<!DOCTYPE article PUBLIC "-//NLM//DTD JATS (Z39.96) Journal Archiving and Interchange DTD v1.0 20120330//EN" "JATS-archivearticle1.dtd">
<article xmlns:xlink="http://www.w3.org/1999/xlink">
  <front>
    <journal-meta />
    <article-meta>
      <title-group>
        <article-title>Correlation Immunity of Many-Valued Logic Component Functions of Modern Cryptographic Algorithm S-Boxes</article-title>
      </title-group>
      <contrib-group>
        <contrib contrib-type="author">
          <string-name>Nadiia Kazakova</string-name>
          <xref ref-type="aff" rid="aff1">1</xref>
        </contrib>
        <contrib contrib-type="author">
          <string-name>Artem Sokolov</string-name>
          <xref ref-type="aff" rid="aff0">0</xref>
        </contrib>
        <contrib contrib-type="author">
          <string-name>Alexander Troyanskiy</string-name>
          <xref ref-type="aff" rid="aff2">2</xref>
        </contrib>
        <aff id="aff0">
          <label>0</label>
          <institution>Department of Cybersecurity and Software, Odessa National Polytechnic University</institution>
          ,
          <addr-line>Odessa, 65044</addr-line>
          ,
          <country country="UA">Ukraine</country>
        </aff>
        <aff id="aff1">
          <label>1</label>
          <institution>Department of Information Technologies, Odessa State Environmental University</institution>
          ,
          <addr-line>Odessa, 65016</addr-line>
          ,
          <country country="UA">Ukraine</country>
        </aff>
        <aff id="aff2">
          <label>2</label>
          <institution>Department of Radioelectronic and Telecommunication Systems, Odessa National Polytechnic University</institution>
          ,
          <addr-line>Odessa, 65044</addr-line>
          ,
          <country country="UA">Ukraine</country>
        </aff>
      </contrib-group>
      <abstract>
        <p>The development of modern cryptanalysis methods, in particular, with the use of many-valued logic functions, leads to the need for a more detailed research of the correlation properties of Sboxes of modern cryptographic algorithms. In this paper, we introduce indicators for the maximum and integral deviation of a many-valued logic functions on the basis of criterion for the independence of its output from the input variables. These indicators are a convenient tool for comparative analysis of the correlation properties of S-boxes of various lengths when they are represented using the many-valued logic functions with different bases q. The research and comparative analysis of the S-boxes of the AES and Kalyna cryptographic algorithms was performed, which showed a general tendency of a decreasing of their correlation properties with an increase in the value of representation base q, and also made it possible to establish that for all values of the representation base q, the correlation properties of the Kalyna cryptographic algorithm S-box are weaker than the correlation properties of the AES cryptographic algorithm S-box.</p>
      </abstract>
      <kwd-group>
        <kwd>1 Cryptography</kwd>
        <kwd>many-valued logic function</kwd>
        <kwd>correlation immunity</kwd>
      </kwd-group>
    </article-meta>
  </front>
  <body>
    <sec id="sec-1">
      <title>-</title>
      <p>and
problem</p>
    </sec>
    <sec id="sec-2">
      <title>1. Introduction statement</title>
      <p>Block symmetric ciphers are one of the most
important components of modern cybersecurity
systems, the task of which is to make it impossible
to access transmitted or stored information
without knowing the key. Nevertheless, the rapid
increase in the computing power of modern
information systems, development of the
mathematical methods of cryptanalysis, as well as
the prospects for the creation of quantum
computers lead to the need for continuous
improvement of the structure and components of
modern block symmetric ciphers.</p>
      <p>The issues of improving modern cryptographic
algorithms are inextricably linked with the theory
of estimation of their cryptographic quality. The
modern theory of estimation of the cryptographic
quality of block symmetric ciphers involves the
representation of their constituent parts (first of
all, S-boxes) using component Boolean functions
[1, 2]. Further, cryptographic quality criteria are
applied to the obtained Boolean functions, each of
which reflects the ability of the Boolean function
to resist one or another cryptanalysis attack, as
well as to provide a sufficient level of diffusion
and confusion [3].</p>
      <p>However, publications in the field of
cryptography [4] show an increasing need to
research all possible representations of S-boxes,
in particular, using functions of many-valued
logic. Research [5] is devoted to the method of
estimation of the nonlinearity, as well as
avalanche characteristics of S-boxes of modern
cryptographic algorithms, represented by
functions of many-valued logic.</p>
      <p>Nevertheless, the results of the analysis of the
correspondence of the constructions of modern
cryptographic algorithms to the criterion of
correlation immunity of functions of many-valued
logic are not presented in the open sources.</p>
      <p>The purpose of this paper is to perform a
comparative analysis of the correlation immunity
of the cryptographic algorithms AES (USA) and
Kalyna (Ukraine) when they are represented by
the many-valued logic functions.
