Keywords

These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

1 Introduction

Binary bent functions are usually called Boolean bent functions. These functions were first introduced by Rothaus in [12]. Bent functions are closely related to other combinatorial and algebraic objects such as Hadamard difference sets, relative difference sets, planar functions and commutative semi-fields. Later, this notion has been generalized to that of \(p\)-ary bent functions [11]. Several studies on \(p\)-ary bent functions have been performed (a non exhaustive list is [5, 710, 13]). Most of them concern constructions of bent functions or studies of their properties. Another important family of binary functions is that of plateaued functions [3]. Like the notion of bent function, the notion of plateaued function can be generalized to \(p\)-ary plateaued functions (see [4] for instance). In this paper, we establish characterizations of bent functions and plateaued functions in terms of sums of powers of the Walsh transform (Theorems 1 and 3). We also introduce the notion of directional difference for \(p\)-ary functions, generalizing the directional derivative of Boolean functions (Definition 1). We then show that one can establish identities linking sums of fourth-powers of the Walsh transform and directional derivatives of a \(p\)-ary function (Proposition 1). We then deduce from our characterizations of all bent \(p\)-ary functions of algebraic degree \(3\) when \(p\) is odd (Theorem 4). We finally establish a link between the bentness of all elements of a family of \(p\)-ary functions and counting zeros of their directional differences (Theorem 6 and Corollary 2).

2 Notation and Preliminaries

Let \(p\) be a prime integer, \(n\ge 1\) be an integer. We will denote \({\mathbb F}_{p^{n}}\) the finite field of size \(p^n\) and \({\mathbb F}_{p^{n}}^\star \) the set of nonzero elements of \({\mathbb F}_{p^{n}}\). Let \(\xi _p\) be a primitive \(p\)th-root of unity and set \(\chi _p(a)=\xi _p^{a}\). Let \(f\) be a function from \({\mathbb F}_{p^{n}}\) to \({\mathbb F}_{p^{}}\). The Walsh transform of \(f\) at \(w\in {\mathbb F}_{p^{n}}\) is defined as

$$\begin{aligned} \widehat{\chi _{f}}(w) = \sum _{x\in {\mathbb F}_{p^{n}}}\chi _p\Big ({f(x)-Tr^{p^{n}}_{p}(wx)}\Big ). \end{aligned}$$

Then \(f\) is bent if and only if \(\big |Wa f(w)\big |^2=p^n\) for every \(w\in {\mathbb F}_{p^{n}}\). It is said to be regular bent if there exists \(f^\star : {\mathbb F}_{p^{n}}\rightarrow {\mathbb F}_{p^{}}\) such that \(\widehat{\chi _{f}}(w)=\chi _p(f^\star (w))p^{\frac{n}{2}}\) for all \(w\in {\mathbb F}_{p^{n}}\). The function \(f^\star \) is called the dual function of \(f\) (in characteristic \(2\), all bent functions are regular bent; when \(p\) is odd, regular bent functions can exist only if \(p\equiv 1\mod 4\)). A function \(f:{\mathbb F}_{p^{n}}\rightarrow {\mathbb F}_{p^{}}\) is said to be weakly regular bent if, for all \(w\in {\mathbb F}_{p^{n}}\), we have \(\widehat{\chi _{f}}(w) = \epsilon \chi _p(f^\star (w))p^{\frac{n}{2}}\) for some complex number with \(\big |\epsilon \big |=1\) (in fact \(\epsilon \) can only be \(\pm 1\) or \(\pm i\)). For every function \(f\) from \({\mathbb F}_{p^{n}}\) to \({\mathbb F}_{p^{}}\), we have

$$\begin{aligned} \sum _{w\in {\mathbb F}_{p^{n}}}\widehat{\chi _{f}}(w) = p^n\chi _p(f(0)). \end{aligned}$$
(1)

Set \(\big | z\vert ^2=z\bar{z}\) where \(\bar{z}\) stands for the conjugate of \(z\). Then

$$\begin{aligned} \sum _{w\in {\mathbb F}_{p^{n}}}\big |\widehat{\chi _{f}}(w)\big |^2 = p^{2n}. \end{aligned}$$
(2)

In the sequel, we shall refer to (2) as the Parseval identity. If \(\big | \widehat{\chi _{f}}(w)\big |\in \big \{0,p^{\frac{n+s}{2}}\big \}\) for some nonnegative integer \(s\) then \(f\) is said to be \(s\)-plateaued. With this definition, bent functions are \(0\)-plateaued functions (in the case where \(s=0\), \(\big | \widehat{\chi _{f}}(w)\big |\in \big \{0,p^{\frac{n}{2}}\big \}\) is equivalent to \(\big | \widehat{\chi _{f}}(w)\big |=p^{\frac{n}{2}}\)). The Parseval identity allows to compute the multiplicity of each value of the Walsh transform (when \(p=2\), a more precise statement has been shown in [2]).

