This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

The cc-differential properties of a class of power functions

Huan Zhou1, Xiaoni Du 1,2,3, Wenping Yuan1, Xingbin Qiao1
1College of Mathematics and Statistics, Northwest Normal University,
Lanzhou, 730070, China
2Key Laboratory of Cryptography and Data Analytics,
Northwest Normal University, 730070, Lanzhou, China
3Gansu Provincial Research Center for Basic Disciplines of Mathematics and Statistics,
Northwest Normal University, 730070, Lanzhou, China
Email: [email protected]
Corresponding author.
Abstract

Power functions with low cc-differential uniformity have been widely studied not only because of their strong resistance to multiplicative differential attacks, but also low implementation cost in hardware. Furthermore, the cc-differential spectrum of a function gives a more precise characterization of its cc-differential properties. Let f(x)=xpn+32f(x)=x^{\frac{p^{n}+3}{2}} be a power function over the finite field 𝔽pn\mathbb{F}_{p^{n}}, where p3p\neq 3 is an odd prime and nn is a positive integer. In this paper, for all primes p3p\neq 3, by investigating certain character sums with regard to elliptic curves and computing the number of solutions of a system of equations over 𝔽pn\mathbb{F}_{p^{n}}, we determine explicitly the (1)(-1)-differential spectrum of ff with a unified approach. We show that if pn3(mod4)p^{n}\equiv 3\pmod{4}, then ff is a differentially (1,3)(-1,3)-uniform function except for pn{7,19,23}p^{n}\in\{7,19,23\} where ff is an APcN function, and if pn1(mod4)p^{n}\equiv 1\pmod{4}, the (1)(-1)-differential uniformity of ff is equal to 44. In addition, an upper bound of the cc-differential uniformity of ff is also given.

Key words: power function, elliptic curve, cc-differential uniformity, (1)(-1)-differential spectrum.

1 Introduction

Let 𝔽pn\mathbb{F}_{p^{n}} be the finite field with pnp^{n} elements, where pp is a prime and nn is a positive integer. The multiplicative cyclic group of the finite field 𝔽pn\mathbb{F}_{p^{n}} is denoted by 𝔽pn\mathbb{F}_{p^{n}}^{*}. Let 𝔽pn[x]\mathbb{F}_{p^{n}}[x] denote the polynomial ring over 𝔽pn\mathbb{F}_{p^{n}}. Any function f:𝔽pn𝔽pnf:\mathbb{F}_{p^{n}}\rightarrow\mathbb{F}_{p^{n}} can be uniquely represented as a univariate polynomial of degree less than pnp^{n}. Therefore, ff can always be seen as a polynomial in 𝔽pn[x]\mathbb{F}_{p^{n}}[x]. A polynomial f𝔽pn[x]f\in\mathbb{F}_{p^{n}}[x] is called a permutation polynomial of 𝔽pn\mathbb{F}_{p^{n}} if the induced mapping xf(x)x\mapsto f(x) is a permutation over 𝔽pn\mathbb{F}_{p^{n}}. Permutation polynomials are a very important class of polynomials as they have applications in coding theory and cryptography. Nowadays, designing infinite classes of permutation polynomials over finite fields with good cryptographic properties remains an interesting research topic.

Substitution boxes (S-boxes for short) are important nonlinear building blocks in block ciphers, which can be seen as permutation functions over finite fields. To quantify the ability of S-boxes to resist the differential attack[1], one of the most powerful attacks on block ciphers, Nyberg[8] introduced the notion of differential uniformity. More importantly, to estimate the resistance against some variants of differential cryptanalysis, the differential spectrums[2] of S-boxes are also shown to be of great interest.

In [3], Borisov et al. proposed a new differential on block ciphers, called a multiplicative differential, it is an extension of differential cryptanalysis, which helps to identify the weakness of IDEA ciphers. Motivated by the concept of multiplicative differential, Ellingsen et al. in [4] first introduced the (multiplicative) cc-derivative.

Dacf(x)=f(x+a)cf(x),forallx𝔽pn.{}_{c}D_{a}f(x)=f(x+a)-cf(x),~{}\mathrm{for~{}all}~{}x\in\mathbb{F}_{p^{n}}.

In the same paper they presented a generalized notion of differential uniformity, the so-called cc-differential uniformity.

Definition 1.

[4] Let 𝔽pn\mathbb{F}_{p^{n}} be the finite field with pnp^{n} elements, where pp is a prime and nn is a positive integer. For a function f:𝔽pn𝔽pnf:\mathbb{F}_{p^{n}}\rightarrow\mathbb{F}_{p^{n}}, and a,b,c𝔽pna,b,c\in\mathbb{F}_{p^{n}}. Let Δfc(a,b)=#{x𝔽pn:f(x+a)cf(x)=b}{}_{c}\Delta_{f}(a,b)=\#\{x\in\mathbb{F}_{p^{n}}:f(x+a)-cf(x)=b\}. The cc-differential uniformity of ff is defined as

Δfc=max{δfc(a,b):a,b𝔽pn,anda0ifc=1}.{}_{c}\Delta_{f}=\max\{{}_{c}\delta_{f}(a,b):a,b\in\mathbb{F}_{p^{n}},~{}\mathrm{and}~{}a\neq 0~{}\mathrm{if}~{}c=1\}.

If Δfc=δ{}_{c}\Delta_{f}=\delta, then ff is called a differentially (c,δ)(c,\delta)-uniform function. Especially, ff is called a perfect cc-nonlinear (PcN) function if Δfc=1{}_{c}\Delta_{f}=1, and an almost perfect cc-nonlinear (APcN) function if Δfc=2{}_{c}\Delta_{f}=2. It is clear that if c=1c=1 and a0a\neq 0, the cc-differential uniformity becomes the usual differential uniformity and one use the corresponding notation by omit the symbol cc. The smaller the value Δfc{}_{c}\Delta_{f} is, the better ff resists against multiplicative differential attacks. Thus, the research on cryptographic functions with low cc-differential uniformity has been a hot issue in recent years, the readers can refer to [4, 7, 6, 9, 10, 15, 13, 12, 16, 20, 19, 17, 14] and the references therein.

For a power function f(x)=xdf(x)=x^{d} with a positive integer dd, one can easily see that δfc(a,b)=δfc(1,b/ad){}_{c}\delta_{f}(a,b)={}_{c}\delta_{f}(1,b/a^{d}), for all a𝔽pna\in\mathbb{F}_{p^{n}}^{*} and b𝔽pnb\in\mathbb{F}_{p^{n}}. For simplicity, denoted δfc(1,b){}_{c}\delta_{f}(1,b) by δfc(b){}_{c}\delta_{f}(b) with b𝔽pnb\in\mathbb{F}_{p^{n}}. More precisely, it was proved in [7, Lemma 1] that the cc-differential uniformity of f(x)=xdf(x)=x^{d} is given by

cΔf=max{{cδf(b):b𝔽pn}{gcd(d,pn1)}}._{c}\Delta_{f}=\max\big{\{}\{_{c}\delta_{f}(b):b\in\mathbb{F}_{p^{n}}\}\cup\{\gcd(d,p^{n}-1)\}\big{\}}. (1)

Moreover, as a generalization of the differential spectrum, the cc-differential spectrum of a power function is defined as follows.

Definition 2.

[13] Let f(x)=xdf(x)=x^{d} with cc-differential uniformity Δfc{}_{c}\Delta_{f}. Denote ωic=#{b𝔽pn:δfc(b)=i}{}_{c}\omega_{i}=\#\{b\in\mathbb{F}_{p^{n}}:{}_{c}\delta_{f}(b)=i\}, for each 0iΔfc0\leq i\leq{}_{c}\Delta_{f}. Then the cc-differential spectrum of ff is defined as the multi-set

𝕊={cωi>0:0iΔfc}.\mathbb{S}=\{_{c}\omega_{i}>0:0\leq i\leq{}_{c}\Delta_{f}\}.

Generally speaking, it is difficult to determine the cc-differential spectrum of power functions. To the best of our knowledge, only a few classes of power functions over odd characteristic finite fields have a nontrivial cc-differential spectrum. The known results on power functions ff over 𝔽pn\mathbb{F}_{p^{n}}, for which cc-differential spectrum has been determined are summarized in Table 1, where Tr()n{}_{n}(\cdot) is the absolute trace mapping from 𝔽2n\mathbb{F}_{2^{n}} to 𝔽2\mathbb{F}_{2}.