2. Correlation immunity of
manyvalued logic functions</p>
      <p>The definition of the correlation immunity of
Boolean functions is widely known [6], which can
be specified through the definition of a
subfunction of a Boolean function.</p>
      <p>Definition 1 [6]. A Boolean function f (x) ,
x Vk , is called as correlation-immune of order
m , 1  m  k if its output y  f (x1k ) and any set
of m its input variables are statistically
independent.</p>
      <p>Definition 2 [6]. A subfunction of a Boolean
function f (x) , x Vk , is a function f  obtained
by substitution in f constants "0" or "1" instead
of a part of variables. If we substitute constants
 i 1 ,  , i s in the function f instead of variables
xi ,  , xi , respectively, then the resulting
1 s
subfunction is denoted f i1 ,,i s . If a constant is
xi1 ,,xis
not substituted for a variable xi , then xi is called
as a free variable.</p>
      <p>For example, let a Boolean function
f (x1, x2 , x3) be given, then its subfunctions will
be
f (x1, x2 ,0) ,
f (x1, x2 ,1) ,
f (x1,0,0) ,
f (x1,0,1) , etc.</p>
      <p>Definition 3 [6]. Boolean function f (x) ,
x Vk , is called correlation-immune of order m ,
1  m  k if weight is equal to
wt( f )  wt( f ) / 2m , for any of its subfunctions
f  of k  m variables.</p>
      <p>In [7], the theoretical foundations for
estimating the correlation immunity of functions
of many-valued logic were developed, while in
[8], definitions of the independence of the output
of a 3-function from its input variables, as well as
the definition of the correlation immunity of a
3function were introduced. The basis of these
definitions is the definition of the imbalance of
functions of many-valued logic.</p>
      <p>Consider an arbitrary sequence over the
alphabet {0,1,..., q 1}</p>
      <p>
        fi {0,1,..., q 1}, i  0,1,..., N 1. (
        <xref ref-type="bibr" rid="ref1">1</xref>
        )
Note that the elements of a given sequence (
        <xref ref-type="bibr" rid="ref1">1</xref>
        )
can be represented in exponential form by an
unambiguous transformation
0  z0  e j2qπ0 , 1  z1  e j2qπ1 ,...,
q 1  zq1  e
j2qπ(q1)
.
      </p>
      <p>
        (
        <xref ref-type="bibr" rid="ref2">2</xref>
        )
      </p>
      <p>
        For a given sequence (
        <xref ref-type="bibr" rid="ref1">1</xref>
        ), it is possible to
introduce a vector K  {K0 , K1,..., Kq1} , where
the coefficients Ku characterize the number of
occurrences of a character u {0,1,..., q 1} in the
sequence f .
      </p>
      <p>
        Definition 4. An imbalance of a sequence f
is the absolute value of the sum of element-wise
products of vector K elements by the
corresponding elements of the exponential
alphabet {z0 , z1,..., zq1}
( f )  K0 z0  K1z1  ...  Kq1zq1 .
