Relationships Between Boolean Functions and Symmetric Groups
Relationships between Bo olean Functions and
Symmetric Groups
Chengxin Qu, Jennifer Seb erry, Josef Pieprzyk
Center for Computer Security Research
Scho ol of Information Technology and Computer Science
University of Wollongong
Email: cxq01,jennie,[email protected]
2 Background Abstract
2.1 Bo olean space and boolean func-
We study the relations b etween b o olean functions
tions
and symmetric groups. We consider elements of a
symmetric group as variable transformation op er-
The set of n-tuple vectors,
ators for b o olean functions. Bo olean function may
V = f =(a ; ;a ) j
n 1 n
b e xed or p ermuted by these op erators. Wegive
a 2 GF (2); i =1; ;ng;
i
some prop erties relating the symmetric group S
n
and b o olean functions on V .
n
is a boolean space if its arithmetic is in a Galois
n
eld. A b o olean space V contains 2 vectors.
n
Clearly, all the vectors in V are binary sequences.
n
A boolean function is de ned on V by the map-
n
ping
f (x): V ! V
n 1
where x is a variable vector in V .
1 Intro duction
n
There are several ways to represent a b o olean
function: bya p olynomial; by a binary sequence;
The values of a b o olean function for each vector
and bya ( 1; 1) sequence. Here we use the p oly-
n
in V form a binary sequence of length 2 called
n
nomial representation to discuss b o olean func-
a a
the trace of the function. The trace of a boolean
a
1 2
n
tions. Let x = x x x denote a monomial
n
1 2
function is widely used in communication systems
on V . Then a b o olean function on V is a linear
n n
such as DE S and S-b ox theory [1 , 2]. To pro-
combination of monomials
tect against cryptographic attacks b o olean func-
M
c x c =0 or1; (1) f (x)=
tions must satisfy some algebraic prop erties such
2V
n
as nonlinearity, balance, the propagation criteria
and correlation immunity. These are called cryp-
where the sign denotes b o olean addition (XOR).
tographic prop erties [6 , 8, 11 ]. In this pap er, we
For anytwo binary sequences and with the
use symmetric groups to study b o olean functions.
same length s, we de ne their multiplication ()
The transformation of variables, x ! x , is called
i j
and binary addition () as follows;
an op eration or a variable transformation op era-
tor. We consider elements in the symmetric group
= (a ;a ; ;a ) (b ;b ; ;b )
1 2 s 1 2 s
(2)
as a variable exchange op erators for b o olean func-
= (a b ;a b ; ;a b )
1 1 2 2 s s
tions. We study the conditions under which a
= (a ;a ; ;a ) (b ;b ; ;b ) b o olean function is xed or transformed by this
1 2 s 1 2 s
(3)
= (a b ;a b ; ;a b ): op eration.
1 1 2 2 s s 1
De nition 4 Let 0 k n. The function f (x) So and are still binary sequences. If f (x)
on V is k -th order correlation immune if the fol- corresp onds to the binary sequence and g (x) cor-
n
lowing equation resp onds to , then the functions f (x)g (x) and
f (x) g (x) corresp ond to formulae (2) and (3)
X
f (x) x
( 1) =0; for 1 wt( ) k;
resp ectively.
x2V
n
We call the numb er of 1s in a binary sequence,
, its Hamming weight that is denoted by wt( ).
is satis ed, where wt( ) is the Hamming weight of
Avector in V is a binary sequence with length n
a vector 2 V .
n
n
and the values of a b o olean function for eachvec-
n
tor in V also form a length 2 binary sequence
n
2.2 Symmetric group
that we call the trace of the function. For anytwo
functions f (x) and g (x), their Hamming distance
For an n-tuple vector, = (a ;a ; ;a ) 2 V ,
1 2 n n
is the numb er of 1s in the sequence of the function
we consider an op eration on the vector which p er-
f (x) g (x). The function (1), with the restriction
mutes the p ositions of a and a . Then the vector
i j
such that c = 0 for all where wt( ) > 1, is
b ecomes
called an ane function and denoted by '(x). Us-
ing the dot pro duct we can write ane functions
(a ; ;a ; ;a ; ;a ):
1 j i n
with the form
We denote the op eration of p ermuting the posi-
'(x)= x c;
tions of a and a by the op erator = (ij ) and
i j
then we write
where 2 V ; c =0; 1. An ane function is called
n
a linear function if c = 0 (which corresp onds c =
0
(a ;a ; ;a )=(a ; ;a ; ;a ; ;a ):
1 2 n 1 j i n
0 in the function (1)). The following de nitions
are the most imp ortant cryptographic parameters
The p ermutations for an n-tuple vector in V may
n
for a b o olean functions in cryptography [3 , 9, 10 ].
apply to more than two entries. Thus the op era-
tion =(ij k )isde nedbytheith entry go es
to j th p osition, the j th entry go es to k th p osition,
De nition 1 Let f (x) be a function on V . If,
n
and so on. Thus the op erator =(ij k ), acting
as x runs through al l vectors in V , f (x) = 1 is
n
n 1
on the vector , for example, gives the vector
true 2 times f (x)=1, then the function f (x)
is said to bebalanced.
(a ; ;a ;a ;a ; ;a ;a ;a ; ;a ):
1 i 1 i+1 j 1 i j +1 n
k
De nition 2 Let f (x) be a function on V . The
n
Let and be anytwo op erators for a vector
i j
nonlinearity (denoted by N ) of the function f (x)
f
2 V . Then the combination of the op erators is
n
is de ned by the minimum Hamming distance
de ned by = such that
i j
from f (x) to al l ane functions over V i.e.
n
=( ) = ( ):
i j i j
N = minfwt(f ') j for al l ' on v g:
n
f
The inverse of an op erator exists. For =