In <MHEGIFHMACFNNIMMBACAIEPAHGAA(_dot_)sethg(_at_)GoodmanAssociates(_dot_)com>
"Seth Goodman" <sethg(_at_)GoodmanAssociates(_dot_)com> writes:
Since people are talking about something shorter and lighter weight than
SHA1 or MD5, and we are not trying to build a crypto system but merely have
an efficient hash that is not easily guessed, how about something like
CRC-16 or CRC-32 but with a random seed? Even if the CRC algorithm is fixed
and known, it's still very hard to invert and crack the random seed, as long
as you run more than a few non-zero bytes through it.
I'm almost certain that CRCs can be "quickly" cracked with a
reasonable number of sample SRS strings.
It may not be obvious to those that have looked at CRC algorithms, but
the math behind CRCs is basically just taking the remainder from a
division operation. The "division" is done in a funny mathmatical
system (a "field"), and if I recall correctly that field is the
polynomials over Z/2.
As a result, I think the Chinese Remainder Theorem can be used to
crack CRCs.
RC4 isn't that much slower than CRCs and, if you use a reasonable
sized IV, it is quite secure.
-wayne