(
        <xref ref-type="bibr" rid="ref3">3</xref>
        )
      </p>
      <p>The definitions of the independence of the
output of a 3-function from its input variables, as
well as the definition of the correlation immunity
of a 3-function, introduced in [8] using
Definition 4 of the imbalance of a many-valued
logic function can be generalized to the case of
qfunctions for an arbitrary value of q.</p>
      <p>Definition 5 [8]. It is said that the output of a
q-function f (x) does not depend on the group of
its input variables {xi} , i  1,..., m if, when
any constants
of these
substituting
 i 1 ,  , i s {0,1,...q 1} instead
variables, the imbalance of the subfunctions
obtained in this way is ( f )  qmf .</p>
      <p>Note, however, that the correlation immunity
of component Boolean functions, as well as
component many-valued logic functions, is a
rather stringent requirement that is not met for
most S-boxes used in practice. This circumstance
poses the task of developing a mathematical
apparatus for performing a comparative analysis
of the degree of correspondence of S-boxes to the
correlation immunity criterion of component
qfunctions.</p>
      <p>The Definition 5 gives us the possibility to
estimate the degree of deviation of the
manyvalued logic functions and S-boxes from the
criterion of correlation immunity by introducing
two basic indicators of cryptographic quality: the
maximum and integral deviation from the
criterion of independence of the output of
manyvalued logic functions from their input.</p>
      <p>Let's consider these indicators on a specific
example, and then apply them to specific
cryptographic algorithms.</p>
      <p>Let an S-box of length N  16 be given, which
can be represented in the form of four component
Boolean functions, as well as in the form of two
component 4-functions.</p>
      <p>S 12 0 2141311 3 5 1 7 15 9 10 6 4 8
f20 0 0 0 0 1 1 1 1 1 1 1 1 0 0 0 0
f21 0 0 1 1 0 1 1 0 0 1 1 0 1 1 0 0
f22 1 0 0 1 1 0 0 1 0 1 1 0 0 1 1 0
f23 1 0 0 1 1 1 0 0 0 0 1 1 1 0 0 1
f40 0 0 2 2 1 3 3 1 1 3 3 1 2 2 0 0
f41 3 0 0 3 3 2 0 1 0 1 3 2 2 1 1 2</p>
      <p>
        We begin our research with the first
component Boolean function f20 , for which, in
accordance with Definition 3, we find all its
subfunctions of three variables
wt( f20 (x1, x2 , x3 , 0)  [00111100])  4;
(
        <xref ref-type="bibr" rid="ref6">6</xref>
        )
(
        <xref ref-type="bibr" rid="ref7">7</xref>
        )
functions. Let us find the subfunctions of the
component 4-function f40 of the S-box (
        <xref ref-type="bibr" rid="ref4">4</xref>
        )
( f40 (3, x2 )  [2200])  0.
      </p>
      <p>
        In a similar way, we can find subfunctions for
the component function f41 of the S-box (
        <xref ref-type="bibr" rid="ref4">4</xref>
        )
( f40 (x1, 0)  [0112])  2;
( f40 (x1,1)  [0332])  2;
( f40 (x1, 2)  [2330])  2;
( f40 (x1,3)  [2110])  2;
( f40 (0, x2 )  [0022])  0;
( f40 (1, x2 )  [1331])  0;
( f40 (2, x2 )  [1331])  0;
( f41(x1, 0)  [3302])  2;
( f41(x1,1)  [0211])  2;
( f41(x1, 2)  [0031])  2;
( f41(x1,3)  [3122])  2;
( f41(0, x2 )  [3003])  8;
( f41(1, x2 )  [3201])  0;
( f41(2, x2 )  [0132])  0;
wt( f20 (x1, x2 , 0, x4 )  [00111100])  4;
wt( f20 (x1, 0, x3, x4 )  [00001111])  4;
wt( f20 (0, x2 , x3 , x4 )  [00001111])  4;
wt( f20 (x1, x2 , x3 ,1)  [00111100])  4;
wt( f20 (x1, x2 ,1, x4 )  [00111100])  4;
wt( f20 (x1,1, x3 , x4 )  [11110000])  4;
wt( f20 (1, x2 , x3 , x4 )  [11110000])  4.
