=Paper=
{{Paper
|id=Vol-2899/paper007
|storemode=property
|title=Method for detecting vulnerabilities of unmanned vehicle interfaces based on continuous values discretization
|pdfUrl=https://ceur-ws.org/Vol-2899/paper007.pdf
|volume=Vol-2899
|authors=Dmitriy Moiseev,Alexey Bryukhovetskiy
}}
==Method for detecting vulnerabilities of unmanned vehicle interfaces based on continuous values discretization==
<pdf width="1500px">https://ceur-ws.org/Vol-2899/paper007.pdf</pdf>
<pre>
Method for detecting vulnerabilities of unmanned vehicle
interfaces based on continuous values discretization
Dmitriy Moiseev1 and Alexey Bryukhovetskiy1
1
    Sevastopol state university, 33 Universitetskaya str., Sevastopol, 299053, Russia


                 Abstract
                 An approach related to the development of methods for ensuring the safety of unmanned
                 vehicles in the smart city information infrastructure is proposed. The method is based on the
                 continuous values discretization of the state vector's features of UMV resources, which
                 include: communication channel, processor, memory. For each of these resources, it is
                 proposed to evaluate the change in such characteristics as the degree of resource load and the
                 speed of its change. The proposed method allows you to build a system of rules for the
                 membership of the analyzed vectors to the specified classes and minimize the conditions
                 number in the generated rules. The problem of ensuring the unmanned vehicles information
                 security operating in the intelligent networks of the smart city transport infrastructure does not
                 lose its relevance due to the fact that modern networks face an unprecedented range of
                 computer threats that lead to a violation of the integrity, confidentiality and availability of
                 resources.

                 Keywords 1
                 UMV resources, vulnerability detection, continuous values discretization, intelligent
                 technology

1. Introduction

    The basis of this article is the material, obtained in the research laboratory of "Intelligent Information
Systems and Critical Computing" at the Department of "Information Technologies and Computer
Systems" of Sevastopol State University within the framework of the RFBR grants (grant No. 19-29-
06015 "Adaptive neural network methods for detecting vulnerabilities in the interfaces of unmanned
vehicles based on artificial immune systems" and grant No. 19-29-06023 "Methods of structural
synthesis of information exchange channels between an unmanned vehicle and a dispatch center based
on stochastic analysis). vector programming with probabilistic criteria"), in which the authors of this
article were co-executors.
    The experience gained so far in setting problems of describing and analyzing vulnerabilities of
information systems of various classes is mostly associated with the analysis of vulnerabilities that
directly affect a certain function of information systems, but the problem of integrating systems and
nesting components give rise to a high degree of variability of solutions and parametric uncertainties of
various types. In the monograph, based on the analysis of the state of the problem, the main vulnerable
elements of UMV information systems are considered; the functional-complete set of models for
evaluating the effectiveness of the protection of UMV information systems is defined; the approach to
variant analysis and selection of vulnerable components based on expert assessments and fuzzy sets is
further developed.
    The problem of ensuring the unmanned vehicles information security operating in the intelligent
networks of the smart city transport infrastructure does not lose its relevance due to the fact that modern

III International Workshop on Modeling, Information Processing and Computing (MIP: Computing-2021), May 28, 2021, Krasnoyarsk,
Russia
EMAIL: dmitriymoiseev@mail.com (Dmitriy Moiseev); a.alexir@mail.ru (Alexey Bryukhovetskiy)
ORCID: 0000-0002-3141-1529 (Dmitriy Moiseev); 0000-0002-2612-2968 (Alexey Bryukhovetskiy)
              © 2021 Copyright for this paper by its authors.
              Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
              CEUR Workshop Proceedings (CEUR-WS.org)