Lemma 1

Let \(f:{\mathbb F}_{p^{n}}\rightarrow {\mathbb F}_{p^{}}\) be \(s\)-plateaued. Then the absolute value of the Walsh transform \(\widehat{\chi _{f}}\) takes \(p^{n-s}\) times the value \(p^{\frac{n+s}{2}}\) and \(p^n-p^{n-s}\) times the value \(0\).

Proof

If \(N\) denotes the number of \(w\in {\mathbb F}_{p^{n}}\) such that \(\big |\widehat{\chi _{f}}(w)\big |=p^{\frac{n+s}{2}}\), then \(\sum _{w\in {\mathbb F}_{p^{n}}}\big |\widehat{\chi _{f}}(w)\big |^2=p^{n+s}N\). Now, according to Eq. (2), one must have that \(p^{n+s}N=p^{2n}\), that is, \(N=p^{n-s}\). The result follows.

A map \(F\) from \({\mathbb F}_{p^{n}}\) to \({\mathbb F}_{p^{n}}\) is said to be planar if and only if the function from \({\mathbb F}_{p^{n}}\) to \({\mathbb F}_{p^{n}}\) induced by the polynomial \(F(X+a)-F(x)-F(a)\) is bijective for every \(a\in {\mathbb F}_{p^{n}}^\star \). We finally introduce the directional difference.

Definition 1

Let \(f :{\mathbb F}_{p^{n}}\rightarrow {\mathbb F}_{p^{}}\). The directional difference of \(f\) at \( a\in {\mathbb F}_{p^{n}}\) is the map \(D_af\) from \({\mathbb F}_{p^{n}}\) to \({\mathbb F}_{p^{}}\) defined by

$$\forall x\in {\mathbb F}_{p^{n}},\quad D_af(x)=f(x+a)-f(x).$$

3 New Characterizations of Plateaued Functions

Let \(p\) be a positive prime integer. For any nonnegative integer \(k\), we set

$$\begin{aligned} S_{k}(f) = \sum _{w\in {\mathbb F}_{p^{n}}} \big |\widehat{\chi _{f}}(w)\big |^{2k} \text{ and } T_{k}(f)=\frac{S_{k+1}(f)}{S_k(f)} \end{aligned}$$

with the convention regarding \(k=0\) that \(S_0(f)=p^n\) (in this case, \(T_0(f)=\frac{S_1(f)}{S_0(f)}=p^n\)). Let us make a preliminary but important remark : for every integer \(A\) and every positive integer \(k\), it holds

$$\begin{aligned} \nonumber&\sum _{w\in {\mathbb F}_{p^{n}}}\left( \big |\widehat{\chi _{f}}(w)\big |^{2}-A\right) ^2\big |\widehat{\chi _{f}}(w)\big |^{2(k-1)}\\&\qquad \quad =S_{k+1}(f)-2A S_k(f) + A^2 S_{k-1}(f). \end{aligned}$$
(3)

We are now going to deduce from (3) a characterization of plateaued functions in terms of moments of the Walsh transform (in Sect. 4, we shall specialize our characterization to bent functions, see Theorem 3).

Theorem 1

Let \(n\) and \(k\) be two positive integers. Let \(f\) be a function from \({\mathbb F}_{p^{n}}\) to \({\mathbb F}_{p^{}}\). Then, the two following assertions are equivalent.

  1. 1.

    \(f\) is plateaued, that is, there exists a nonnegative integer \(s\) such that \(f\) is \(s\)-plateaued.

  2. 2.

    \(T_{k+1}(f)=T_{k}(f)\).

Proof

 

  1. 1.

    Suppose that \(f\) is \(s\)-plateaued for some nonnegative integer \(s\), that is, \(\big |\widehat{\chi _{f}}(w)\big |\in \{0,p^{\frac{n+s}{2}}\}\). Then, by Lemma 1,

    $$\begin{aligned} S_{k}(f)&= \sum _{w\in {\mathbb F}_{p^{n}}}\big |\widehat{\chi _{f}}(w)\big |^{2k}=p^{n-s}\times p^{k(n+s)}=p^{(k+1)n+(k-1)s}\\ S_{k+1}(f)&= p^{n-s}\times p^{(k+1)(n+s)}=p^{(k+2)n+ks}\\ S_{k+2}(f)&= p^{n-s}\times p^{(k+2)(n+s)}=p^{(k+3)n+(k+1)s}. \end{aligned}$$

    Therefore

    $$T_k(f)=\frac{p^{(k+2)n+ks}}{p^{(k+1)n+(k-1)s}}=p^{n+s}$$

    and

    $$T_{k+1}(f)=\frac{p^{(k+3)n+(k+1)s}}{p^{(k+2)n+ks}}=p^{n+s}=T_k(f).$$
  2. 2.

