Live data from Hacker News

Beyond passwords: New tools to identify humans

bits.blogs.nytimes.com

21–23 of 23 posts

Re: Beyond passwords: New tools to identify humans

#21
post #20

In my opinion there are 2 main issues with Biometrics: - The first is that it may identify you with precision but it can not tell your will, so basically, you can force someone to use the sensor. - The second issue is that it does not allow you to give authorization to a third party in exceptional situations.

and third: You can't change whatever attribute of yours is being measured when the data leaks out by accident. And fourth: You always share the same attribute with every service identifying you. We've already seen password databases being compromised. If that happens, you change your password and move on. If biometric data leaks out, you a) can't change your attribute and b) you will have shared that attribute with o…

plus your biometric data may change over time. what do you do for those unfortunate edge cases where someone has an eye infection or a skin rash?

plus it costs a lot more to install, maintain and verify.

plus you can 'steal' someones biometric data without their knowledge (eg. from coffee cups, pictures etc.) without any real defense short of never opening your eyes or touching anything without gloves on.

plus they rely on heuristics and this lack of determinism means that either there will be false positives or false negatives. both cause issues.

... but at least it looks cool. biometrics does have that going for it.

Re: Beyond passwords: New tools to identify humans

#22
post #20

In my opinion there are 2 main issues with Biometrics: - The first is that it may identify you with precision but it can not tell your will, so basically, you can force someone to use the sensor. - The second issue is that it does not allow you to give authorization to a third party in exceptional situations.

and third: You can't change whatever attribute of yours is being measured when the data leaks out by accident. And fourth: You always share the same attribute with every service identifying you. We've already seen password databases being compromised. If that happens, you change your password and move on. If biometric data leaks out, you a) can't change your attribute and b) you will have shared that attribute with o…

To me, that seems to suggest that biometrics aren't really passwords (which are private, and authenticate you), so much as they are usernames (which are public, and identify you.)

Just like a username, you can know/have someone else's thumbprint, and "type it in" for the identification phase of login. And just like a username, you need a password in addition, in order to test that you're really who you say you are.

The main difference is that biometrics, unlike usernames, are unique to a person (two people in separate places can't decide to use the same thumbprint without knowing about each-other), so there's no equivalent to trying to log into a machine by using the top 50 most common username+password combinations. In other words, biometrics can always be compromised by HUMINT, and sometimes, given a flawed implementation, by SIGINT--but they're completely immune to MASINT (http://en.wikipedia.org/wiki/Measurement_and_signature_intel...).

Another way to think about it: fingerprints and the like are pretty much UUIDs you're assigned at birth. Just like a textual UUID, anyone who can read it can replicate it or retransmit it fairly easily. There's no security there. What there is, is collision-prevention. Nobody will be using a UUID you just generated unless they're trying to collide with it, specifically. If they just generate their own UUIDs, they'll never pick yours by coincidence. Much easier to detect and flag malicious logins once you can be sure that there's no chance of accidental collisions (e.g., someone just typing in their username-on-some-other-machine that happens to be already owned by someone else on this one.)

Re: Beyond passwords: New tools to identify humans

#23
post #15

> last month, it was discovered that even passwords as long as 55 characters could be broken. I stopped reading there. This is utter bullshit. If 55 is doable, then 16 should be peanuts (like, many orders of magnitude easier). Here crack this md5 hash for me: 2419219bcd13f7a6dfebeca7cf94108e. We all know md5 is broken, right? By the way, this was generated by using: head -c 16 /dev/urandom | md5sum Of course we can c…

Some of the characters you used are not allowed by many password systems. This decreases the entropy quite a bit, but still not enough to make your point invalid.
Post reply on HN