                                                                                    43
networks face an unprecedented range of computer threats that lead to a violation of the integrity,
confidentiality and availability of resources. To date, there are a large number of methods for detecting
vulnerabilities in UMV interfaces, which quite effectively perform detailed researches of the UMV
resources information state and search for intrusions sources. The heterogeneity of applications and
wireless communications in the smart city infrastructure significantly complicates the facilities security
[1]. Therefore, the development and implementation of approaches to the creation of information
technologies that ensure the security of the smart city critical information infrastructure are relevant
and of scientific and practical interest. The need to solve this problem is associated with significant
changes in the field of applied digital technologies in Vanet networks, which use technologies
implemented using interfaces: vehicle-to-vehicle, vehicle-to-infrastructure, vehicle-to-pedestrian,
vehicle-to-grid, vehicle-to-device [2].
    In the works [2, 3] the applied methods and solutions in intelligent transport networks are considered
in order to ensure the safety of the UMV operation. The paper presents the classification of attacks on
UMV and means of ensuring information security in Vanet networks. The requirements for the UMV
architecture and data exchange between smart city infrastructure objects are defined. The creation of
management systems for such tools implies the need to study methods and approaches related not only
to the conceptual organization of such systems architecture, but also to their software implementation.
When developing software, special attention is paid to ensuring UMV security. For autonomous driving
of a vehicle, it is necessary to systematically update its software. Upgrading over a wireless network
can bring many benefits to both consumers and manufacturers (Figure 1:).




Figure 1: xr Types of main BTS interfaces

    The authors of the article also obtained some results in the works, the solution of problems of
intrusion detection in computer networks based on the assessment of changes in the network traffic
state using statistical criteria, nonparametric statistics methods [4], multi-agent model of UMV
information interaction [5], mechanisms adaptation of artificial immune systems to control the
parameters of UMV resources state[6], anomalies detection using Markov sequences [7], and also on
the concept development of an intelligent monitoring system for solving large-scale tasks in cloud
computing environments [8]
    The problem of information security is multifaceted, costly and knowledge-intensive. The search
for effective solutions to ensure information security leads to the need to create new structural elements
in systems and networks. Their main purpose is to determine the presence of an attack. Timely detection
of an attack leads to a reduction in the latent period of its action, minimizes the amount of damage
caused, as well as the costs associated with subsequent reengineering.

                                                     44
2. Problem statement

    In this article, we propose a method for detecting vulnerabilities in UMV interfaces based on
continuous values discretization. Discretization is a technology for separating continuous attribute
values in a finite set of adjacent intervals in order to obtain a set of attribute values belonging to a single
class [9,10]. One of the main features of the proposed approach is to build a system of rules for
determining the ownership of a feature vector describing UMV resources state, containing the minimum
number of conditions to be checked.
    Given a training data set Х, containing m objects xj (j=1, m) each of which belongs to a single class
Ck (k=1, s). Each xj is an l-dimensional feature vector describing the state of system resources in the
UMV-dispatch center channel at time t. System resources include: data link, memory, and processor.
We will assume that the UMV can be in one of three states: normal, precritical, critical. The permissible
limits of state changes ranges are set on the interval [0;1], Ck є [0;1]. We will use metrics as resource
characteristics:
        D – loading a resource (M-memory, Ch- channel capacity, Pr-processor),
        V – the rate of decrease in the volume of the resource.
    Thus, the vector xj is represented by a attributes tuple
                                      xj = (DM, DCh, DPr, VM,VCh,VPr ,t | Ck )
    We need to find a discretization scheme that will establish a relationship between the impact of
attacks and the system resources states that are subject to change under external influence.

3. Method description

  We will assume that values range of each attribute xr is represented on n discrete intervals, each of
which is represented by a values pair:
                                     {[z0, z1], (z1, z2], . . . , (zn-1, zn]},

where z0 – minimum value, zn – maximum attribute value xr for any r (0=< r< n), zr<zr+1. Set of values
                                                  {z1, z2, . . . , zn-1}
is the split points for the attribute xr. The main idea of the algorithm is as follows. Let the split point in
the first iteration be the average value between two adjacent attribute values xr. If the values zr fall into
the intervals (zr-1, zr] and (zr, zr+1] and belong to the same class– remove zr from the list of xr -attribute
split points until we find a pair of values that fall into two adjacent intervals but do not belong to the
same class. The process of selecting the split points is proposed to be optimized using the criterion DCR
(Discretization using Class Information to Reduce Number of Intervals) [9 - 12].
    Each attribute value can only be classified in one of the n intervals. For each attribute, a discretization
scheme is constructed (Figure 2:) in the form of a two – dimensional array, the columns of which are
associated with the intervals of loading values and the rate of reduction of resources, and the rows-
object classes.

