Peter Cordes wrote: > that has the same hash as the file you're trying to spoof. (you don't get > the advantage of the "birthday paradox" (29 people in a room -> 50% chance > at least one pair has the same birthday) because the other member of the > pair is already picked: it is the md5 hash of the original file.
a) I seem to recall it's 23 people. FWIW. Can't be bothered to check, though. b) The other member of the pair is not necessarily already picked; you might be happy to match any one of a given number of base "nasty" files you had with any one of the signed packages that someone might want to download. Not sure that this would help significantly enough though ;) > Besides, I'm almost certain that no system cracker would bother to get the > md5 digests the same on all the files they changed, since most people don't > check. I'd say you would be able to find changed files > 99% of the time, > and either you wouldn't find any changed, which would mean a _very_ > sophisticated cracker, or you would find every file she changed. (the > chance of one changed file randomly staying unchanged is 1/(2^128)) In this case only one file needs to be changed - the .deb for a package (*any* package will do - I'm sure the preinst could do sufficiently nasty stuff before you realised it wasn't the package you'd thought - or even without you ever realising). I'm not saying it's necessarily feasible, just that there are a few invalid assumptions flying around. Cheers, Nick
