4 ms·
>Fingerprints are usernames, not passwords. Here is an excellent (and timeless) post on this fact No, that is complete absolute shit post that isn't even self
by xoa 5y ago
>Fingerprints are usernames, not passwords. Here is an excellent (and timeless) post on this fact
No, that is complete absolute shit post that isn't even self coherent. Like, it literally whines about needing something that can be "independently chosen, changed, and rotated", which obviously describes usernames so obviously biometrics can't possibly be usernames by that very post! Why is this dumb meme so fucking persistent? Fingerprints are one of many biometrics. They aren't usernames, which aren't an authentication factor at all. They aren't passwords. They aren't tokens. They are their own thing. They have their own pluses and minuses as part of a comprehensive response to a given threat scenario. That's it. Trying to shoehorn them into something else is the same as trying to shoehorn everything into a car analogy.
All security exists solely in the context of an equation of threat scenario (the word "threat" doesn't even appear in that post), defender vs attacker resources and the value of what is being defended. Real security must work for actual real humans too. For example, rotating passwords every day/week/month is "secure" except that it's also a huge PITA or even outright impossible for many humans and defending against what should be a non-existent threat scenario anyway. So the obvious and inevitable result is that everyone starts to use crappy passwords, write them all down on sticky notes and text files and such everywhere, or both. That is not the fault of the users, it's the fault of a shitty system.
Another word that doesn't appear in that post? "Camera". Biometrics is an enormously rich potential field, fingerprints are about the worst lowest hanging fruit and in no way represent everything particularly as we use more and more wearables (there are bits of entropy to be found in your body's cardiac cycle for example). But even for fingerprints, which is really lower resource for attackers: getting a reproducing a fingerprint, or having AI go through every single networked look-down camera for the obvious obvious pattern of a human pulling out a slab of screen and then entering a PIN or passcode into it then recording that? Are people expected to never ever unlock a device anywhere but a physically secure area? Because see above, that is not realistic for real humans and thus a worthless security response.
As is usually the case, the best answer is hybrid, with multiple levels of factor usage to try to combine the strengths of each. And indeed that is the way things are going.
Edit to add: And if I sound irritated about this I am. This is the same kind of user hostile shallow anti-security thinking that brought us things like "security" questions, password rotation policies, lengthy and baroque "must contain 2 caps 1 number 3 special characters but not those special characters and cannot START with a number" password policies, etc. All of which add aggravation and failure points to no good end. Bad security practices affect our entire industry to the detriment of us all, but "bad security" isn't just a technical thing it's a human UX thing.
- mrtranscendence 5y ago> lengthy [...] password policies Bizarrely, my organization limits passwords to a length of 12 characters or shorter. I agree with you, I don't want a password the size of a paragraph, but c'mon... 12 characters?
- xoa 5y agoI think you misread me, or I didn't communicate clearly. By "lengthy" I was referring to the policy, not password length. Indeed max password length itself is another common bit of foolishness, for sanity reasons arguably it shouldn't be infinite but ~150 characters should be fine so that if people want to have a long diceware passphrase that's fine. To the extent passwords are used at all it should be exclusively as input to a KDF or adaptive-hash anyway so storage-side it should all be normalized regardless of input length.
- mrtranscendence 5y agoAh, gotcha, sorry. "Lengthy (password policies)", not "(lengthy password) policies". I wouldn't call the policies themselves particularly lengthy, though we do have multiple systems with different policies for which we're supposed to use the same password, so there's that -- it's possible to set a password in one place that can't be set in the other. (Would something bad happen if they weren't in sync? I can't see how, other than it wouldn't be clear half the time which password to use.)
- xoa 5y agoSorry for not being clearer. Really though, the only "password policy" should be "no password reuse/dictionary" (check it against haveibeenpwned.com or the like, there is a nice API), and some minimum decent length. Preferably with a decent user friendly generator option for default suggestions too, and password manager friendly. It's probably not the weakest link at that point. "Multiple systems with different policies for which we're supposed to use the same password" seems like it should just be SSO? But I recognize in reality when using archaic systems at businesses with no budget sometimes hacks are just the best that can be done, and that's how it is. I mean, obviously best of all is no shared password, use proper key via hardware token instead and the password/PIN or (gasp :)) biometrics is purely something the user uses to activate the token. Unfortunately it'll probably be awhile until we get there. But the general use of baroque password policies, particular when interfacing with the general public, is still an anti-feature for security which has finally started to fade away.