    Suppose \(T_{k+1}(f)=T_{k}(f)\). According to (3)

    $$\begin{aligned}&\sum _{w\in {\mathbb F}_{p^{n}}}\left( \big |\widehat{\chi _{f}}(w)\big |^2-T_k(f)\right) ^2\big |\widehat{\chi _{f}}(w)\big |^{2k}\\&\qquad =S_{k+2}(f) - 2T_k(f)S_{k+1}(f)+T_k^2(f)S_k(f)\\&\qquad =S_{k+1}(f)\left( T_{k+1}(f) - 2 T_k(f) + T_k(f)\right) = 0 \end{aligned}$$

    proving that \(\big |\widehat{\chi _{f}}(w)\big |\in \{0,\sqrt{T_k(f)}\}\) for every \(w\in {\mathbb F}_{p^{n}}\). Thus,

    $$ \sum _{w\in {\mathbb F}_{p^{n}}}\big |\widehat{\chi _{f}}(w)\big |^2=T_k(f)\#\{w\in {\mathbb F}_{p^{n}}\mid \big |\widehat{\chi _{f}}(w)\big |=\sqrt{T_k(f)}\}. $$

    Now, the Parseval identity (2) states that

    $$ \sum _{w\in {\mathbb F}_{p^{n}}}\big |\widehat{\chi _{f}}(w)\big |^2=p^{2n}. $$

    Therefore \(T_k(f)\) divides \(p^{2n}\) proving that \(T_k(f)=p^{\rho }\) for some positive integer \(\rho \). Now, one has \(\#\{w\in {\mathbb F}_{p^{n}}\big | \big |\widehat{\chi _{f}}(w)\big |=\sqrt{T_k(f)}\}=p^{2n-\rho }\le p^n\) which implies that \(\rho \ge n\), that is, \(\rho =n+s\) for some nonnegative integer \(s\).

Remark 1

Specializing Theorem 1 to the case where \(k=1\), we get that \(f\) is plateaued if and only if \(T_{2}(f)=T_1(f)\), that is

$$ S_3(f)S_1(f) - S_2^2(f) = p^{2n}S_3(f) - S_2^2(f) = 0. $$

Remark 2

In the proof, we have shown more than the sole equivalence between (1) and (2). Indeed, we have shown that if (2) holds then \(f\) is \(s\)-plateaued and \(\big |\widehat{\chi _{f}}(w)\big |\in \{0,\sqrt{T_k(f)}\}\).

In Theorem 1, we have considered the ratio of two consecutive sums \(S_k(f)\). In fact, one can get a more general result than Theorem 1. Indeed, for every positive integer \(k\) and every nonnegative integer \(l\), we have

$$\begin{aligned}&\sum _{w\in {\mathbb F}_{p^{n}}}\left( \big |\widehat{\chi _{f}}(w)\big |^{2l}-A\right) ^2\big |\widehat{\chi _{f}}(w)\big |^{2(k-1)}\\\nonumber&\qquad =S_{k+2l-1}(f)-2A S_{k+l-1}(f) + A^2 S_{k-1}(f). \end{aligned}$$
(4)

Then, one can make the same kind of proof as that of Theorem 1 but with (4) in place of (3) (the proof being very similar, we omit it).

Theorem 2

Let \(n\), \(k\) and \(l\) be positive integers and \(f : {\mathbb F}_{p^{n}}\rightarrow {\mathbb F}_{p^{}}\). Then, the two following assertions are equivalent

  1. 1.

    \(f\) is plateaued, that is, there exists a nonnegative integer \(s\) such that \(f\) is \(s\)-plateaued.

  2. 2.

    \(\frac{S_{k+2l}(f)}{S_{k+l}(f)}=\frac{S_{k+l}(f)}{S_{k}(f)}\).

4 The Case of Bent Functions

In this section, we shall specialize our study to bent functions and suppose that \(p\) is a positive prime integer. In the whole section, \(n\) is a positive integer. In Theorem 1, we have excluded the possibility to for the integer \(k\) to be equal to \(0\) because it does concern both plateaued functions and bent functions. In fact, if we aim to characterize only bent functions, we are going to show that it follows from comparing \(T_1(f)=\frac{S_2(f)}{S_1(f)}=\frac{S_2(f)}{p^{2n}}\) to \(T_0(f)=\frac{S_1(f)}{S_0(f)}=p^n\).

Theorem 3

Let \(n\) be a positive integer. Let \(f\) be a function from \({\mathbb F}_{p^{n}}\) to \({\mathbb F}_{p^{}}\). Then

$$S_2(f)=\sum _{w\in {\mathbb F}_{p^{n}}}\big |\widehat{\chi _{f}}(w)\big |^4 \ge p^{3n}$$

and \(f\) is bent if and only if \(S_2(f)=p^{3n}\).

