Commutative Presemifields and Semifields: Difference between revisions

From Boolean
No edit summary
Line 82: Line 82:
* <math>x^{p^s+1}-a^{p^t-1}x^{p^t+p^{2t+s}}</math> over <math>\mathbb{F}_{p^{3t}}, a</math> primitive, <math>gcd(3,t)=1, t-s\equiv0</math> mod <math>3, 3t/gcd(s,3t)</math> odd;
* <math>x^{p^s+1}-a^{p^t-1}x^{p^t+p^{2t+s}}</math> over <math>\mathbb{F}_{p^{3t}}, a</math> primitive, <math>gcd(3,t)=1, t-s\equiv0</math> mod <math>3, 3t/gcd(s,3t)</math> odd;
* <math>x^{p^s+1}-a^{p^t-1}x^{p^{3t}+p^{t+s}}</math> over <math>\mathbb{F}_{p^{4t}}, a</math> primitive, <math>p^s\equiv p^t\equiv1</math> mod 4, <math>2t/gcd(s,2t)</math> odd;
* <math>x^{p^s+1}-a^{p^t-1}x^{p^{3t}+p^{t+s}}</math> over <math>\mathbb{F}_{p^{4t}}, a</math> primitive, <math>p^s\equiv p^t\equiv1</math> mod 4, <math>2t/gcd(s,2t)</math> odd;
* <math>a^{1-p}x^2+x^{2p^m}+a^{1-p}\sum_{i=0}^k(-1)^ix^{p^{2i}(p^2+1)}+\sum_{j=0}^{k-1}(-1)^{k+j}x^{p^{2j+1}(p^2+1)}+</math>
* <math>a^{1-p}x^2+x^{2p^m}+a^{1-p}T(x)-T(x)^{p^m}</math>, with <math>T(x)=\sum_{i=0}^k(-1)^ix^{p^{2i}(p^2+1)}+a^{p-1}\sum_{j=0}^{k-1}(-1)^{k+j}x^{p^{2j+1}(p^2+1)}</math>, over <math>\mathbb{F}_{p^{2m}}</math> for <math>a\in\mathbb{F}^\star_{p^2}, m=2k+1</math>.
<math>-(\sum_{i=0}^k(-1)^ix^{p^{2i}(p^2+1)}+a^{p-1}\sum_{j=0}^{k-1}(-1)^{k+j}x^{p^{2j+1}(p^2+1)})^{p^m}</math> over <math>\mathbb{F}_{p^{2m}}</math> for <math>a\in\mathbb{F}^star_{p^2}, m=2k+1</math>.

Revision as of 07:22, 18 September 2019

Background

For a prime p and a positive integer n let 𝔽pn be the finite field with pn elements. Let F be a map from the finite field to itself. Such function admits a unique representation as a polynomial of degree at most pn−1, i.e.

F(x)=∑j=0pn−1ajxj,aj∈𝔽pn.

The function F is

  • linear if F(x)=∑j=0n−1ajxpj,
  • affine if it is the sum of a linear function and a constant,
  • DO (Dembowski-Ostrim) polynomial if F(x)=∑0≤i≤j<naijxpi+pj,
  • quadratic if it is the sum of a DO polynomial and an affine function.

For δ a positive integer, the function F is called differentially δ-uniform if for any pairs a,b∈𝔽pn, with a≠0, the equation F(x+a)−F(x)=b admits at most δ solutions.

A function F is called planar or perfect nonlinear (PN) if δF=1. Obviously such functions exist only for p an odd prime. In the even case the smallest possible case for δ is two (APN function).

For planar function we have that the all the nonzero derivatives, DaF(x)=F(x+a)−F(x), are permutations.

Equivalence Relations

Two functions F and F′ from 𝔽pn to itself are called:

  • affine equivalent if F′=A1∘F∘A2, where A1,A2 are affine permutations;
  • EA-equivalent (extended-affine) if F′=F″+A, where A is affine and F″ is afffine equivalent to F;
  • CCZ-equivalent if there exists an affine permutation ℒ of 𝔽pn×𝔽pn such that ℒ(GF)=GF′, where GF={(x,F(x)):x∈𝔽pn}.

CCZ-equivalence is the most general known equivalence relation for functions which preserves differential uniformity. Affine and EA-equivalence are its particular cases. For the case of quadratic planar functions the isotopic equivalence is more general than CCZ-equivalence, where two maps are isotopic equivalent if the corresponding presemifields are isotopic.

On Presemifields and Semifields

A presemifield is a ring with left and right distributivity and with no zero divisor. A presemifield with a multiplicative identity is called a semifield. Any finite presemifield can be represented by 𝕊=(𝔽pn,+,⋆), for p a prime, n a positive integer, 𝕊=(𝔽pn,+) additive group and x⋆y multiplication linear in each variable. Every commutative presemifield can be transformed into a commutative semifield[1].