      </p>
      <p>
        Since the weight of each of the subfunctions
(
        <xref ref-type="bibr" rid="ref5">5</xref>
        ) is equal to 4, i.e. they are balanced, the first
component Boolean function f20 (
        <xref ref-type="bibr" rid="ref4">4</xref>
        ) corresponds
to the criterion of correlation immunity. It is
possible to verify that all other component
Boolean functions of the S-box (
        <xref ref-type="bibr" rid="ref4">4</xref>
        ) also
corresponds to the criterion of correlation
immunity.
      </p>
      <p>
        Consider the S-box (
        <xref ref-type="bibr" rid="ref4">4</xref>
        ) from the point of view
of its possible representation by component
4(
        <xref ref-type="bibr" rid="ref4">4</xref>
        )
(
        <xref ref-type="bibr" rid="ref5">5</xref>
        )
      </p>
      <p>( f41(3, x2 )  [2112])  8.</p>
      <p>
        Analysis of expressions (
        <xref ref-type="bibr" rid="ref6">6</xref>
        ) and (
        <xref ref-type="bibr" rid="ref7">7</xref>
        ) shows that
not all subfunctions of the component 4-function
f40 , as well as subfunctions of the component
function f41 , are balanced, which means that the
S-box (
        <xref ref-type="bibr" rid="ref4">4</xref>
        ) does not correspond to the criterion of
correlation immunity of component 4-functions.
      </p>
      <p>To solve the problem of quantitative
estimating of the compliance of the S-box with the
criterion of correlation immunity, we introduce
the indicators of the maximum and integral
deviation from the compliance with the criterion
of the independence of the S-box output from its
input variables.</p>
      <p>Definition 6. The maximum deviation of an
Sbox from the criterion of independence of the
output from the input variables when it is
represented by component q-functions is the
maximum among all deviations from the criterion
of independence of the output from the input
variables of its component q-functions.</p>
      <p>
        In our case, the S-box (
        <xref ref-type="bibr" rid="ref4">4</xref>
        ) corresponds to the
criterion of correlation immunity of component
Boolean functions, therefore, its maximum
deviation from the criterion of independence of
the output from the input variables when it is
represented by component Boolean functions is
equal to max f2i  0 .
      </p>
      <p>In the case of representation by component
4functions, the maximum deviation from the
criterion for the independence of the output from
the input variables is max f4i  8  2.8284 .</p>
      <p>
        It is obvious that the maximum possible value
of the maximum deviation of the q-function from
the criterion of the independence of the output
from the input variables is the maximum value of
the imbalance of the q-function of k  1 variables,
that is max{max fqi}  N q . In the case of
representing an S-box of length N  16 using
component 4-functions the maximum value of the
maximum deviation from the criterion of
independence of the output from the input
variables would reach max{max f4i} 16 4  4 .
I.e. the maximum deviation from the criterion of
independence of the output from the input
variables among the component functions for our
S-box (
        <xref ref-type="bibr" rid="ref4">4</xref>
        ) is 70.71% of its maximum value.
      </p>
      <p>Definition 7. The integral deviation of the
Sbox from the criterion of independence of the
output from the input variables when it is
represented by component q-functions is the total
value of all deviations from the criterion of
independence of the output from the input
variables of its component q-functions</p>
      <p>
        k
fqi  fqi . (
        <xref ref-type="bibr" rid="ref8">8</xref>
        )
      </p>
      <p>i0</p>
      <p>
        In our example, in view of the compliance of
the S-box (
        <xref ref-type="bibr" rid="ref4">4</xref>
        ) with the criterion of correlation
immunity of component Boolean functions, its
integral deviation from the criterion of the
independence of the output from the input
variables is equal to f2i  0 .