Table 1: Power functions f(x)=xd𝔽pn[x]f(x)=x^{d}\in\mathbb{F}_{p^{n}}[x] with known cc-differential spectrum
dd Conditions Δfc{}_{c}\Delta_{f} Ref.
23m+22m+2m12^{3m}+2^{2m}+2^{m}-1 p=2,n=4m,0,1c𝔽2n,c1+22m=1p=2,~{}n=4m,~{}0,1\neq c\in\mathbb{F}_{2^{n}},~{}c^{1+2^{2m}}=1 2 [12]
pk+1p^{k}+1  pp is odd, gcd(n,k)=e,1c𝔽pe\gcd(n,k)=e,~{}1\neq c\in\mathbb{F}_{p^{e}} and ne\frac{n}{e} is odd 2 [13]
pk+1p^{k}+1  pp is odd, gcd(n,k)=e,c𝔽pe\gcd(n,k)=e,~{}c\notin\mathbb{F}_{p^{e}}nn is even, k=n2k=\frac{n}{2} 2 [13]
2n22^{n}-2  p=2,c0p=2,c\neq 0, Tr(c)n={}_{n}(c)=Tr(c1)n=1{}_{n}(c^{-1})=1 2 [19]
2n22^{n}-2  p=2,c0p=2,c\neq 0, Tr(c)n=0{}_{n}(c)=0 or Tr(c1)n=0{}_{n}(c^{-1})=0 3 [19]
pn2p^{n}-2  pp is odd, c0,1,4,41c\neq 0,1,4,4^{-1} 2 or 3 [19]
3n33^{n}-3  p=3,c=1p=3,~{}c=-1, any nn 4 or 6 [19]
pk+12\frac{p^{k}+1}{2}  pp is odd,  c=1,gcd(n,k)=1c=-1,~{}\gcd(n,k)=12ngcd(2n,k)\frac{2n}{\gcd(2n,k)} is even p+12\frac{p+1}{2} [19]
pn+12\frac{p^{n}+1}{2}  any ppc1c\neq-1 2,4, pn+34\frac{p^{n}+3}{4} or pn+54\frac{p^{n}+5}{4} [10]
5n32\frac{5^{n}-3}{2} p=5,c=1,n2p=5,~{}c=-1,~{}n\geq 2 2 [19] and [9]
pn32\frac{p^{n}-3}{2} c=1,pn3(mod4),pn>3c=-1,~{}p^{n}~{}\equiv 3~{}(\bmod~{}4),~{}p^{n}>3 2 or  4 [17]
3n+32\frac{3^{n}+3}{2}  p=3p=3c=1c=-1n2n\geq 2 is even 2 [19]
pn+32\frac{p^{n}+3}{2}  c=1c=-1pn1(mod4)p^{n}~{}\equiv 1~{}(\bmod~{}4)p3p\neq 3 4 This paper
pn+32\frac{p^{n}+3}{2}  c=1c=-1pn3(mod4)p^{n}~{}\equiv 3~{}(\bmod~{}4)p3p\neq 3 3 This paper

Throughout this paper, let f(x)=xpn+32𝔽pn[x]f(x)=x^{\frac{p^{n}+3}{2}}\in\mathbb{F}_{p^{n}}[x], where pp is an odd prime and nn is a positive integer. We should mention that when p=3p=3, ff is PccN if n3n\geq 3 is odd, and is APccN if n2n\geq 2 is even [7] with its (1)(-1)-differential spectrum being given in [19, Theorem 9]. When p>3p>3, the (1)(-1)-differential uniformity of ff was discussed by Mesnager et al. in [7, Theorem 11] and they proved that Δf14{}_{-1}\Delta_{f}\leq 4 if pn1(mod4)p^{n}\equiv 1\pmod{4} and Δf13{}_{-1}\Delta_{f}\leq 3 if pn3(mod4)p^{n}\equiv 3\pmod{4}.

Inspired by the works above, in this paper, we mainly study the (1)(-1)-differential spectrum of f(x)=xpn+32f(x)=x^{\frac{p^{n}+3}{2}} for all positive integer nn when p3p\neq 3. More precisely, we prove that Δf1=3{}_{-1}\Delta_{f}=3 if pn3(mod4)p^{n}\equiv 3\pmod{4} except for p{7,19,23}p\in\{7,19,23\} where ff is an APcN function, and Δf1=4{}_{-1}\Delta_{f}=4 if pn1(mod4)p^{n}\equiv 1\pmod{4}, that is, the upper bound of the (1)(-1)-differential uniformity of ff given by Mesnager et al. can be achieved.

The rest of this paper is organized as follows. In Section 2, we introduce some basic notation about quadratic character and results on quadratic character sums, which will be employed in the sequel. In Section 3, we first determine the (1)(-1)-differential spectrum of ff, and some examples are also presented. In Section 4, we give an upper bound of the cc-differential uniformity of ff. Section 5 concludes this paper.

2 Preliminaries

In this section, we mainly introduce some basic results on quadratic character sums over 𝔽pn\mathbb{F}_{p^{n}}. Let η\eta be the quadratic character over 𝔽pn\mathbb{F}_{p^{n}}, i.e., for any x𝔽pnx\in\mathbb{F}_{p^{n}},

η(x)=xpn12={1,if x is a square,0,if x=0,1,if x is a nonsquare.\displaystyle\eta(x)=x^{\frac{p^{n}-1}{2}}=\left\{\begin{array}[]{ll}1,&\hbox{if $x$ is a square,}\\ 0,&\hbox{if $x=0$,}\\ -1,&\hbox{if $x$ is a nonsquare.}\end{array}\right.

It is well-known that x𝔽pnη(x)=0\sum\limits_{x\in\mathbb{F}_{p^{n}}}\eta(x)=0, and η(1)=1\eta(-1)=1 (resp. 1-1) if pn1(mod4)p^{n}\equiv 1~{}(\bmod~{}4) (resp. pn3(mod4)p^{n}\equiv 3~{}(\bmod~{}4)), which are used extensively in the calculations of character sums.

We consider now sums involving the quadratic character of the form

x𝔽pnη(f(x))\sum_{x\in\mathbb{F}_{{p^{n}}}}\eta(f(x))

with f(x)𝔽pn[x]f(x)\in\mathbb{F}_{p^{n}}[x]. Recall that for deg(f(x))=1\deg(f(x))=1, the sums above is trivial, and for deg(f(x))=2\deg(f(x))=2, the following explicit formula was established in [5] and [6], respectively.

Lemma 1.

[5] Let f(x)=a2x2+a1x+a0𝔽pn[x]f(x)=a_{2}x^{2}+a_{1}x+a_{0}\in\mathbb{F}_{p^{n}}[x], pp is odd, and a20a_{2}\neq 0. Put Δ:=a124a0a2\Delta:=a_{1}^{2}-4a_{0}a_{2} be the discriminant of f(x)=0f(x)=0 and let η\eta be the quadratic character of 𝔽pn\mathbb{F}_{p^{n}}. Then

x𝔽pnη(f(x))={η(a2),if Δ0,(pn1)η(a2),if Δ=0.\displaystyle\sum_{x\in\mathbb{F}_{{p^{n}}}}\eta(f(x))=\left\{\begin{array}[]{ll}-\eta(a_{2}),&\hbox{\rm{if} $\Delta\neq 0$,}\\ (p^{n}-1)\eta(a_{2}),&\hbox{\rm{if} $\Delta=0$.}\end{array}\right.

Lemma 2.

[6] Let f(x)=a2x2+a1x+a0𝔽pn[x]f(x)=a_{2}x^{2}+a_{1}x+a_{0}\in\mathbb{F}_{p^{n}}[x] with pp odd and a20a_{2}\neq 0. Then the equation f(x)=0f(x)=0 has two (resp. one) solutions in 𝔽pn\mathbb{F}_{p^{n}} if and only if the discriminant Δ=a124a0a2\Delta=a_{1}^{2}-4a_{0}a_{2} is a nonzero (resp. zero) square in 𝔽pn\mathbb{F}_{p^{n}}. That is to say, the number of solutions of ff is 1+η(Δ)1+\eta(\Delta).

