On Mon, May 07, 2007 at 11:57:50AM +0200, Martin Schr?der wrote: > 2007/5/7, Adam Hawes <[EMAIL PROTECTED]>: > >MD5 is proven weak. It's possible to take almost any file and its > >MD5 then create an identically sized file with the same hash in a > >reasonable time. This can be used to pass out an arbitrary CD > >image that completely trashes the contents of your hard disk. It > >doesn't even need to be OpenBSD on the CD. > > Your mixing collision and preimage attacks. The former are possible, > the latter not. > > Still, it's certainly time to switch to something better. PGP comes to > mind... > > Best > Martin > >
Not specifically to you, Martin.. - Instead of writing silly emails about theoretical md5 attacks and wasting everyones time, how about sending a damn patch to tech@ that 'fixes' it? MD5 sums are meant to be used for verification of a downloaded file in case of transmit errors. If you own ftp.openbsd.org and upload trojaned binaries, how hard is it to update the damn checksums file? It's like rocket sience, yes!! Really hard! "But, but, but, i'm clever, i will use checksums from another server!!1!" Yes, of course, the only problem is that these other servers rsync in 2-8 hour intervals, which is a very tiny window to detect anything. Even if you do, it's highly questionable that you will be clever enough to ask yourself why they updated the filesets and run a bindiff on them to check if it is trojaned or a legitimate update. When was the last commit to any of these projects from you guys: http://netbsd-soc.sourceforge.net/projects/bpg/TODO http://openpgp.nominet.org.uk/cgi-bin/trac.cgi hmm? Btw, pgp requires a working web of trust, it's not secure just because you can sign something. Joe Cracker can easily generate a key with "Theo de Raadt <[EMAIL PROTECTED]>" and provides you with "signed" filesets. Who steps up to organise key signing parties, worldwide? SCNR, Tobias