Proof

If we apply (3) with \(A=p^n\) at \(k=1\), we get that

$$\begin{aligned} \sum _{w\in {\mathbb F}_{p^{n}}}\Big (\big |\widehat{\chi _{f}}(w)\big |^2-p^n\Big )^2&= S_2(f) - 2p^n S_1(f) +p^{2n}S_0(f). \end{aligned}$$

Now, \(S_0(f)=p^n\) and \(S_1(f)=p^{2n}\) (Parseval identity, Eq. 2). Hence

$$\begin{aligned} \sum _{w\in {\mathbb F}_{p^{n}}}\Big (\big |\widehat{\chi _{f}}(w)\big |^2-p^n\Big )^2&= S_2(f) -p^{3n}. \end{aligned}$$
(5)

Since \(\Big (\big |\widehat{\chi _{f}}(w)\big |^2-p^n\Big )^2\ge 0\) for every \(w\in {\mathbb F}_{p^{n}}\), it implies that \(S_2(f)\ge p^{3n}\). Now, \(f\) is bent if and only if \(\big |\widehat{\chi _{f}}(w)\big |^2=p^n\) for every \(w\in {\mathbb F}_{p^{n}}\). Therefore, \(f\) is bent if and only if the left-hand side of Eq. (5) vanishes, that is, if and only if \(S_2(f)=p^{3n}\).

In characteristic \(2\), identities have been established involving the Walsh transform of a Boolean function and its directional derivatives (see [1, 3]). For instance, for every Boolean function \(f\), \(S_2(f)\) and the second-order derivatives of \(f\) have been linked. We now show that one can link \(S_2(f)\) and the directional difference defined in Definition 1.

Proposition 1

Let \(n\) be a positive integer. Let \(f\) be a function from \({\mathbb F}_{p^{n}}\) to \({\mathbb F}_{p^{}}\). Then

$$\begin{aligned} \sum _{w\in {\mathbb F}_{p^{n}}}\big |\widehat{\chi _{f}}(w)\big |^4 = p^{n}\sum _{(a,b,x)\in {\mathbb F}_{p^{n}}^3}\chi _p(D_aD_bf(x)). \end{aligned}$$
(6)

Proof

Since \(\big | z\vert ^4=z^2\overline{z}^2\) where \(\overline{z}\) stands for the conjugate of \(z\) and \(\overline{\xi _p}=\xi _p^{-1}\), we have

$$\begin{aligned}&\sum _{w\in {\mathbb F}_{p^{n}}}\big |\widehat{\chi _{f}}(w)\big |^4\\&=\sum _{w\in {\mathbb F}_{p^{n}}}\sum _{(x_1,x_2,x_3,x_4)\in {\mathbb F}_{p^{n}}^4}\chi _p\big (f(x_1)-f(x_2)+f(x_3)-f(x_4)\\&\qquad \qquad \qquad \qquad \qquad \qquad -Tr^{p^{n}}_{p}(w(x_1-x_2+x_3-x_4))\big ). \end{aligned}$$

Now,

$$\begin{aligned} \sum _{w\in {\mathbb F}_{p^{n}}}\chi _p\big (-Tr^{p^{n}}_{p}(w(x_1-x_2+x_3-x_4))\big )=\left\{ \begin{array}{ll} p^n&{}\text{ if } x_1-x_2+x_3-x_4=0\\ 0 &{} \text{ otherwise. }\end{array}\right. \end{aligned}$$

Hence,

$$\begin{aligned} \sum _{w\in {\mathbb F}_{p^{n}}}\big |\widehat{\chi _{f}}(w)\big |^4=p^n\sum _{(x_1,x_2,x_3)\in {\mathbb F}_{p^{n}}^3}\chi _p\big (f(x_1)-f(x_2)+f(x_3)-f(x_1-x_2+x_3) \big ). \end{aligned}$$

Now note that

$$ D_{x_2-x_1}D_{x_3-x_2}f(x_1)=f(x_1)+f(x_3)-f(x_2)-f(x_1+x_3-x_2). $$

Then, since \((x_1,x_2,x_3)\mapsto (x_1,x_2-x_1,x_3-x_2)\) is a permutation of \({\mathbb F}_{p^{n}}^3\), we get

$$\begin{aligned} \sum _{w\in {\mathbb F}_{p^{n}}}\big |\widehat{\chi _{f}}(w)\big |^4=p^n\sum _{(a,b,x)\in {\mathbb F}_{p^{n}}^3}\chi _p\big (D_aD_bf(x)\big ). \end{aligned}$$

Remark 3

In odd characteristic \(p\), when \(f\) is a quadratic form over \({\mathbb F}_{p^{n}}\), that is, \(f(x)=\phi (x,x)\) for some symmetric bilinear map \(\phi \) from \({\mathbb F}_{p^{n}}\times {\mathbb F}_{p^{n}}\) to \({\mathbb F}_{p^{n}}\), then, \(f(x+y)=f(x)+f(y)+2\phi (x,y)\). Let us now compute the directional differences of \(f\) at \((a,b)\in {\mathbb F}_{p^{n}}\) :

$$\begin{aligned} D_b f(x)&= f(x+b)-f(x) = f(b)+2\phi (b,x)\\ D_aD_bf(x)&= 2\phi (b,x+a)-2\phi (b,x)=2\phi (b,a). \end{aligned}$$