      </p>
      <p>
        In the case of representation using component
4-functions, and using (
        <xref ref-type="bibr" rid="ref8">8</xref>
        ), we obtain that the
integral deviation of the S-box (
        <xref ref-type="bibr" rid="ref4">4</xref>
        ) from the
criterion for the independence of the output from
the input variables is fq0  8 for the first
component 4-function and fq1  13.6569 for the
second component 4-function.
      </p>
      <p>
        It is obvious that the maximum value of the
integral deviation of the q-function from the
criterion for the independence of the output from
the input variables is
max{fqi}  qk1z ,
(
        <xref ref-type="bibr" rid="ref9">9</xref>
        )
where z is the number of subfunctions of the
qfunction.
      </p>
      <p>In our case, the maximum value of the integral
deviation of the q-function from the criterion of
the independence of the output from the input
variables is max{f4i}  4 8  32 , i.e. the
4function f40 is characterized by the integral
deviation from the criterion of the independence
of the output from the input variables equal to
25% of the maximum value, while for the
component function f41 this indicator is 42.68%.</p>
      <p>
        Note also that since the S-box (
        <xref ref-type="bibr" rid="ref4">4</xref>
        ) consists of
two component 4-functions, the total value of the
integral deviation for the two component
4functions is equal to fq0  fq1  21.6569 ,
which is 33.84% of the maximum value equal to
max{S4}  64 .
      </p>
      <p>Note also that for the cryptographic algorithms
AES and Kalyna which are researched in this
paper, the length of the S-boxes used is N  256 ,
respectively, the maximum value of the maximum
deviation of the Boolean function from the
criterion for the independence of the output from
the input variables is
max{max f2i}  256 2 128, while the maximum
value of the maximum deviation of the
4functions from the criterion for the independence
of the output from the input variables is
max{max f4i}  256 4  64 .</p>
      <p>In this case, the maximum value of the
maximum deviation of the 16-function from the
criterion for the independence of the output from
the input variables is
max{max f16i}  256 16 16.</p>
      <p>The maximum value of the integral deviation
of the Boolean function from the criterion for the
independence of the output from the input
variables is max{f2i} 128 16  2048 , while
this indicator for the entire S-box will be
max{S2}  8  2048=16384 .</p>
      <p>The maximum value of the integral deviation
of the 4-function from the criterion for the
independence of the output from the input
variables is max{f4i}  64 16 1024 , while this
indicator for the entire S-box will be
max{S4}  4 1024=4096 .</p>
      <p>The maximum value of the integral deviation
of the 16-function from the criterion for the
independence of the output from the input
variables is max{f16i} 16  32  512 , while this
indicator for the entire S-box will be
max{S16}  2  512=1024 .
3. Indicators of deviation from the
criterion of correlation immunity
of S-boxes of modern ciphers</p>
      <p>In this paper, we research the deviation from
the criterion of correlation immunity of S-boxes
of cryptographic algorithms AES (USA) [9], as
well as Kalyna (Ukraine) [10].</p>
      <p>
        The cryptographic algorithm AES is based on
a Nyberg construction S-box [11], which is
defined using a mapping in the form of
multiplicatively inverse elements of the Galois
field GF (2k )
y  x1 modd[f (z), p],
y, x GF (2k ) ,
(
        <xref ref-type="bibr" rid="ref10">10</xref>
        )
which is generally combined with the affine
transformation
      </p>
      <p>
        b  A  y  a, a,b GF (2k ) , (
        <xref ref-type="bibr" rid="ref11">11</xref>
        )
where the standard AES irreducible over the field
GF (28 ) polynomial is used as a polynomial
f (z),
      </p>
      <p>A is a non-singular affine transformation
matrix,
a is a the shift vector,
p  2 is the characteristic of the extended
Galois field, k  8 , and 01  0 ,</p>
      <p>a,b, x, y are the elements of the extended Galois
field GF (2k ) , which are considered as decimal
numbers, or binary vectors, or polynomials of
degree k  1.</p>
      <p>
        In this paper, we consider the AES S-box of
the Nyberg construction without using the affine
transformation (
        <xref ref-type="bibr" rid="ref11">11</xref>
        ), as well as with its
application.