For deg(f(x))3\deg(f(x))\geq 3, it is challenging to obtain an explicit formula for the character sum x𝔽pnη(f(x))\sum\limits_{x\in\mathbb{F}_{p^{n}}}\eta(f(x)). However, when deg(f(x))=3\deg(f(x))=3, such a sum can be computed by considering 𝔽pn\mathbb{F}_{p^{n}}-rational points of elliptic curves over 𝔽p\mathbb{F}_{p}. More specifically, we denote λp,n\lambda_{p,n} as

λp,n=x𝔽pnη(f(x)).\lambda_{p,n}=\sum_{x\in\mathbb{F}_{p^{n}}}\eta(f(x)).

To calculate λp,n\lambda_{p,n}, we will use some elementary concepts from the theory of elliptic curves [11]. Let E/𝔽pE/\mathbb{F}_{p} be the elliptic curve over 𝔽p\mathbb{F}_{p}

E:y2=f(x)E:y^{2}=f(x)

and Np,nN_{p,n} denote the number of 𝔽pn\mathbb{F}_{p^{n}}-rational points (remember the extra point at infinity) on the curve E/𝔽pE/\mathbb{F}_{p}. From the results in [11, P.139, P.142], we have, for all n1n\geq 1,

Np,n=pn+1+λp,nN_{p,n}=p^{n}+1+\lambda_{p,n}

with

λp,n=αnβn,\displaystyle\lambda_{p,n}=-\alpha^{n}-\beta^{n},

where α\alpha and β\beta are the two conjugate complex roots of the polynomial T2+λp,1T+pT^{2}+\lambda_{p,1}T+p. We obtain an explicit and efficient formula of λp,n\lambda_{p,n}.

Define the following two specific character sums

λp,n(1)=x𝔽pnη(x(x+1)(x3))\displaystyle\lambda^{(1)}_{p,n}=\sum_{x\in\mathbb{F}_{p^{n}}}\eta(x(x+1)(x-3)) (2)

and

λp,n(2)=x𝔽pnη(x(x+1)(x2)),\displaystyle\lambda^{(2)}_{p,n}=\sum_{x\in\mathbb{F}_{p^{n}}}\eta(x(x+1)(x-2)), (3)

which play a significant role in our main results. As we will see later, the determination of the (1)(-1)-differential spectrum of f(x)=xpn+32f(x)=x^{\frac{p^{n}+3}{2}} over 𝔽pn\mathbb{F}_{p^{n}} heavily depend on the computations of λp,n(1)\lambda^{(1)}_{p,n} and λp,n(2)\lambda^{(2)}_{p,n}.

In the following examples, we will give the exact values of λp,n(1)\lambda^{(1)}_{p,n} and λp,n(2)\lambda^{(2)}_{p,n}, respectively, for specific values of pp.

Example 1.

Let p=5p=5. We can obtain λ5,1(1)=2\lambda^{(1)}_{5,1}=2 by Magma program. So we have α,β=1±21\alpha,\beta=-1\pm 2\sqrt{-1}. Hence λ5,n(1)=(1+21)n(121)n\lambda^{(1)}_{5,n}=-(-1+2\sqrt{-1})^{n}-(-1-2\sqrt{-1})^{n}.

Example 2.

Let p=13p=13. We can obtain λ13,1(2)=2\lambda^{(2)}_{13,1}=2 by Magma program. So we have α,β=1±23\alpha,\beta=-1\pm 2\sqrt{-3}. Hence λ13,n(2)=(1+23)n(123)n\lambda^{(2)}_{13,n}=-(-1+2\sqrt{-3})^{n}-(-1-2\sqrt{-3})^{n}.

In addition, we have the following bound on λp,n(i)\lambda^{(i)}_{p,n} for i=1,2i=1,2.

Lemma 3.

[11] With the notation as above, we have |λp,n(i)|2pn2|\lambda^{(i)}_{p,n}|\leq 2p^{\frac{n}{2}}, for i=1,2i=1,2.

In the end of this section, we present the following results concerning the exact values specific character sums used in Section 3.

Lemma 4.

If pn3(mod4)p^{n}\equiv 3\pmod{4} and p3p\neq 3, then relevant consequences as follows,

1)x𝔽pnη((2x2)(2x+1)(2x+2))=λp,n(1)1)\sum\limits_{x\in\mathbb{F}_{p^{n}}}\eta\left((2x-2)(2x+1)(2x+2)\right)=\lambda_{p,n}^{(1)},

2)x𝔽pnη((2x2)(2x1)(2x+2))=λp,n(1)2)\sum\limits_{x\in\mathbb{F}_{p^{n}}}\eta\left((2x-2)(2x-1)(2x+2)\right)=-\lambda_{p,n}^{(1)},

3)x𝔽pnη((2x2)(2x1)(2x+1)(2x+2))=λp,n(1)13)\sum\limits_{x\in\mathbb{F}_{p^{n}}}\eta((2x-2)(2x-1)(2x+1)(2x+2))=\lambda_{p,n}^{(1)}-1,

4)x𝔽pnη((2x1)(2x+1)(2x+2))=λp,n(2)4)\sum\limits_{x\in\mathbb{F}_{p^{n}}}\eta\left((2x-1)(2x+1)(2x+2)\right)=\lambda_{p,n}^{(2)},

5)x𝔽pnη((2x2)(2x1)(2x+1))=λp,n(2)5)\sum\limits_{x\in\mathbb{F}_{p^{n}}}\eta\left((2x-2)(2x-1)(2x+1)\right)=-\lambda_{p,n}^{(2)}.

Proof.

1) It is clear that 2x+12x+1 is a permutation of 𝔽pn\mathbb{F}_{p^{n}}. Set t=2x+1t=2x+1. The left side of equation 1) leads to

t𝔽pnη(t(t+1)(t3))=λp,n(1).\sum\limits_{t\in\mathbb{F}_{p^{n}}}\eta(t(t+1)(t-3))=\lambda_{p,n}^{(1)}.

2) Set t=2x1t=2x-1. The left side of equation 2) leads to

t𝔽pnη(t(t1)(t+3))\displaystyle\sum\limits_{t\in\mathbb{F}_{p^{n}}}\eta(t(t-1)(t+3))
=\displaystyle= t𝔽pnη((t)(t1)(t+3))\displaystyle\sum\limits_{t\in\mathbb{F}_{p^{n}}}\eta((-t)(-t-1)(-t+3))
=\displaystyle= t𝔽pnη(t(t+1)(t3))\displaystyle-\sum\limits_{t\in\mathbb{F}_{p^{n}}}\eta(t(t+1)(t-3))
=\displaystyle= λp,n(1)\displaystyle-\lambda_{p,n}^{(1)}

since η(1)=1\eta(-1)=-1.

3) First observe that the left side of equation 3) can be written as

4x210η(4x244x21).\displaystyle\sum\limits_{4x^{2}-1\neq 0}\eta\Big{(}\frac{4x^{2}-4}{4x^{2}-1}\Big{)}. (4)

If we set 4x244x21=u\frac{4x^{2}-4}{4x^{2}-1}=u, then uu and xx satisfy

(44u)x2+u4=0.(4-4u)x^{2}+u-4=0.

When u1u\neq 1, it is a quadratic equation with connection on the variable xx, with discriminant Δ=16(u4)(u1)\Delta=16(u-4)(u-1). Therefore, according to Lemma 2 for all u1\ u\neq 1, it corresponds to (1+η(Δ))(1+\eta(\Delta)) xx’s, and then

Eq.(4)=\displaystyle\mathrm{Eq.}\eqref{ux}= u𝔽pnη(u)(1+η(Δ))η(1)\displaystyle\sum\limits_{u\in\mathbb{F}_{p^{n}}}\eta(u)(1+\eta(\Delta))-\eta(1)
=\displaystyle= u𝔽pnη(u(u4)(u1))1\displaystyle\sum\limits_{u\in\mathbb{F}_{p^{n}}}\eta(u(u-4)(u-1))-1
=\displaystyle= t𝔽pnη(t(t+1)(t3))1\displaystyle\sum\limits_{t\in\mathbb{F}_{p^{n}}}\eta(t(t+1)(t-3))-1
=\displaystyle= λp,n(1)1.\displaystyle\lambda_{p,n}^{(1)}-1.