According to Proposition 1, one has

$$\begin{aligned} S_2(f)&= p^n\sum _{(a,b,x)\in {\mathbb F}_{p^{n}}^3} \chi _p(2\phi (b,a))\\&= p^{2n}\sum _{b\in {\mathbb F}_{p^{n}}}\sum _{a\in {\mathbb F}_{p^{n}}}\chi _p(2\phi (b,a)). \end{aligned}$$

Now, classical results about character sums over finite abelian groups say that

$$\begin{aligned} \sum _{a\in {\mathbb F}_{p^{n}}}\chi _p(2\phi (b,a))=\left\{ \begin{array}{ll} p^n &{} \text{ if } \phi (b,\bullet )=0\\ 0 &{}\text{ otherwise. }\end{array}\right. \end{aligned}$$

Hence,

$$ S_2(f) = p^{3n}\#\mathfrak {rad}(\phi ) $$

where \(\mathfrak {rad}(\phi )\) stands for the radical of \(\phi \) : \(\mathfrak {rad}(\phi )=\{b\in {\mathbb F}_{p^{n}}\mid \phi (b,\bullet )=0\}\). One can then conclude thanks to Theorem 3 that \(f\) is bent if and only if \(\mathfrak {rad}(\phi )=\{0\}\).

Suppose that \(p\) is odd and consider now functions of the form

$$\begin{aligned} f(x)=Tr^{p^{n}}_{p}\left( \sum _{\begin{array}{c} i,j,k=0\\ i\not =j,j\not =k,k\not =i \end{array}}^{n-1}a_{ijk} x^{p^i+p^j+p^k} + \sum _{\begin{array}{c} i,j=0\\ i\not =j \end{array}}^{n-1}b_{ij}x^{p^i+p^j} \right) . \end{aligned}$$
(7)

We are going to characterize bent functions of that form thanks to Theorem 3 and Proposition 1. But before, let us note that we can rewrite the expression of \(f\) as follows

$$\begin{aligned} f(x)&= Tr^{p^{n}}_{p}\left( \sum _{\begin{array}{c} i,j,k=0\\ i\not =j,j\not =k,k\not =i \end{array}} ^{n-1}a_{ijk}x^{p^i+p^j+p^k}\right) + Tr^{p^{n}}_{p}\left( \sum _{\begin{array}{c} i,j=0\\ i\not =j \end{array}}^{n-1}b_{ij}x^{p^i+p^j}\right) \\&= Tr^{p^{n}}_{p}\left( \sum _{\begin{array}{c} i,j,k=0\\ i\not =j,j\not =k,k\not =i \end{array}} ^{n-1}a_{ijk}^{p^{-i}}x^{1+p^{j-i}+p^{k-i}}\right) + Tr^{p^{n}}_{p}\left( \sum _{\begin{array}{c} i,j=0\\ i\not =j \end{array}}^{n-1}b_{ij}x^{p^i+p^j}\right) \\&= Tr^{p^{n}}_{p}\left( x\sum _{\begin{array}{c} i,j,k=0\\ i\not =j,j\not =k,k\not =i \end{array}} ^{n-1}a_{ijk}^{p^{-i}}x^{p^{j-i}+p^{k-i}}\right) + Tr^{p^{n}}_{p}\left( \sum _{\begin{array}{c} i,j=0\\ i\not =j \end{array}}^{n-1}b_{ij}x^{p^i+p^j}\right) . \end{aligned}$$

In the second equality, we have used the fact that \(Tr^{p^{n}}_{p}\) is invariant under the Frobenius map \(x\mapsto x^p\). Set

$$\begin{aligned} \psi (x,y)&= \frac{1}{2}\sum _{\begin{array}{c} i,j,k=0\\ i\not =j,j\not =k,k\not =i \end{array}}^{n-1}a_{ijk}^{p^{-i}}(x^{p^{j-i}}y^{p^{k-i}} +x^{p^{k-i}}y^{p^{j-i}})\\ \phi (x,y)&= \frac{1}{2} Tr^{p^{n}}_{p}\left( \sum _{\begin{array}{c} i,j=0\\ i\not =j \end{array}} ^{n-1}b_{ij}(x^{p^i}y^{p^j}+x^{p^j}y^{p^i})\right) , \end{aligned}$$

Therefore, a function \(f\) of the form (7) can be written

$$\begin{aligned} f(x) = Tr^{p^{n}}_{p}(x\psi (x,x)) + \phi (x,x) \end{aligned}$$
(8)

where \(\psi : {\mathbb F}_{p^{n}}\rightarrow {\mathbb F}_{p^{n}}\) is a symmetric bilinear map and \(\phi : {\mathbb F}_{p^{n}}\rightarrow {\mathbb F}_{p^{n}}\) is a symmetric bilinear form. We can now state our characterization.