     Class                                           Intervals                                      Total in
                                                                                                    class
                     [z0, z1)         ……             [zr-1, zr)      ……            [zn-1, zn]
       C1              q11            ……               q 1r          ……              q 1n            q1+
      ……              ……              …….              ……            …….             ……             ……
       Ci              q i1           ……                q ir         ……               q in           qi+
     …….              ……              …….              ……            …….             ……             ……
       Cs              q s1           ……                q sr         ……               q sn           qs+
 Total number          q+1            …….               q+r          …….              q+n             Q
  in intervals
Figure 2: xr attribute discretization scheme



                                                        45
     In this Figure 1:, the following designations are accepted:
     qir – total number of values belonging to the class Сi, which are in the interval (zr-1, zr];
     Qi+ – total number of objects belonging to the class Сi;
     Q+r – total number of attribute values xr, which are in the interval (zr-1, zr].
     The criterion used allows us to find a discretization scheme in which each interval belongs to objects
of only one class. We will also assume that the resources Dj, Vj, can be in one of the following states:
          normal,
          pre-critical,
          critical,
where Dj, Vj є [0;1]. The number and values of the resource state boundaries are set by the expert
depending on the nature of the task to be solved: the UMV purpose, the movement dynamics,
environmental conditions, etc.
     The algorithm for constructing the sampling scheme contains the following sequence of actions.
     1. Perform steps 1 – 6, for j=1..m, where m – number of attributes.
     2. Arrange attribute values xj in ascending order. Minimum value – z0, maximum value – zn.
     3. Create a set of all possible split points Z, for xr attribute.
     4. Construct a discretizatioin scheme for the xr attribute using the obtained partition points Z.
     5. Calculate the value of the DCR criterion for all possible obtained partition points Z:
                                                 ∑      ∑ q ⁄q
                                        DCR
                                                           n
     6. Select the division boundary that gives the highest value of the criterion DCR.
     7. Repeat steps 2 – 6 for intervals that contain objects belonging to different classes.
     8. Build a rules system for classifying input vectors that describe the resources state.
     9. End.
     The proposed method allows us to build a rules system for the membership of the analyzed vectors
to the specified classes and minimize the conditions number in the generated rules. This circumstance
plays an important role in the analysis of information received from the UMV in real time. In addition,
it is proposed to use several rules sets built separately for the parameters combinations specified by the
expert for the specified sets of resource states {DM , DCh , DPr, VM ,VCh ,VPr } , for example, the rules for
parameter D, the rules for parameter V, and others. Then the decision about UMV state– SUMV can, for
example, be accepted in accordance with the following rules: SUMV є {critical}, if they are in a critical
condition:
          one of the resources {M, Ch, Pr} by one criteria {Dj, Vj},
          one of the resources based on two criteria {Dj, Vj} etc.