5) Set t=2x1t=2x-1. The left side of equation 5) leads to

t𝔽pnη(t(t1)(t+2))\displaystyle\sum\limits_{t\in\mathbb{F}_{p^{n}}}\eta(t(t-1)(t+2))
=\displaystyle= t𝔽pnη((t)(t1)(t+2))\displaystyle\sum\limits_{t\in\mathbb{F}_{p^{n}}}\eta((-t)(-t-1)(-t+2))
=\displaystyle= t𝔽pnη(t(t+1)(t2))\displaystyle-\sum\limits_{t\in\mathbb{F}_{p^{n}}}\eta(t(t+1)(t-2))
=\displaystyle= λp,n(2).\displaystyle-\lambda_{p,n}^{(2)}.

The proof of following lemma is very similar to that of [18, Lemma 5], we will no longer prove that.

Lemma 5.

Let λp,n(1)\lambda^{(1)}_{p,n} and λp,n(2)\lambda^{(2)}_{p,n} be defined as above. If pn1(mod4)p^{n}\equiv 1\pmod{4} and p3p\neq 3, then we have

(1)x𝔽pnη(x(x2+x+1))=λp,n(1)(1)\sum\limits_{x\in\mathbb{F}_{p^{n}}}\eta(x(x^{2}+x+1))=\lambda^{(1)}_{p,n},

(2)x𝔽pnη((x+1)(x2+x+1))=λp,n(1)(2)\sum\limits_{x\in\mathbb{F}_{p^{n}}}\eta((x+1)(x^{2}+x+1))=\lambda^{(1)}_{p,n},

(3)x𝔽pnη((x2+x)(x2+x+1))=λp,n(1)1(3)\sum\limits_{x\in\mathbb{F}_{{p^{n}}}}\eta((x^{2}+x)(x^{2}+x+1))=\lambda^{(1)}_{p,n}-1,

(4)x𝔽pnη(x(3x2+2x+3))=λp,n(2)(4)\sum\limits_{x\in\mathbb{F}_{p^{n}}}\eta(x(3x^{2}+2x+3))=\lambda^{(2)}_{p,n},

(5)x𝔽pnη((x2+x+1)(3x2+2x+3))=λp,n(2)η(3)(5)\sum\limits_{x\in\mathbb{F}_{{p^{n}}}}\eta((x^{2}+x+1)(3x^{2}+2x+3))=\lambda^{(2)}_{p,n}-\eta(3),

(6)x𝔽pnη(x(x2+x+1)(3x2+2x+3))=2λp,n(1)(6)\sum\limits_{x\in\mathbb{F}_{p^{n}}}\eta(x(x^{2}+x+1)(3x^{2}+2x+3))=2\lambda^{(1)}_{p,n}.

3 The cc-differential properties of f(x)=xpn+32f(x)=x^{\frac{p^{n}+3}{2}} over 𝔽pn\mathbb{F}_{p^{n}}

In this section, we are about to determine the (1)(-1)-differential spectrum of ff explicitly.

We first introduce some properties of the cc-differential spectrum of power function presented by Yan and Zhang in [19], which will be used to examine the (1)(-1)-differential spectrum of ff.

Lemma 6.

[19] Let f(x)=xdf(x)=x^{d} be a power function over 𝔽pn\mathbb{F}_{p^{n}} with cc-differential uniformity Δfc{}_{c}\Delta_{f} for some 1c𝔽pn1\neq c\in\mathbb{F}_{p^{n}}, where dd is a positive integer. Then we have

i=0Δfcωic=i=0Δfc(iωic)=pn.\displaystyle\sum_{i=0}^{{}_{c}\Delta_{f}}{{}_{c}\omega_{i}}=\sum_{i=0}^{{}_{c}{\Delta_{f}}}(i\cdot{{}_{c}{\omega_{i}}})=p^{n}. (5)

Moreover,

i=0Δfc(i2ωic)=N4c1pn1gcd(d,pn1),\sum\limits_{i=0}^{{}_{c}\Delta_{f}}(i^{2}\cdot{}_{c}\omega_{i})=\frac{{}_{c}N_{4}-1}{p^{n}-1}-\gcd(d,p^{n}-1), (6)

where

N4c=#{(x1,x2,x3,x4)(𝔽pn)4:{x1x2+x3x4=0x1dcx2d+cx3dx4d=0}.{}_{c}N_{4}=\#\left\{{(x_{1},x_{2},x_{3},x_{4})\in(\mathbb{F}_{p^{n}})^{4}:\Bigg{\{}\begin{array}[]{ll}x_{1}-x_{2}+x_{3}-x_{4}&=0\\ x_{1}^{d}-cx_{2}^{d}+cx_{3}^{d}-x_{4}^{d}&=0\end{array}}\right\}. (7)

3.1 pn3(mod4)p^{n}\equiv 3\pmod{4} and p3p\neq 3

In this subsection, we give the (1)(-1)-differential spectrum of the power function f(x)=xpn+32f(x)=x^{\frac{p^{n}+3}{2}} over 𝔽pn\mathbb{F}_{p^{n}}, pn3(mod4)p^{n}\equiv 3\pmod{4} and p3p\neq 3. Before that, we give some notions.

For any square α𝔽pn\alpha\in\mathbb{F}_{p^{n}}^{*}α\sqrt{\alpha} denotes either solution of the equation x2=αx^{2}=\alpha in 𝔽pn\mathbb{F}_{p^{n}}. Let C0C_{0} and C1C_{1} denote the sets of squares and nonsquares in 𝔽pn\mathbb{F}_{p^{n}}^{*}, respectively. The cyclotomic number (i,j)(i,j) is defined as the cardinality of the set Ci,j={x𝔽pn\{0,1}:xCi,x+1Cj}C_{i,j}=\{{x\in\mathbb{F}_{p^{n}}\backslash\{{0,-1}\}:x\in C_{i},x+1\in C_{j}}\}, where i,j{0,1}i,j\in\{{0,1}\}.

Theorem 1.

Let f(x)=xpn+32f(x)=x^{\frac{p^{n}+3}{2}} over 𝔽pn\mathbb{F}_{p^{n}}, where pn3(mod4)p^{n}\equiv 3\pmod{4} and p3p\neq 3. Then the (1)(-1)-differential spectrum of ff is

𝕊={\displaystyle\mathbb{S}=\{ ω01=14(pn+λp,n(1)+1),ω11=116(9pn5λp,n(1)+2λp,n(2)7),{}_{-1}\omega_{0}=\frac{1}{4}(p^{n}+\lambda_{p,n}^{(1)}+1),~{}{}_{-1}\omega_{1}=\frac{1}{16}(9p^{n}-5\lambda_{p,n}^{(1)}+2\lambda_{p,n}^{(2)}-7),
ω21=18(pnλp,n(1)2λp,n(2)+1),ω31=116(pn+3λp,n(1)+2λp,n(2)+1)}.\displaystyle{}_{-1}\omega_{2}=\frac{1}{8}(p^{n}-\lambda_{p,n}^{(1)}-2\lambda_{p,n}^{(2)}+1),~{}{}_{-1}\omega_{3}=\frac{1}{16}(p^{n}+3\lambda_{p,n}^{(1)}+2\lambda_{p,n}^{(2)}+1)\}.
Proof.

In order to determine the (1)(-1)-differential spectrum of ff, the number of solutions Nf(b)N_{f}(b) of the equation

(x+1)d+xd=b,forallb𝔽pn(x+1)^{d}+x^{d}=b,~{}\mathrm{for~{}all}~{}b\in\mathbb{F}_{p^{n}} (8)

needs to be determined.

Evidently, if x=0x=0, then b=1b=1, and if x=1x=-1, then b=1b=-1 since d=pn+32d=\frac{p^{n}+3}{2} is odd. Now we always assume x𝔽pn\{1}x\in\mathbb{F}_{p^{n}}^{*}\backslash\{-1\}, Eq.(8) can be written as

(η(x+1)+η(x))x2+2η(x+1)x+η(x+1)b=0.(\eta(x+1)+\eta(x))x^{2}+2\eta(x+1)x+\eta(x+1)-b=0. (9)

Note that 𝔽pn\{1}=𝒞0,0𝒞0,1𝒞1,0𝒞1,1\mathbb{F}_{p^{n}}^{*}\backslash\{-1\}=\mathcal{C}_{0,0}\cup\mathcal{C}_{0,1}\cup\mathcal{C}_{1,0}\cup\mathcal{C}_{1,1}. The following four cases are discussed.