Theorem 4

Suppose that \(p\) is odd. Let \(\phi \) be a symmetric bilinear form over \({\mathbb F}_{p^{n}}\times {\mathbb F}_{p^{n}}\) and \(\psi \) be a symmetric bilinear map from \({\mathbb F}_{p^{n}}\times {\mathbb F}_{p^{n}}\) to \({\mathbb F}_{p^{n}}\). Define \(f :{\mathbb F}_{p^{n}}\rightarrow {\mathbb F}_{p^{}}\) by \(f(x)=Tr^{p^{n}}_{p}( x\psi (x,x))+\phi (x,x))\) for \(x\in {\mathbb F}_{p^{n}}\). For \((a,b)\in {\mathbb F}_{p^{n}}\), set \(\ell _{a,b}(x) = Tr^{p^{n}}_{p}(\psi (a,b)x+a\psi (b,x)+b\psi (a,x))\). For every \(a\in {\mathbb F}_{p^{n}}\), define the vector space \(\mathfrak K_a=\{b\in {\mathbb F}_{p^{n}}\mid \ell _{a,b}=0\}\). Then \(f\) is bent if and only if \(\{a\in {\mathbb F}_{p^{n}}, \phi (a,\bullet )\big |_{\mathfrak K_a}=0\}=\{0\}\).

Proof

According to Theorem 3 and Proposition 1, \(f\) is bent if and only if

$$\begin{aligned} \sum _{(a,b,x)\in {\mathbb F}_{p^{n}}^3}\chi _p(D_bD_af(x))=p^{2n}. \end{aligned}$$
(9)

Now, for \((a,b)\in {\mathbb F}_{p^{n}}^2\),

$$\begin{aligned} D_af(x)&= Tr^{p^{n}}_{p}( (x+a)\psi (a+x,a+x)-x\psi (x,x)) \\&\quad +\phi (x+a,x+a)-\phi (x,x)\\&= Tr^{p^{n}}_{p}\big (a\psi (x,x)+2x\psi (a,x)+2a\psi (a,x)+x\psi (a,a)+a\psi (a,a)\big ) \\&\quad +2\phi (a,x)+\phi (a,a).\\ D_bD_af(x)&= Tr^{p^{n}}_{p}\big (2a\psi (b,x)+a\psi (b,b)+2b\psi (a,x)+ 2x\psi (a,b)+2b\psi (a,b)\\&\quad + 2a\psi (a,b) + b\psi (a,a)\big ) + 2\phi (a,b))\\&= 2\ell _{a,b}(x) +Tr^{p^{n}}_{p}(a\psi (b,b)+b\psi (a,a)+2(a+b)\psi (a,b))+2\phi (a,b). \end{aligned}$$

Note that, \(\ell _{a,b}\) is a linear map from \({\mathbb F}_{p^{n}}\) to \({\mathbb F}_{p^{n}}\). Furthermore, for any \(a\in {\mathbb F}_{p^{n}}\) and \(b\in \mathfrak K_a\), one has

$$\begin{aligned}&\ell _{a,b}(a)=Tr^{p^{n}}_{p}(\psi (a,b)a+a\psi (b,a)+b\psi (a,a))=0,\\&\ell _{a,b}(b)=Tr^{p^{n}}_{p}(\psi (a,b)b+a\psi (b,b)+b\psi (a,b))=0 \end{aligned}$$

which implies, summing those two equations, that

$$Tr^{p^{n}}_{p}(a\psi (b,b)+b\psi (a,a)+2(a+b)\psi (a,b))=0.$$

Hence,

$$\begin{aligned} \sum _{(a,b,x)\in {\mathbb F}_{p^{n}}^3}\chi _p(D_bD_af(x))&= \sum _{(a,b)\in {\mathbb F}_{p^{n}}^3}\chi _p(2\phi (a,b))\sum _{x\in {\mathbb F}_{p^{n}}}\chi _p(2\ell _{a,b}(x))\\&= p^n\sum _{a\in {\mathbb F}_{p^{n}}}\sum _{b\in \mathfrak K_a}\chi _p(2\phi (a,b)).\\ \end{aligned}$$

Now, for every \(a\in {\mathbb F}_{p^{n}}\), the map \(b\in \mathfrak K_a\mapsto \phi (a,b)\) is linear over \(\mathfrak K_a\). Therefore

$$\begin{aligned} \sum _{b\in \mathfrak K_a}\chi _p(2\phi (a,b))=\left\{ \begin{array}{ll} \#\mathfrak K_a&{}\text{ if } \phi (a,\bullet )\big |_{\mathfrak K_a}=0\\ 0&{}\text{ otherwise }\end{array}\right. \end{aligned}$$