4. Conclusions

    The ongoing research in the field of intelligent transport systems is based on a theoretical and
methodological basis in the areas of self-organization of complex natural and artificial immune systems,
which provide a balanced strategy for finding a solution and combine local and global search for a
solution. It has been established that the solution to the problem of detecting BTS vulnerabilities is
characterized by multidimensionality, multi-criteria, the influence of information presentation forms on
the classification accuracy, the need to use minimal a priori information, a combination of determinism
and fuzziness, the possibility of combining formal methods and taking into account expert judgments.
    Currently, most of the problems of data analysis are associated with studies of stochastic dynamical
systems, in which the detection of significant, but rare information situations is often of decisive
importance. The necessity of building information technology is revealed, since an analytical solution
is impossible under the given conditions.
The development of an intelligent technology for detecting vulnerabilities in BTS interfaces, based on
the use of new approaches and methods, will lead to an increase in the validity, reliability and efficiency
of decision support processes for assessing the probability of accepting hypotheses about the presence
of anomalous values, taking into account errors of the first and second kind.
    Adaptive decision-making methods under conditions of uncertainty will eliminate the shortcomings
and limitations inherent in classical approaches in the case of noisy data and incomplete information.

                                                      46
On the basis of Big Data technology and a special modeling stand being developed, the quality of
evaluating decisions is improved.
   The proposed approach is focused on real-time use, as it has a relatively low computational
complexity. Using the simulation mode of decision-making processes allows the expert to: first,
implement the training mode, and secondly, gives the system as a whole adaptive properties. A
promising direction is to study the sensitivity and stability of the UMV state to the impact of attacks on
a resources variety, to determine the probabilities estimates of accepting hypotheses P(H0|H0),
P(H0|H1), P(H1|H0), P(H1|H1) when recognizing the UMV resources states.

5. Acknowledgements

   The research was carried out with the financial support of the RFBR in the framework of scientific
projects № 19-29-06015 and № 19-29-06023.

6. References

[1] P. D. Zegzhda, M. A. Poltavtseva, D. S. Lavrov, Systematization of cyberphysical systems and
     assessment of their security? Problems of information security Computer system (2017) 127–138.
[2] H. Hasrouny, A. Samhat, C. Bassil, A Laouiti, Vanet security challenges and solutions: A survey
     Vehicular Communications 7 (2017) 7–20.
[3] F. Sakiz, S. Sen, A survey of attacks and detection mechanisms on intelligent transportation
     systems: Vanets and IoV Ad Hoc Networks 61 (2017) 33-50.
[4] B. Mokhtar, M. Azab, Survey on security issues in vehicular adhoc networks Alexandria
     Engineering Journal 4 (2015) 1115–1126.
[5] A. V. Skatkov, A. A. Bryukhovetskiy, D. V. Moiseev, Kullback measure in dynamic clustering
     problems of environment state observations, Environmental monitoring systems 3 (2019) 35–38.
[6] S. Kulbak, Information theory and statistics, Moscow, Nauka, 1967.
[7] A. V. Skatkov, D. V. Moiseev, A. A. Bryukhovetskiy, Model for vulnerabilities detection in
     unmanned vehicle interfaces based on artificial immune systems, IOP Publishing Ltd Journal of
     Physics: Conference Series 1515 (2020) 022043.
[8] A. V. Skatkov, A. A. Bryukhovetskiy, D. V. Moiseev, V. I. Shevchenko, Detecting vulnerabilities
     of information resources of unmanned vehicles method based on dynamic evaluation of Markov
     sequences properties, IOP Publishing Ltd Journal of Physics: Conference Series 1515 (2020)
     022033.
[9] F. Sakiz, S. Sen, A survey of attacks and detection mechanisms on intelligent transportation
     systems: Vanets and IoV Ad Hoc Networks 61 (2017) 33–50.
[10] D. A. Zighed, A method for discretization of continuous attributes for supervised learning,
     International Journal of Uncertainty, Fuzziness and Knowledge – Based Systems 17 (2001) 307–
     326.
[11] A. H. Mohammed, A. L. Junaid, A. A. Syed, Classification of Security Attacks in VANET: A
     Review of Requirements and Perspectives, MATEC Web of Conferences, 150 06038 (2018)
     doi:/10.1051/matecconf/201815006038.
[12] A. V. Skatkov, A. A. Bryukhovetskiy, V. I, Shevchenko Monitoring of qualitative changes of
     network traffic states based on the heteroscedasticity effect, Application of Information and
     Communication Technologies (2016) 7991765.




                                                     47

</pre>