(1) x𝒞0,0x\in\mathcal{C}_{0,0}, that is η(x)=η(x+1)=1\eta(x)=\eta(x+1)=1. Eq.(9) becomes

2x2+2x+1b=02x^{2}+2x+1-b=0 (10)

with discriminant Δ1=8b4\Delta_{1}=8b-4.

If Δ1=0\Delta_{1}=0, then b=12b=\frac{1}{2}, by Eq.(10), we get x=12x=-\frac{1}{2}. So η(x+1)=η(12+1)=η(12)=η(x)=η(x)\eta(x+1)=\eta(-\frac{1}{2}+1)=\eta(\frac{1}{2})=\eta(-x)=-\eta(x), i.e., η(x)η(x+1)\eta(x)\neq\eta(x+1), which means that x=12𝒞0,0x=-\frac{1}{2}\notin\mathcal{C}_{0,0}.

If η(Δ1)=1\eta\left(\Delta_{1}\right)=1, we can acquire two solutions of Eq.(10) over 𝔽pn\{1}\mathbb{F}_{p^{n}}^{*}\backslash\{-1\}, denote by x1=1+2b12x_{1}=\frac{-1+\sqrt{2b-1}}{2}, x1=12b12x_{1}^{\prime}=\frac{-1-\sqrt{2b-1}}{2}. Apparently, x1+1=x1x_{1}+1=-x_{1}^{\prime}, x1+1=x1x_{1}^{\prime}+1=-x_{1} and x1(x1+1)=b12x_{1}\left(x_{1}+1\right)=\frac{b-1}{2}. Since η(1)=1\eta(-1)=-1, we assert that Eq.(9) has at most one solution in 𝒞0,0\mathcal{C}_{0,0} when bD1b\in D_{1}, where

D1={b𝔽pn:η(2b1)=η(2b2)=1}.D_{1}=\{b\in\mathbb{F}_{p^{n}}:\eta(2b-1)=\eta(2b-2)=1\}.

(2) x𝒞0,1x\in\mathcal{C}_{0,1}, that is η(x)=1,η(x+1)=1\eta(x)=1,~{}\eta(x+1)=-1. Eq.(9) becomes 2x+1+b=02x+1+b=0, one can be deduced that x=1b2x=\frac{-1-b}{2}, x+1=1b2x+1=\frac{1-b}{2}. Hence Eq.(9) has at most one solution when bD2b\in D_{2}, where

D2={b𝔽pn:η(2b2)=1,η(2b+2)=1}.D_{2}=\{b\in\mathbb{F}_{p^{n}}:\eta(2b-2)=1,~{}\eta(2b+2)=-1\}.

(3) x𝒞1,0x\in\mathcal{C}_{1,0}, that is η(x)=1,η(x+1)=1\eta(x)=-1,\eta(x+1)=1. Eq.(9) becomes 2x+1b=02x+1-b=0, then x=b12x=\frac{b-1}{2}, x+1=b+12x+1=\frac{b+1}{2}. Therefore, Eq.(9) has at most one solution in 𝒞1,0\mathcal{C}_{1,0} if bD3b\in D_{3}, where

D3={b𝔽pn:η(2b2)=1,η(2b+2)=1}.D_{3}=\{b\in\mathbb{F}_{p^{n}}:\eta(2b-2)=-1,~{}\eta(2b+2)=1\}.

(4) x𝒞1,1x\in\mathcal{C}_{1,1}, that is η(x)=η(x+1)=1\eta(x)=\eta(x+1)=-1. Eq.(9) becomes

2x2+2x+1+b=0,2x^{2}+2x+1+b=0, (11)

with discriminant Δ2=48b\Delta_{2}=-4-8b.

If Δ2=0\Delta_{2}=0, then b=12b=-\frac{1}{2} and x=12x=-\frac{1}{2}. So η(x+1)=η(12)=η(x)=η(x)\eta(x+1)=\eta(\frac{1}{2})=\eta(-x)=-\eta(x), i.e., η(x)η(x+1)\eta(x)\neq\eta(x+1), which means that x=12𝒞1,1x=-\frac{1}{2}\notin\mathcal{C}_{1,1}.

If η(Δ2)=1\eta\left(\Delta_{2}\right)=1, we can acquire two solutions of Eq.(11) over 𝔽pn\{1}\mathbb{F}_{p^{n}}^{*}\backslash\{-1\}, denote by x2=1+2b12x_{2}=\frac{-1+\sqrt{-2b-1}}{2}, x2=12b12x_{2}^{\prime}=\frac{-1-\sqrt{-2b-1}}{2}. Obviously, x2+1=x2x_{2}+1=-x_{2}^{\prime}, x2+1=x2x_{2}^{\prime}+1=-x_{2} and x2(x2+1)=1+b2x_{2}\left(x_{2}+1\right)=-\frac{1+b}{2}. Since η(1)=1\eta(-1)=-1, we assert that Eq.(9) has at most one solution in 𝒞1,1\mathcal{C}_{1,1} when bD4b\in D_{4}, where

D4={b𝔽pn:η(2b+1)=η(2b+2)=1}.D_{4}=\{b\in\mathbb{F}_{p^{n}}:\eta(2b+1)=\eta(2b+2)=-1\}.

We can get that b=±1i=14Dib=\pm 1\notin\bigcup\limits_{i=1}^{4}D_{i}, D3(D1D2D4)=D_{3}\cap\left(D_{1}\cup D_{2}\cup D_{4}\right)=\emptyset and D1D4=D1D2D4D_{1}\cap D_{4}=D_{1}\cap D_{2}\cap D_{4}, The discussion above can be summed up in Table 2.

Table 2: Solutions to Eq.(8) in 𝔽pn\mathbb{F}_{p^{n}}
bb solutions
b=1b=1 x=0x=0
b=1b=-1 x=1x=-1
bD1b\in D_{1} xA{1+2b12,12b12}x\in A\subset\{\frac{1+\sqrt{2b-1}}{2},\frac{1-\sqrt{2b-1}}{2}\}
bD2b\in D_{2} x=b+12x=-\frac{b+1}{2}
bD3b\in D_{3} x=b12x=\frac{b-1}{2}
bD4b\in D_{4} xB{1+2b12,12b12}x\in B\subset\{\frac{-1+\sqrt{-2b-1}}{2},\frac{-1-\sqrt{-2b-1}}{2}\}

Hence Eq.(8) has at most three solutions in 𝔽pn\mathbb{F}_{p^{n}}, and the number of the solutions of Eq.(8) is

Nf(b)={3,ifbD1D2D4,2,ifb((D1D2)\D4)((D2D4)\D1),1,ifb=±1,orbD3,orb(D2\(D1D4))(D1\(D2D4))(D4\(D1D2)),0,otherwise.\displaystyle N_{f}(b)=\left\{\begin{array}[]{ll}3,&\mathrm{if}~{}b\in D_{1}\cap D_{2}\cap D_{4},\\ 2,&\mathrm{if}~{}b\in(\left(D_{1}\cap D_{2}\right)\backslash D_{4})\cup(\left(D_{2}\cap D_{4}\right)\backslash D_{1}),\\ 1,&\mathrm{if}~{}b=\pm 1,~{}\mathrm{or}~{}b\in D_{3},\mathrm{or}~{}b\in(D_{2}\backslash\left(D_{1}\cup D_{4}\right))\\ &~{}\cup(D_{1}\backslash\left(D_{2}\cup D_{4}\right))\cup(D_{4}\backslash\left(D_{1}\cup D_{2}\right)),\\ 0,&\mathrm{otherwise}.\end{array}\right.

Now we calculate the cardinality of the sets D1D2D4D_{1}\cap D_{2}\cap D_{4}, (D1D2)\D4(D_{1}\cap D_{2})\backslash D_{4}, (D2D4)\D1(D_{2}\cap D_{4})\backslash D_{1} to determine (1)(-1)-differential spectrum of ff.