Two presemifields 𝕊1=(𝔽pn,+,⋆) and 𝕊2=(𝔽pn,+,∘) are called isotopic if there exist three linear permutations T,M,N of 𝔽pn such that T(x⋆y)=M(x)∘N(y), for any x,y∈𝔽pn. If M=N then they are called strongly isotopic. Each commutative presemifields of odd order defines a planar DO polynomial and viceversa:

  • given 𝕊=(𝔽pn,+,⋆) let F𝕊(x)=12(x⋆x);
  • given F let 𝕊F=(𝔽pn,+,⋆) defined by x⋆y=F(x+y)−F(x)−F(y).

Given 𝕊=(𝔽pn,+,⋆) a finite semifield, the subsets

Nl(𝕊)={α∈𝕊:(α⋆x)⋆y=α⋆(x⋆y) for all x,y∈𝕊}

Nm(𝕊)={α∈𝕊:(x⋆α)⋆y=x⋆(α⋆y) for all x,y∈𝕊}

Nr(𝕊)={α∈𝕊:(x⋆y)⋆α=x⋆(y⋆α) for all x,y∈𝕊}

are called left, middle and right nucleus of 𝕊.

The set N(𝕊)=Nl(𝕊)∩Nm(𝕊)∩Nr(𝕊) is called the nucleus. All these sets are finite field and, when 𝕊 is commutative, Nl(𝕊)=Nr(𝕊)⊆Nm(𝕊). The order of the different nuclei are invariant under isotopism.

Properties

Hence two quadratic planar functions F,F′ are isotopic equivalent if their corresponding presemifields are isotopic. Moreover, we have:

  • F,F′ are CCZ-equivalent if and only if 𝕊F,𝕊F′ are strongly isotopic[2];
  • for n odd, isotopic coincides with strongly isotopic;
  • if F,F′ are isotopic equivalent, then there exists a linear map L such that F′ is EA-equivalent to F(x+L(x))−F(x)−F(L(x));
  • any commutative presemifield of odd order can generate at most two CCZ-equivalence classes of planar DO polynomials;
  • if 𝕊1 and 𝕊2 are isotopic commutative semifields of characteristic p with order of middle nuclei and nuclei pm and pk respectively, then either one of the following is satisfied:
    • m/k is odd and the semifields are strongly isotopic,
    • m/k is even and the semifields are strongly isotopic or the only isotopisms are of the form (α⋆N,N,L) with α∈Nm(𝕊1) non-square.

Known cases od planar functions and commutative semifields

Among the known example of planar functions, the only ones that are non-quadratic are the power functions x3t+12 defined over 𝔽3n, with t is odd and gcd(t,n)=1.

In the following the list of some known infinite families of planar functions (and corresponding commutative semifields):

  • x2 over 𝔽pn (finite field 𝔽pn);
  • xpt+1 over 𝔽pn with n/gcd(t,n) odd (Albert's commutative twisted fields);
  • L(t2(x))+12x2 over 𝔽p2km with L(x)=18(xpk−x),t(x)=xpkm−x (Dickson semifields);
    • (ax)ps+1−(ax)pk(ps+1)+xpk+1
    • bxps+1+(bxps+1)pk+cxpk+1

over 𝔽p2k where a,b∈𝔽22k⋆,b not square, c∈𝔽22k∖𝔽2k,gcd(k+s,2k)=gcd(k+s,k) and for the first one also gcd(ps+1,pk+1)≠gcd(ps+1,(pk+1)/2). Without loss of generality it is possible to take a=1 and fix a value for c;

  • xps+1−apt−1xpt+p2t+s over 𝔽p3t,a primitive, gcd(3,t)=1,t−s≡0 mod 3,3t/gcd(s,3t) odd;
  • xps+1−apt−1xp3t+pt+s over 𝔽p4t,a primitive, ps≡pt≡1 mod 4, 2t/gcd(s,2t) odd;
  • a1−px2+x2pm+a1−pT(x)−T(x)pm, with T(x)=∑i=0k(−1)ixp2i(p2+1)+ap−1∑j=0k−1(−1)k+jxp2j+1(p2+1), over 𝔽p2m for a∈𝔽p2⋆,m=2k+1.
  1. ↑ Coulter R. S., Henderson M. Commutative presemifields and semifields. Advances in Math. 217, pp. 282-304, 2008
  2. ↑ Budaghyan L., Helleseth T. On Isotopism of Commutative Presemifields and CCZ-Equivalence of Functions. Special Issue on Cryptography of International Journal of Foundations of Computer Science, v. 22/6), pp- 1243-1258, 2011