Thomas Seeling wrote in a message to Mike Bilow:

 MB>> It would surprise me if two distinct three-byte strings had 
 MB>> identical CRC-32, but I see no reason that it would be 
 MB>> mathematically impossible, and I can even see ways to construct 
 MB>> such strings if they exist.

 TS> A CRC32 can be falsificated if you are able to change 31 bits 
 TS> in the data for which the crc is calculated. There may be cases 
 TS> where fewer bits need to be changed, but in general it is 31 
 TS> bits.

Yes, that is fairly clear.

 TS> I think you need to search some time to construct two 24 bit 
 TS> numbers with the same CRC32 :-)

As I said, I am not sure if it is possible for two sequences much shorter
than the 32-bit CRC itself to give the same CRC-32, but I don't see any
reason why it would be mathematically impossible.  It is clearly not
possible for an arbitrary 24-bit sequence, but two such chosen 24-bit
sequences could exist.

The fastest way to answer this question would be a brute force search,
since 24 bits is not all that many.  If two such sequences could be found
to exist, it would probably become immediately obvious how to find more. 
The answer to this question has a lot to do with the particular generating
polynomial used for the CRC, which makes it a fairly complicated question.
 
-- Mike


---