when loading data for an already saved password.