      </p>
      <p>The indicators of the maximum and integral
deviation from the criterion of independence of
the output vectors from the input variables for the
AES cryptographic algorithm are presented in
Table 1.</p>
      <p>Analysis of the data presented in Table 1
shows a tendency towards an increase in both the
maximum and integral deviation from the
criterion of independence of the output vectors
from the input variables of the AES cipher with an
increase in the representation base q. At the same
time, the use of an affine transformation
insignificantly reduces the growth of this value
(except for the maximum deviation from the
criterion of independence of the output vectors
from the input variables of 4-functions), but the
deviation even in this case is quite strong.</p>
      <p>Ukraine has developed its own block
symmetric cryptographic algorithm Kalyna,
which was adopted as the standard DSTU 7624:
2014 “Information technologies. Cryptographic
data security. Symmetric block transformation
algorithm” [12].</p>
      <p>Today, there are several options for the
implementation of the Kalyna cryptographic
algorithm, which differ in the key length:
Kalyna128, Kalyna-256 and Kalyna-512. However, they
all use the same cryptographic primitives on
which the cryptographic quality of the
cryptographic algorithm relies.</p>
      <p>The Kalyna block symmetric cipher is
characterized by the use of an SP-network and
thus has an AES-like structure. The basis of the
Kalyna cryptographic transformation is its
nonlinear elements, which are four permutations
specified in [12].</p>
      <p>The indicators of the maximum and integral
deviations from the criterion for the independence
of the output vectors from the input variables for
the permutations  0 , 1, 2 , 3 are presented in
Table 2.</p>
      <p>Note at the same time that in view of the fact
that the overall quality of the cipher is determined
by the weakest of its constituent elements [1], in
order to demonstrate the overall quality of the
Kalyna cryptographic algorithm, the smallest
values of the maximum and integral deviations
from the criterion for independence of the output
from the input vectors are selected in the last row
of Table 2.</p>
      <p>Analysis of the data presented in Table 2
shows a steady increase in both the maximum and
integral deviation from the criterion for
independence of the output vectors from the input
variables of the S-boxes of the Kalyna cipher with
an increase in the representation base q.
For the convenience of comparing the correlation
properties of the AES cipher S-box (with affine
transformation) and the Kalyna cipher S-box the
Fig. 1 shows the graphs of changes in the
maximum (a) and integral (b) deviation from the
criterion of independence of the output vectors
from the input variables for these cryptographic
algorithms.</p>
      <p>Analysis of the data presented in Fig. 1 shows
that both the maximum and integral deviation of
the S-boxes of the AES and Kalyna cryptographic
algorithms from the criterion for the
independence of the output vectors from the input
variables show an increase with an increase in the
representation base q. At the same time, the
Kalyna cryptographic algorithm has a
significantly higher level of the maximum and
integral deviation from the criterion for the
independence of the output vectors from the input
variables.
