I think this is wrong. It claims that the problem in IDEA comes from the bias of the multiplication operation... but the multiply in IDEA is over the integers modulo 65537, with the 0 value representing 65536, not 0... this is in fact an unbiased operation. However I have not attempted to verify the rest of the logic.
Correct, multiplication by a non-zero residue modulo 65537 is a permutation of the non-zero residues. The central notion behind IDEA is to use this operation, which can be expressed expeditiously in terms of the unsigned 16x16->32 multiply, as a builtin wide S-Box. This is how IDEA manages to do strong encryption using only a few ordinary arithmetic and logical instructions with no necessity for table lookup. I didn't bother to read the rest of the rant. -- Sponsor the DES Analytic Crack Project http://www.cyberspace.org/~enoch/crakfaq.html