On this topic, I might actually say that a concatenation of two hashes
is very secure if the two hashes are sufficiently different.

Although MD5 and SHA1 are reasonably similar, let's suppose for a
moment that they use entirely different mechanisms. If this were so
and the crack time for MD5 was 2**50, for SHA1 2**65, then the crack
time for CONCAT(MD5(string), SHA1(string)) would be about 2**115.
This is because you need to find something that colides for both.
This is *far* more difficult and therefore far more collision
resistant. Unfortunately this requires 288 bits - far more than each
of the old hashes. Still, that much memory doesn't tend to be a
problem ;).

Regards,

Michael Cordover

--
http://mine.mjec.net/

[ reply ]