4. Conclusions</p>
      <p>1. The development of cryptanalysis
methods necessitates a more detailed research of
the structure of modern cryptographic algorithms,
not only when they are represented by Boolean
functions, but also when they are represented by
functions of many-valued logic. In this paper, on
the basis of the criterion for the independence of
the output of many-valued logic functions from
their input variables, the indicators of the
maximum and integral deviation from the
criterion for the independence of the output of
many-valued logic functions from their input
variables are introduced. For these indicators, the
maximum possible values for the given N and q
are obtained. These indicators are applicable to
individual functions of many-valued logic, as well
as to S-boxes of various lengths with all their
possible representations, and allow a comparative
analysis of the correlation properties of S-boxes
of modern cryptographic algorithms when they
are represented by functions of many-valued
logic.</p>
      <p>2. The analysis of S-boxes of modern
cryptographic algorithms AES and Kalyna was
performed, which showed that S-boxes of both
ciphers demonstrate a decrease in correlation
properties with an increase in the representation
base q. At the same time, for all values of the
representation base q, the correlation properties of
the Kalyna cryptographic algorithm are weaker
than the correlation properties of the S-box of the
AES cryptographic algorithm.</p>
      <p>3. Research shows the possibilities for
further improvement of the nonlinear
transformation of the Ukrainian cryptographic
algorithm Kalyna due to its representation by
functions of many-valued logic, in conjunction
with the fact that the DSTU 7624: 2014 standard
allows the use of other, more advanced nonlinear
elements, increases the relevance of the
development of new cryptographic constructions,
which are optimal from the point of view of
criteria for the cryptographic quality of functions
of many-valued logic.</p>
    </sec>
    <sec id="sec-3">
      <title>5. References</title>
    </sec>
  </body>
  <back>
    <ref-list>
      <ref id="ref1">
        <mixed-citation>
          [1]
          <string-name>
            <given-names>O.N.</given-names>
            <surname>Zhdanov</surname>
          </string-name>
          ,
          <article-title>Methodology for selecting the key information for a block cipher algorithm</article-title>
          , Moscow, INFRA-M,
          <year>2013</year>
          , 90 p.
        </mixed-citation>
      </ref>
      <ref id="ref2">
        <mixed-citation>
          [2]
          <string-name>
            <given-names>A.V.</given-names>
            <surname>Sokolov</surname>
          </string-name>
          ,
          <article-title>New methods for synthesis of nonlinear transformations of modern ciphers</article-title>
          , Lap Lambert Academic Publishing, Germany,
          <year>2015</year>
          , 100 p.
        </mixed-citation>
      </ref>
      <ref id="ref3">
        <mixed-citation>
          [3]
          <string-name>
            <given-names>C.E.</given-names>
            <surname>Shannon</surname>
          </string-name>
          ,
          <source>A Mathematical Theory of Cryptography, Bell System Technical Memo</source>
          ,
          <year>1945</year>
          , MM 45-110-02.
        </mixed-citation>
      </ref>
      <ref id="ref4">
        <mixed-citation>
          [4]
          <string-name>
            <given-names>T.</given-names>
            <surname>Baigneres</surname>
          </string-name>
          ,
          <string-name>
            <given-names>J.</given-names>
            <surname>Stern</surname>
          </string-name>
          ,
          <string-name>
            <given-names>S.</given-names>
            <surname>Vaudenay</surname>
          </string-name>
          ,
          <article-title>Linear cryptanalysis of non binary ciphers</article-title>
          ,
          <source>International Workshop on Selected Areas in Cryptography</source>
          ,
          <year>2007</year>
          , Springer, Berlin, Heidelberg, pp.
          <fpage>184</fpage>
          -
          <lpage>211</lpage>
          .
        </mixed-citation>
      </ref>
      <ref id="ref5">
        <mixed-citation>
          [5]
          <string-name>
            <given-names>A.V.</given-names>
            <surname>Sokolov</surname>
          </string-name>
          ,
          <string-name>
            <given-names>O.N.</given-names>
            <surname>Zhdanov</surname>
          </string-name>
          ,
          <article-title>Cryptographic constructions based on many-valued logic functions</article-title>
          .
          <source>Monograph</source>
          , Moscow: Scientific Thought,
          <year>2020</year>
          , 192 p.
          <source>doi:10.12737/1045434</source>
        </mixed-citation>
      </ref>
      <ref id="ref6">
        <mixed-citation>
          [6]
          <string-name>
            <given-names>A.A.</given-names>
            <surname>Salnikov</surname>
          </string-name>
          ,
          <string-name>
            <given-names>O.A.</given-names>
            <surname>Logachev</surname>
          </string-name>
          , Boolean Functions in
          <source>Coding Theory &amp; Cryptography</source>
          , Universities Press (India) Private Limited,
          <year>2017</year>
          , 334 p.