Hence, according to (9), \(f\) is bent if and only if

$$\begin{aligned} \sum _{(a,b,x)\in {\mathbb F}_{p^{n}}^3}\chi _p(D_aD_bf(x)) =p^{n}\sum _{a\in {\mathbb F}_{p^{n}},\,\phi (a,\bullet )\big |_{\mathfrak K_a}=0 } \#\mathfrak K_a=p^{2n}, \end{aligned}$$

that is, if and only if,

$$\begin{aligned} \sum _{a\in {\mathbb F}_{p^{n}},\,\phi (a,\bullet )\big |_{\mathfrak K_a}=0 } \#\mathfrak K_a=p^{n}. \end{aligned}$$

Now, if \(a=0\), then \(\mathfrak K_0={\mathbb F}_{p^{n}}\) because \(\ell _{0,b}=0\) for every \(b\in {\mathbb F}_{p^{n}}\). Therefore, \(f\) is bent if and only if

$$\begin{aligned} \sum _{a\in {\mathbb F}_{p^{n}}^\star ,\,\phi (a,\bullet )\big |_{\mathfrak K_a}=0 } \#\mathfrak K_a=0 \end{aligned}$$

which is equivalent to \(\#\mathfrak K_a=0\) for every \(a\in {\mathbb F}_{p^{n}}^\star \) such that \(\phi (a,\bullet )\big |_{\mathfrak K_a}=0\).

We now turn our attention towards maps from \({\mathbb F}_{p^{n}}\) to \({\mathbb F}_{p^{m}}\). Let us extend the notion of bentness to those maps as follows.

Definition 2

Let \(F\) be a Boolean map from \({\mathbb F}_{p^{n}}\) to \({\mathbb F}_{p^{m}}\). For every \(\lambda \in {\mathbb F}_{p^{n}}^\star \), define \(f_\lambda :{\mathbb F}_{p^{n}}\rightarrow {\mathbb F}_{p^{}}\) as : \(f_\lambda (x)=Tr^{p^{m}}_{p}(\lambda F(x))\) for every \(x\in {\mathbb F}_{p^{n}}\). Then \(F\) is said to be bent if and only if \(f_\lambda \) is bent for every \(\lambda \in {\mathbb F}_{p^{n}}^\star \).

Theorem 3 implies

Theorem 5

Let \(F\) be a map from \({\mathbb F}_{p^{n}}\) to \({\mathbb F}_{p^{m}}\). Then, \(F\) is bent if and only if

$$\begin{aligned} \sum _{\lambda \in {\mathbb F}_{p^{m}}^\star }S_2(f_\lambda )=p^{3n}(p^m-1). \end{aligned}$$
(10)

Proof

According to Theorem 3, for every \(\lambda \in {\mathbb F}_{p^{m}}^\star \), \(f_\lambda \) is bent if and only if \(S_2(f_\lambda )=p^{3n}\) which gives (10). Conversely, suppose that (10) holds. Theorem 3 states that \(S_2(f_\lambda )\ge p^{3n}\) for every \(\lambda \in {\mathbb F}_{p^{m}}^\star \). Thus, one has necessarily, for every \(\lambda \in {\mathbb F}_{p^{n}}^\star \), \(S_2(f_\lambda )=p^{3n}\) implying that \(f_\lambda \) is bent for every \(\lambda \in {\mathbb F}_{p^{n}}\), proving that \(F\) is bent.

We now show that one can compute the left-hand side of (10) by counting the zeros of the second-order directional differences.

Proposition 2

Let \(F\) be a Boolean map from \({\mathbb F}_{p^{n}}\) to \({\mathbb F}_{p^{m}}\). Then

$$ \sum _{\lambda \in {\mathbb F}_{p^{m}}^\star }S_2(f_\lambda )=p^{n+m}\mathfrak N(F)-p^{4n} $$

where \(\mathfrak N(F)\) is the number of elements of \(\{(a,b,x)\in {\mathbb F}_{p^{n}}^3\mid D_aD_bF(x) = 0\}\).

Proof

According to Proposition 1, we have

$$ \sum _{\lambda \in {\mathbb F}_{p^{m}}^\star }S_2(f_\lambda )=p^n\sum _{\lambda \in {\mathbb F}_{p^{m}}^\star } \sum _{a,b,x\in {\mathbb F}_{p^{n}}}\chi _p\big (D_{a}D_{b}f_\lambda (x)\big ). $$

Next, \(D_aD_bf_\lambda = Tr^{p^{m}}_{p}(\lambda D_aD_bF)\). Therefore

$$ \sum _{\lambda \in {\mathbb F}_{p^{m}}^\star }S_2(f_\lambda )=p^n\sum _{a,b,x\in {\mathbb F}_{p^{n}}} \sum _{\lambda \in {\mathbb F}_{p^{m}}^\star }\chi _p\big (Tr^{p^{m}}_{p}(\lambda D_aD_bF(x))\big ). $$