First, note that

D1D2D4={b𝔽pn:η(2b2)=η(2b1)=1,η(2b+1)=η(2b+2)=1}.D_{1}\cap D_{2}\cap D_{4}=\{b\in\mathbb{F}_{p^{n}}:\eta(2b-2)=\eta(2b-1)=1,~{}\eta(2b+1)=\eta(2b+2)=-1\}.

By the definition of the (1)(-1)-differential spectrum of ff, we have

ω31={}_{-1}\omega_{3}= #(D1D2D4)\displaystyle\#(D_{1}\cap D_{2}\cap D_{4})
=\displaystyle= 116b𝔽pn((1+η(2b2))(1+η(2b1))(1η(2b+1))(1η(2b+2)))\displaystyle\frac{1}{16}\sum_{b\in\mathbb{F}_{p^{n}}}((1+\eta(2b-2))(1+\eta(2b-1))(1-\eta(2b+1))(1-\eta(2b+2)))
116b{±1,±12}((1+η(2b2))(1+η(2b1))(1η(2b+1))(1η(2b+2)))\displaystyle-\frac{1}{16}\sum_{b\in\{\pm 1,\pm\frac{1}{2}\}}((1+\eta(2b-2))(1+\eta(2b-1))(1-\eta(2b+1))(1-\eta(2b+2)))
=\displaystyle= M1M2.\displaystyle M_{1}-M_{2}.

where M1M_{1} and M2M_{2} denote by the two summations above respectively.

It can be easy to see that M2=0M_{2}=0. Expanding the expression of M1M_{1}, then we have that

M1=116(pn+3λp,n(1)+2λp,n(2)+1)M_{1}=\frac{1}{16}(p^{n}+3\lambda_{p,n}^{(1)}+2\lambda_{p,n}^{(2)}+1)

after a calculation based on the fact in Lemmas 1, 2 and 4.
Therefore, ω31=M1M2=116(pn+3λp,n(1)+2λp,n(2)+1){}_{-1}\omega_{3}=M_{1}-M_{2}=\frac{1}{16}(p^{n}+3\lambda_{p,n}^{(1)}+2\lambda_{p,n}^{(2)}+1).

Similarly, we have

(D1D2)\D4={b𝔽pn:η(2b2)=η(2b±1)=1,η(2b+2)=1},(D_{1}\cap D_{2})\backslash D_{4}=\{b\in\mathbb{F}_{p^{n}}:\eta(2b-2)=\eta(2b\pm 1)=1,~{}\eta(2b+2)=-1\},
(D2D4)\D1={b𝔽pn:η(2b+2)=η(2b±1)=1,η(2b2)=1},(D_{2}\cap D_{4})\backslash D_{1}=\{b\in\mathbb{F}_{p^{n}}:\eta(2b+2)=\eta(2b\pm 1)=-1,~{}\eta(2b-2)=1\},

and

ω21={}_{-1}\omega_{2}= #((D1D2)\D4)+#((D2D4)\D1)\displaystyle\#((D_{1}\cap D_{2})\backslash D_{4})+\#((D_{2}\cap D_{4})\backslash D_{1})
=\displaystyle= 116(pnλp,n(1)2λp,n(2)+1)+116(pnλp,n(1)2λp,n(2)+1)\displaystyle\frac{1}{16}(p^{n}-\lambda_{p,n}^{(1)}-2\lambda_{p,n}^{(2)}+1)+\frac{1}{16}(p^{n}-\lambda_{p,n}^{(1)}-2\lambda_{p,n}^{(2)}+1)
=\displaystyle= 18(pnλp,n(1)2λp,n(2)+1).\displaystyle\frac{1}{8}(p^{n}-\lambda_{p,n}^{(1)}-2\lambda_{p,n}^{(2)}+1).

Plugging ω21{}_{-1}\omega_{2} and ω31{}_{-1}\omega_{3} into Eq.(5), we get

{1ω1=116(9pn5λp,n(1)+2λp,n(2)7),1ω0=14(pn+λp,n(1)+1).\displaystyle\left\{\begin{array}[]{ll}_{-1}\omega_{1}=\frac{1}{16}(9p^{n}-5\lambda_{p,n}^{(1)}+2\lambda_{p,n}^{(2)}-7),\\ _{-1}\omega_{0}=\frac{1}{4}(p^{n}+\lambda_{p,n}^{(1)}+1).\end{array}\right.

In conclusion, which completes the proof.

Corollary 1.

Let f(x)=xpn+32f(x)=x^{\frac{p^{n}+3}{2}} over 𝔽pn\mathbb{F}_{p^{n}}, where pn3(mod4)p^{n}\equiv 3\pmod{4} and p3p\neq 3 is an odd prime. Then

Δf1{}_{-1}\Delta_{f} ={2,if pn{7,19,23},3,otherwise.\displaystyle=\left\{\begin{array}[]{ll}2,&\hbox{if $p^{n}\in\{7,19,23\}$,}\\ 3,&\hbox{$\mathrm{otherwise}$.}\end{array}\right.

Proof.

By Lemma 3, we obtain

ω31=116(pn+3λp,n(1)+2λp,n(2)+1)116(pn10pn2+1)>0{}_{-1}\omega_{3}=\frac{1}{16}(p^{n}+3\lambda_{p,n}^{(1)}+2\lambda_{p,n}^{(2)}+1)\geq\frac{1}{16}(p^{n}-10p^{\frac{n}{2}}+1)>0

when pn103p^{n}\geq 103. Moreover, after calculation by Magma program, we find that the Δf1=2{}_{-1}\Delta_{f}=2 (resp. 3) for pn{7,19,23}p^{n}\in\{7,19,23\} (resp. pn{11,31,43,47,59,67,71,83}p^{n}\in\{11,31,43,47,59,67,71,83\}).

Therefore, we get ff is an APcN power function when pn{7,19,23}p^{n}\in\{7,19,23\}, and a differentially (1,3)(-1,3)-uniform power function otherwise.

There are some examples as follows, which are consistent with that computed directly by Magma program.

Example 3.

Let n=1n=1. If p=7,19,23p=7,19,23, then f(x)=xpn+32f(x)=x^{\frac{p^{n}+3}{2}} over 𝔽p\mathbb{F}_{p} is APcN for c=1c=-1 with (1)(-1)-differential spectrum

𝕊\displaystyle\mathbb{S} ={1ω0=2,ω11=3,ω21=2},\displaystyle=\{_{-1}\omega_{0}=2,~{}{}_{-1}\omega_{1}=3,~{}{}_{-1}\omega_{2}=2\},
𝕊\displaystyle\mathbb{S} ={1ω0=4,ω11=11,ω21=4},\displaystyle=\{_{-1}\omega_{0}=4,~{}{}_{-1}\omega_{1}=11,~{}{}_{-1}\omega_{2}=4\},
𝕊\displaystyle\mathbb{S} ={1ω0=4,ω11=15,ω21=4},\displaystyle=\{_{-1}\omega_{0}=4,~{}{}_{-1}\omega_{1}=15,~{}{}_{-1}\omega_{2}=4\},

respectively.

Example 4.

Let p=7p=7 and n=3n=3. Then f(x)=x173f(x)=x^{173} over 𝔽73\mathbb{F}_{7^{3}} is differentially (1,3)(-1,3)-uniform with (1)(-1)-differential spectrum

𝕊={1ω0=86,ω11=195,ω21=38,ω31=24}.\mathbb{S}=\{_{-1}\omega_{0}=86,~{}{}_{-1}\omega_{1}=195,~{}{}_{-1}\omega_{2}=38,~{}{}_{-1}\omega_{3}=24\}.

3.2 pn1(mod4)p^{n}\equiv 1\pmod{4} and p3p\neq 3

In this subsection, we will focus on studying the (1)(-1)-differential spectrum of the power function f(x)=xdf(x)=x^{d} over 𝔽pn\mathbb{F}_{p^{n}}, where d=pn+32d=\frac{p^{n}+3}{2}, pn1(mod4)p^{n}\equiv 1\pmod{4} and p3p\neq 3. It is clear that

gcd(pn+32,pn1)=2or 4.\gcd(\frac{p^{n}+3}{2},~{}p^{n}-1)=2\ \mathrm{or}\ 4.

This subsection adopts the method from [18, P.11-13], and we only provide a sketch due to the process is quite complicated and miscellaneous.