        </mixed-citation>
      </ref>
      <ref id="ref7">
        <mixed-citation>
          [7]
          <string-name>
            <given-names>K.</given-names>
            <surname>Gopalakrishnan</surname>
          </string-name>
          ,
          <string-name>
            <given-names>D.R.</given-names>
            <surname>Stinson</surname>
          </string-name>
          ,
          <article-title>Three characterizations of non-binary correlationimmune and resilient functions</article-title>
          ,
          <source>Designs, Codes and Cryptography</source>
          , 5,
          <string-name>
            <surname>P.</surname>
          </string-name>
          241-
          <fpage>251</fpage>
          . doi:
          <volume>10</volume>
          .1007/bf01388386
        </mixed-citation>
      </ref>
      <ref id="ref8">
        <mixed-citation>
          [8]
          <string-name>
            <given-names>A.V.</given-names>
            <surname>Sokolov</surname>
          </string-name>
          ,
          <string-name>
            <given-names>O.N.</given-names>
            <surname>Zhdanov</surname>
          </string-name>
          ,
          <article-title>Correlation immunity of three-valued logic functions</article-title>
          ,
          <source>Journal of Discrete Mathematical Sciences and Cryptography</source>
          ,
          <year>2020</year>
          . P. 1-
          <fpage>17</fpage>
          . doi:
          <volume>10</volume>
          .1080/09720529.
          <year>2020</year>
          .1781882
        </mixed-citation>
      </ref>
      <ref id="ref9">
        <mixed-citation>
          <source>[9] FIPS 197</source>
          ,
          <year>2001</year>
          . [Electronic resource]
          <article-title>Advanced encryption standard</article-title>
          . http://csrc.nist.gov/publications/
        </mixed-citation>
      </ref>
      <ref id="ref10">
        <mixed-citation>
          [10]
          <string-name>
            <given-names>I. D.</given-names>
            <surname>Gorbenko</surname>
          </string-name>
          ,
          <string-name>
            <given-names>R.V.</given-names>
            <surname>Olyynikov</surname>
          </string-name>
          ,
          <string-name>
            <given-names>O.V.</given-names>
            <surname>Kazimirov</surname>
          </string-name>
          et al.,
          <article-title>Symmetric block cipher Kalyna is a new national standard of Ukraine, Radio engineering</article-title>
          ,
          <year>2015</year>
          , (
          <issue>181</issue>
          ), pp.
          <fpage>5</fpage>
          -
          <lpage>22</lpage>
          .
        </mixed-citation>
      </ref>
      <ref id="ref11">
        <mixed-citation>
          [11]
          <string-name>
            <given-names>K.</given-names>
            <surname>Nyberg</surname>
          </string-name>
          ,
          <article-title>Differentially uniform mappings for cryptography, I Advances in cryptology</article-title>
          ,
          <source>Proc. of EUROCRYPT'93</source>
          . Berlin, Heidelberg, New York, Lecture Notes in Compuer Science, Springer-Verlag, vol.
          <volume>765</volume>
          , pp.
          <fpage>55</fpage>
          -
          <lpage>65</lpage>
          ,
          <year>1994</year>
          . doi:
          <volume>10</volume>
          .1007/3-540-48285-
          <issue>7</issue>
          _
          <fpage>6</fpage>
        </mixed-citation>
      </ref>
      <ref id="ref12">
        <mixed-citation>
          <source>[12] DSTU</source>
          <volume>7624</volume>
          :
          <year>2014</year>
          <article-title>Information technologies</article-title>
          .
          <article-title>Cryptographic data security. Symmetric block transformation algorithm</article-title>
          ,
          <source>Ministry of Economic Development of Ukraine</source>
          ,
          <year>2016</year>
          . 221 p.
        </mixed-citation>
      </ref>
    </ref-list>
  </back>
</article>