That is

$$ \sum _{\lambda \in {\mathbb F}_{p^{m}}^\star }S_2(f_\lambda )=p^n\sum _{a,b,x\in {\mathbb F}_{p^{n}}}\Big ( \sum _{\lambda \in {\mathbb F}_{p^{m}}}\chi _p\big (Tr^{p^{m}}_{p}(\lambda D_aD_bF(x))\big )\Big ) - p^{4n}. $$

We finally get the result from

$$ \sum _{\lambda \in {\mathbb F}_{p^{m}}}\chi _p\big (Tr^{p^{m}}_{p}(\lambda D_aD_bF(x))\big )=\left\{ \begin{array}{ll} 0&{}\text{ if } D_aD_bF(x)\not =0 \\ p^m &{} \text{ if } D_aD_bF(x)=0\end{array}\right. $$

We then deduce from Theorem 3 a characterization of bentness in terms of zeros of the second-order directional differences.

Theorem 6

Let \(F\) be a map from \({\mathbb F}_{p^{n}}\) to \({\mathbb F}_{p^{m}}\). Then \(F\) is bent if and only if \(\mathfrak N(F)=p^{3n-m}+p^{2n}-p^{2n-m}\).

Proof

\(F\) is bent if and only if all the functions \(f_\lambda \), \(\lambda \in {\mathbb F}_{p^{n}}^\star \), are bent. Therefore, according to Proposition 3, if \(F\) is bent then

$$ \sum _{\lambda \in {\mathbb F}_{p^{m}}^\star }S_2(f_\lambda )=(p^m-1)p^{3n}. $$

Now, according to Proposition 2, one has

$$ \sum _{\lambda \in {\mathbb F}_{p^{m}}^\star }S_2(f_\lambda )=p^{n+m}\mathfrak N(F)-p^{4n}. $$

We deduce from the two above equalities that

$$\begin{aligned} \mathfrak N(F)&= p^{-n-m}(p^{4n}+(p^m-1)p^{3n})\\ {}&= p^{3n-m}+p^{2n}-p^{2n-m}. \end{aligned}$$

Conversely, suppose that \(\mathfrak N(F)=p^{3n-m}+p^{2n}-p^{2n-m}\). Then

$$ \sum _{\lambda \in {\mathbb F}_{p^{m}}^\star }S_2(f_\lambda )=p^{n+m}\mathfrak N(F)-p^{4n}= p^{4n}+p^{3n+m}-p^{3n}-p^{4n}=p^{3n}(p^m-1). $$

We then conclude by Theorem 5 that \(F\) is bent.

Note that when \(a=0\) or \(b=0\), \(D_aD_bF\) is trivially equal to \(0\). We state below a slightly different version of Theorem 6 to exclude those trivial cases to characterize the bentness of \(F\).

Corollary 1

Let \(F\) be a map from \({\mathbb F}_{p^{n}}\) to \({\mathbb F}_{p^{m}}\). Then \(F\) is bent if and only if \(\mathfrak N^\star (F)=p^n(p^n-1)(p^{n-m}-1)\) where \(\mathfrak N^\star (F)\) is the number of elements of \(\{(a,b,x)\in {\mathbb F}_{p^{n}}^\star \times {\mathbb F}_{p^{n}}^\star \times {\mathbb F}_{p^{n}}\mid D_aD_bF(x) = 0\}\).

Proof

It follows from Proposition 2 by noting that \(\{(a,b,x)\in {\mathbb F}_{p^{n}}^3\mid D_aD_bF(x) \!= 0\}\) contains the set \(\{(a,0,x),\,a,x\in {\mathbb F}_{p^{n}},\}\cup \{(0,a,x),\,a,x\in {\mathbb F}_{p^{n}}\}\) whose cardinality equals \(p^n(1+2(p^n-1))=2p^{2n}-p^n\). Hence, the cardinality of \(\mathfrak N^\star (F)\) equals \(p^{3n-m}+p^{2n}-p^{2n-m}-(2p^{2n}-p^n)=p^{3n-m}-p^{2n-m}+p^n-p^{2n}=p^{2n-m}(p^n-1)+p^n(1-p^n)=p^n(p^n-1)(p^{n-m}-1)\).

In the particular case of planar functions, Theorem 1 rewrites as follows

Corollary 2

Let \(F : {\mathbb F}_{p^{n}}\rightarrow {\mathbb F}_{p^{n}}\). Then, \(F\) is planar if and only if, \(D_aD_bF\) does not vanish on \({\mathbb F}_{p^{n}}\) for every \((a,b)\in {\mathbb F}_{p^{n}}^\star \times {\mathbb F}_{p^{n}}^\star \).

Proof

\(F\) is planar if and only if \(F\) is bent ([6, Lemma 1.1]). Hence, according to Corollary 1, \(F\) is planar if and only if \(\mathfrak N^\star (F)=0\) proving the result.