It's no joke:
http://it.slashdot.org/it/05/11/15/2...tid=93&tid=228

http://www.stachliu.com.nyud.net:8090/md5coll.c
http://www.stachliu.com.nyud.net:8090/md4coll.c

According to what I've read it doesn't, in practice, make it any easier or faster to reverse (ie. crack) password hashes such as are used in Linux's /etc/shadow. Still it'd be good to see distros getting away from it as a default ASAP and moving to something like SHA-256.

You still can't reverse a hash, that remains impossible. All this does is generate a string that will generate the same hash as another string, at least that's my understanding.

Its not impossible - it only takes a few days or hours with a modern computer to reverse an MD5 hash, and that was regardless of the recent collision discoveries. You'd hope with SHA-256 or something like that it'd be at least impractical to reverse it - ie. it'd take years or centuries or more.

Yeah that's the impression I got as well. It's a real danger because it means that you can generate, for example, a trojaned ISO file that has the same MD5 sum as the real one.

(the slashdot post has very misleading information)

What the algorithm does is not reverse a hash, not even (still) finding M2 such that H(M1) = X = H(M2), but finding both M1 and M2, the so-called birthday attack. A hash is theoretically secure if the birthday attack is approached by brute-force, but now there's an algorithm to do so.

I still don't know how much it affects shadow hashes. Salts only protects us from rainbow tables (that is, a collection of hashes of known passwords). Fortunately, recent distros have crypt-blowfish. You may configure it with /etc/login.conf. Just add / change the line to: ":passwd_format=blf:\"

What distro is that on? I've just had a look on both Mandriva2006 and Centos4 and couldn't see anything like that. I think the blowfish stuff needs to be setup and some packages have it built in for it to work. Suse offers the option of blowfish passwords doesn't it?

Try with PAM in /etc/pam.d/system-auth and maybe /etc/libuser.conf

Have you tried the C program for that published md5 weakness? Could you found out a md5 collision?
I couldn’t .... and I let it working for 2 days on a P3 1Gz processor...