Search All of the Math Forum:
Views expressed in these public forums are not endorsed by
Drexel University or The Math Forum.
|
|
jdm
Posts:
16
Registered:
11/22/10
|
|
Almost perfect nonlinear permutations
Posted:
Jun 14, 2011 2:32 PM
|
|
I've been wrestling with a bit of a thorny problem lately, and if
anyone knows whether a particular result relating to APN permutations
has been proven it would help enormously.
Consider an APN bijection S operating on GF(2^n) (n >= 5). It is easy
to show that the first 2^{n-1} entries of the truth table of S cannot
all be < 2^{n-1}, since if they were, all input differences < 2^{n-1}
would result in output differences < 2^{n-1}, and this would mean
that, for instance, input difference 1 would require 2^{n-1} 2s to be
placed in row 1 of the difference distribution table of S in the
columns corresponding to nonzero output differences < 2^{n-1}. As
there are only 2^{n-1}-1 such columns, this would lead to a
contradiction.
Now, consider a generalisation of this problem. Where S is an APN
bijection operating on GF(2^n), is there any value m (3 <= m < n) such
that it is possible for the first 2^m entries of the truth table of S
all to be < 2^{m} (and hence for the first 2^m entries of the truth
table of S to define an APN bijection over GF(2^m))? If not, can
anyone point me towards a proof?
I have tried to prove that this is not possible for the case m=(n-2),
but was unable to do so. More generally, I believe that no m exists
such that this is possible, but have not been able to prove this.
Many thanks,
James McLaughlin.
|
|
|
|