Note that x0x_{0} is a solution of (x+1)d+xd=b(x+1)^{d}+x^{d}=b if and only if x01-x_{0}-1 is a solution of (x+1)d+xd=b(x+1)^{d}+x^{d}=b since dd is even. We assert that δf1(b){}_{-1}\delta_{f}(b) is even except for

b=(12+1)d+(12)d=η(2)2.b=(-\frac{1}{2}+1)^{d}+(-\frac{1}{2})^{d}=\frac{\eta(2)}{2}.

In the following, we will investigate the value of δf1(η(2)2){}_{-1}\delta_{f}(\frac{\eta(2)}{2}). Since the proof of Lemma 7 is similar to that of [18, Lemma 8], we omit the proof here.

Lemma 7.

With the notation as above, we have

δf1(η(2)2){}_{-1}\delta_{f}(\frac{\eta(2)}{2}) ={3,if η(2)=η(3)=1η(1+22)=1 or η(2)=η(3)=1,1,otherwise.\displaystyle=\left\{\begin{array}[]{ll}3,&\hbox{if $\eta(2)=\eta(3)=1$, $\eta(\frac{-1+\sqrt{-2}}{2})=-1$ or $\eta(2)=\eta(3)=-1$,}\\ 1,&\hbox{$\mathrm{otherwise}$.}\end{array}\right.

From the discussion above, one can immediately derive the values of ω11{}_{-1}\omega_{1} and ω31{}_{-1}\omega_{3} in the following corollary.

Corollary 2.

With the notation as above, we have ω11=0,ω31=1{}_{-1}\omega_{1}=0,{}_{-1}\omega_{3}=1 if η(2)=η(3)=1\eta(2)=\eta(3)=1, η(1+22)=1\eta(\frac{-1+\sqrt{-2}}{2})=-1 or η(2)=η(3)=1\eta(2)=\eta(3)=-1, and ω11=1,ω31=0{}_{-1}\omega_{1}=1,{}_{-1}\omega_{3}=0 otherwise.

To determine the (1)(-1)-differential spectrum of ff, it remains to calculate N41{}_{-1}N_{4} in Eq.(7), and this calculation method is similar to that of [18, Theorem 11].

Moreover, when c=1c=-1, Eq.(7) can be rewritten as

{x1(x3)+(x2)x4=0,x1d(x3)d+(x2)dx4d=0,\left\{\begin{array}[]{ll}x_{1}-(-x_{3})+(-x_{2})-x_{4}&=0,\\ x_{1}^{d}-(-x_{3})^{d}+(-x_{2})^{d}-x_{4}^{d}&=0,\end{array}\right.

so we use Lemma 5 to calculate that N41{}_{-1}N_{4} is

N41=1+18(pn1)(21pn+7λp,n(1)2λp,n(2)+13),{}_{-1}N_{4}=1+\frac{1}{8}(p^{n}-1)(21p^{n}+7\lambda^{(1)}_{p,n}-2\lambda^{(2)}_{p,n}+13),

where λp,n(1)\lambda^{(1)}_{p,n} and λp,n(2)\lambda^{(2)}_{p,n} are defined in Eqs.EQ.(2) and Eq.(3), respectively.

Based on the previous preparation work, we are now in a position to determine 𝕊\mathbb{S} of ff, where λp,n(1)\lambda^{(1)}_{p,n} and λp,n(2)\lambda^{(2)}_{p,n} are defined in Eqs.(2) and (3), respectively.

Theorem 2.

Let ff be defined as above. If gcd(d,pn1)=\gcd(d,~{}p^{n}-1)= 2 (resp. 4), then the (1)(-1)-differential spectrum of ff is

𝕊={ω01\displaystyle\mathbb{S}=\{{}_{-1}\omega_{0} =164(37pn+7λp,n(1)2λp,n(2)+5),\displaystyle=\frac{1}{64}(37p^{n}+7\lambda^{(1)}_{p,n}-2\lambda^{(2)}_{p,n}+5),
ω21\displaystyle{}_{-1}\omega_{2} =132(11pn7λp,n(1)+2λp,n(2)21),\displaystyle=\frac{1}{32}(11p^{n}-7\lambda^{(1)}_{p,n}+2\lambda^{(2)}_{p,n}-21),
ω31\displaystyle{}_{-1}\omega_{3} =1,\displaystyle=1,
ω41\displaystyle{}_{-1}\omega_{4} =164(5pn+7λp,n(1)2λp,n(2)27)}\displaystyle=\frac{1}{64}(5p^{n}+7\lambda^{(1)}_{p,n}-2\lambda^{(2)}_{p,n}-27)\}

(resp.

𝕊={ω01\displaystyle\mathbb{S}=\{{}_{-1}\omega_{0} =164(37pn+7λp,n(1)2λp,n(2)11),\displaystyle=\frac{1}{64}(37p^{n}+7\lambda^{(1)}_{p,n}-2\lambda^{(2)}_{p,n}-11),
ω21\displaystyle{}_{-1}\omega_{2} =132(11pn7λp,n(1)+2λp,n(2)5),\displaystyle=\frac{1}{32}(11p^{n}-7\lambda^{(1)}_{p,n}+2\lambda^{(2)}_{p,n}-5),
ω31\displaystyle{}_{-1}\omega_{3} =1,\displaystyle=1,
ω41\displaystyle{}_{-1}\omega_{4} =164(5pn+7λp,n(1)2λp,n(2)43)}\displaystyle=\frac{1}{64}(5p^{n}+7\lambda^{(1)}_{p,n}-2\lambda^{(2)}_{p,n}-43)\}

when η(2)=η(3)=1\eta(2)=\eta(3)=1, η(1+22)=1\eta(\frac{-1+\sqrt{-2}}{2})=-1 or η(2)=η(3)=1\eta(2)=\eta(3)=-1, and is

𝕊={ω01\displaystyle\mathbb{S}=\{{}_{-1}\omega_{0} =164(37pn+7λp,n(1)2λp,n(2)27),\displaystyle=\frac{1}{64}(37p^{n}+7\lambda^{(1)}_{p,n}-2\lambda^{(2)}_{p,n}-27),
ω11\displaystyle{}_{-1}\omega_{1} =1,\displaystyle=1,
ω21\displaystyle{}_{-1}\omega_{2} =132(11pn7λp,n(1)+2λp,n(2)21),\displaystyle=\frac{1}{32}(11p^{n}-7\lambda^{(1)}_{p,n}+2\lambda^{(2)}_{p,n}-21),
ω41\displaystyle{}_{-1}\omega_{4} =164(5pn+7λp,n(1)2λp,n(2)+5)}\displaystyle=\frac{1}{64}(5p^{n}+7\lambda^{(1)}_{p,n}-2\lambda^{(2)}_{p,n}+5)\}

(resp.

𝕊={ω01\displaystyle\mathbb{S}=\{{}_{-1}\omega_{0} =164(37pn+7λp,n(1)2λp,n(2)43),\displaystyle=\frac{1}{64}(37p^{n}+7\lambda^{(1)}_{p,n}-2\lambda^{(2)}_{p,n}-43),
ω11\displaystyle{}_{-1}\omega_{1} =1,\displaystyle=1,
ω21\displaystyle{}_{-1}\omega_{2} =132(11pn7λp,n(1)+2λp,n(2)5),\displaystyle=\frac{1}{32}(11p^{n}-7\lambda^{(1)}_{p,n}+2\lambda^{(2)}_{p,n}-5),
ω41\displaystyle{}_{-1}\omega_{4} =164(5pn+7λp,n(1)2λp,n(2)11)}\displaystyle=\frac{1}{64}(5p^{n}+7\lambda^{(1)}_{p,n}-2\lambda^{(2)}_{p,n}-11)\}

otherwise.

Corollary 3.

Let f(x)=xpn+32f(x)=x^{\frac{p^{n}+3}{2}} over 𝔽pn\mathbb{F}_{p^{n}}, where pn1(mod4)p^{n}\equiv 1\pmod{4} and p3p\neq 3 is an odd prime. Then Δf1=4{}_{-1}\Delta_{f}=4.

Proof.

By Lemma 3, we obtain

ω41164(5pn+7λp,n(1)2λp,n(2)43)164(5pn18pn243)>0{}_{-1}\omega_{4}\geq\frac{1}{64}(5p^{n}+7\lambda^{(1)}_{p,n}-2\lambda^{(2)}_{p,n}-43)\geq\frac{1}{64}(5p^{n}-18p^{\frac{n}{2}}-43)>0

when pn29p^{n}\geq 29. The numerical results show that ω411{}_{-1}\omega_{4}\geq 1 for pn{13,17,25}p^{n}\in\{13,17,25\}.

