News for hackers. Stuff that matters.
Re:Ron was Wrong, Whit is Right by mindslight on Tuesday February 14, 2012 @04:43PM (#3591918)
Utterly simple and completely obvious in retrospect -> game changing.

Maybe it's a good idea to tag a public key with the software/version used to create it. Debian fiasco keys are easy to pick out - the entropy was so low that the bad keys can be enumerated. This new attack relies on any small weakness in the distribution of primes, and it seems quite hard to track down responsible versions of software.

I wonder how many of these shared factors aren't the actual value of p or q chosen at generation time (primality tests are probabilistic).

Re:Ron was Wrong, Whit is Right by marshray on Tuesday February 14, 2012 @06:16PM (#3592246)
> Maybe it's a good idea to tag a public key with the software/version used to create it.

I'm a full-disclosure guy at heart, but why would I want to advertise to the attacker my potential weaknesses?

Wouldn't I find my own personal selfish security is maximized when I falsify that field when generating my keys?

Re:Ron was Wrong, Whit is Right by djcapelis on Tuesday February 14, 2012 @06:58PM (#3592355)
> Wouldn't I find my own personal selfish security is maximized when I falsify that field when generating my keys?

It depends if you're the type of user who would like to rely on hearing themselves that their keys are vulnerable because of the software used or if you're the type of user who would like to be informed by an automated script that their keys are vulnerable.

I am guessing there are many in the latter category and a few in the former category but at the same time most of the people who decide these things will live in the former.

As for you, marshray, personally, I'm going to go with yes. But does that really make things worse for those who opt in to the potential for notifications? I think you can make your security tradeoff and others can make theirs and both categories of users will end up happy.

Re:Ron was Wrong, Whit is Right by mindslight on Tuesday February 14, 2012 @06:31PM (#3592287)
Hence the 'maybe'. But what information would actually be leaked by knowing the generation software (assuming it's decoupled from the server software)? It seems if a certain implementation's keys are weak, one could simply try to exploit that over all keys. I don't need to know you're running Debian before I go scanning for those particular weak keys.
Re:Ron was Wrong, Whit is Right by dekz on Tuesday February 14, 2012 @06:14PM (#3592241)
The responsible version of software are armchair cryptographers or average programmers trying to get things to 'work'. I bet this all results simply using OpenSSL with a SAMPLE to seed a PRNG or a bad seed of 0's.

To see if it could potentially be a afflicted version of a toolkit, a scatter plot of dates and frequencies of these shared primes would provide a little more information.

Re:Ron was Wrong, Whit is Right by mindslight on Tuesday February 14, 2012 @06:45PM (#3592327)
Bad system entropy is probably causing a lot of the duplicate moduli (complete keys), but not the cases where only one factor is shared. To me, that seems like an implementation weakness skewing the distribution of (probable) primes.