The case pn=5p^{n}=5 can be calculated by computing directly the (1)(-1)-differential spectrum of f(x)=x4f(x)=x^{4} over 𝔽5\mathbb{F}_{5}, which is

𝕊={ω01=3,ω21=1,ω31=1},\mathbb{S}=\{{}_{-1}\omega_{0}=3,~{}{}_{-1}\omega_{2}=1,~{}{}_{-1}\omega_{3}=1\},

and hence Δf1=4{}_{-1}\Delta_{f}=4 due to gcd(d,pn1)=4\gcd(d,~{}p^{n}-1)=4.

Therefore, we conclude that the Δf1=4{}_{-1}\Delta_{f}=4. This completes the proof.

Example 5.

Let p=7p=7 and n=4n=4. Then f(x)=x1202f(x)=x^{1202} over 𝔽74\mathbb{F}_{7^{4}} is differentially (1,4)(-1,4)-uniform with (1)(-1)-differential spectrum

𝕊={1ω0=1374,ω11=1,ω21=852,ω41=174}.\mathbb{S}=\{_{-1}\omega_{0}=1374,~{}{}_{-1}\omega_{1}=1,~{}{}_{-1}\omega_{2}=852,~{}{}_{-1}\omega_{4}=174\}.

Example 6.

Let p=29p=29 and n=1n=1. Then f(x)=x16f(x)=x^{16} over 𝔽29\mathbb{F}_{29} is differentially (1,4)(-1,4)-uniform with (1)(-1)-differential spectrum

𝕊={1ω0=16,ω21=11,ω31=1,ω41=1}.\mathbb{S}=\{_{-1}\omega_{0}=16,~{}{}_{-1}\omega_{2}=11,~{}{}_{-1}\omega_{3}=1,~{}{}_{-1}\omega_{4}=1\}.

The above two examples both are consistent with that computed directly by Magma program.

4 The cc-differential uniformity of f(x)=xpn+32f(x)=x^{\frac{p^{n}+3}{2}} over 𝔽pn\mathbb{F}_{p^{n}}

In this section, we give the following result on the cc-differential uniformity of ff, and we omit the proof here since it is similar to that of [17, Theorem 12].

Theorem 3.

Let f(x)=xdf(x)=x^{d} be a power function over 𝔽pn\mathbb{F}_{p^{n}}, where pp is an odd prime and d=pn+32d=\frac{p^{n}+3}{2} is a positive integer. For ±1c𝔽pn\pm 1\neq c\in\mathbb{F}_{p^{n}}, we have Δfc9{}_{c}\Delta_{f}\leq 9.

5 Concluding remarks

In this paper, we first investigated the (1)(-1)-differential spectrum of f(x)=xpn+32f(x)=x^{\frac{p^{n}+3}{2}} over 𝔽pn\mathbb{F}_{p^{n}}, where p3p\neq 3 is an odd prime. In fact, we prove that if pn3(mod4)p^{n}\equiv 3\pmod{4}, then Δf1=3{}_{-1}\Delta_{f}=3 except for pn{7,19,23}p^{n}\in\{7,19,23\} where ff is an APcN function, and if pn1(mod4)p^{n}\equiv 1\pmod{4}, the (1)(-1)-differential uniformity of ff is equal to 4. The results indicate that the (1)(-1)-differential spectrum of ff has closely connection with two character sums λp,n(1)\lambda^{(1)}_{p,n} and λp,n(2)\lambda^{(2)}_{p,n}. Meanwhile, the character sums can be evaluated by employing the theory of elliptic curves over finite fields. Finally, we obtained an upper bound of the cc-differential uniformity of ff.

References

  • [1] Eli Biham and Adi Shamir. Differential cryptanalysis of DES-like cryptosystems. J. Cryptology, 4(1):3–72, 1991.
  • [2] Céline Blondeau, Anne Canteaut, and Pascale Charpin. Differential properties of power functions. Int. J. Inf. Coding Theory, 1(2):149–170, 2010.
  • [3] Nikita Borisov, Monica Chew, Robert Johnson, and David Wagner. Multiplicative differentials. In Fast Software Encryption, 9th International Workshop, FSE 2002, Leuven, Belgium, February 4-6, 2002, Revised Papers, pages 17–33, 2002.
  • [4] På l Ellingsen, Patrick Felke, Constanza Riera, Pantelimon Stănică, and Anton Tkachenko. CC-differentials, multiplicative uniformity, and (almost) perfect cc-nonlinearity. IEEE Trans. Inform. Theory, 66(9):5781–5789, 2020.
  • [5] Rudolf Lidl and Harald Niederreiter. Finite fields, volume 20 of Encyclopedia of Mathematics and its Applications. Cambridge University Press, Cambridge, second edition, 1997. With a foreword by P. M. Cohn.
  • [6] Sihem Mesnager, Bimal Mandal, and Mounira Msahli. Survey on recent trends towards generalized differential and boomerang uniformities. Cryptogr. Commun, 14(4):691–735, 2022.
  • [7] Sihem Mesnager, Constanza Riera, Pantelimon Stănică, Haode Yan, and Zhengchun Zhou. Investigations on cc-(almost) perfect nonlinear functions. IEEE Trans. Inform. Theory, 67(10):6916–6925, 2021.
  • [8] Kaisa Nyberg. Differentially uniform mappings for cryptography. In Advances in cryptology—EUROCRYPT ’93 (Lofthus, 1993), volume 765 of Lecture Notes in Comput. Sci., pages 55–64. Springer, Berlin, 1994.
  • [9] Tingting Pang, Nian Li, Xiangyong Zeng, and Haiying Zhu. A note on the cc-differential spectrum of an APcN{\rm AP}c{\rm N} function. Adv. Math. Commun, 16(4):935–945, 2022.
  • [10] Constanza Riera, Pantelimon Stanica, and Haode Yan. The cc-differential spectrum of xxpn+12x\mapsto x^{\frac{p^{n}+1}{2}} in finite fields of odd characteristics, 2022.
  • [11] Joseph H. Silverman. The arithmetic of elliptic curves, volume 106 of Graduate Texts in Mathematics. Springer, Dordrecht, second edition, 2009.
  • [12] Ziran Tu, Nian Li, Yanan Wu, Xiangyong Zeng, Xiaohu Tang, and Yupeng Jiang. On the differential spectrum and the APcN{\rm AP}c{\rm N} property of a class of power functions over finite fields. IEEE Transactions on Information Theory, 69(1):582–597, 2023.
  • [13] Xiaoqiang Wang, Dabin Zheng, and Lei Hu. Several classes of PcN{\rm P}c{\rm N} power functions over finite fields. Discrete Appl. Math, 322:171–182, 2022.
  • [14] Zhexin Wang, Sihem Mesnager, Nian Li, and Xiangyong Zeng. On differential properties of a class of niho-type power function, 2023.
  • [15] Yanan Wu, Nian Li, and Xiangyong Zeng. New PcN{\rm P}c{\rm N} and APcN{\rm AP}c{\rm N} functions over finite fields. Des. Codes Cryptogr, 89(11):2637–2651, 2021.
  • [16] Haode Yan. On (-1)-differential uniformity of ternary APN power functions. Cryptogr. Commun, 14(2):357–369, 2022.
  • [17] Haode Yan, Sihem Mesnager, and Xiantong Tan. On the differential spectrum of a class of apn power functions over odd characteristic finite fields and their cc-differential properties, 2022.
  • [18] Haode Yan, Sihem Mesnager, and Xiantong Tan. The complete differential spectrum of a class of power permutations over odd characteristic finite fields. IEEE Transactions on Information Theory, 69(11):7426–7438, 2023.
  • [19] Haode Yan and Kun Zhang. On the cc-differential spectrum of power functions over finite fields. Des. Codes Cryptogr, 90(10):2385–2405, 2022.
  • [20] Zhengbang Zha and Lei Hu. Some classes of power functions with low cc-differential uniformity over finite fields. Des. Codes Cryptogr, 89(6):1193–1210, 2021.