Abstract: Executive practices on desertification control should be based on recognizing the current desertification state and its severity. So, it is essential to assess the ways to give zoning based on logic, active principles, and theoretical foundation for the management of desert regions. For this aim, 30 useful indices on desertification were determined in two human and natural sections. The significance of indices relative to each other and each index's importance per work unit was determined using the Delphi method. The Bonissone method in the framework of the Fuzzy Multiple Attribute Decision Making (FMADM) method was used to combine indices and determine desertification intensity in each working unit. Then, data were converted to the Fuzzy layer using Chen and Wang method, and Fuzzy analysis was performed on data. Finally, Fuzzy data were changed to non-Fuzzy, and desertification intensity was estimated. The results showed that 9.35% of the study area was in a very high class regarding desertification intensity and 9.36% of the region was in relatively high class. Desertification with moderate intensity (50.64%) and a relatively moderate intensity (29.45%) had the most shares in the study area, respectively. The quantitative value of desertification potential in the whole area from all of the components was obtained as 0.083, relatively high. This study shows the efficiency and ease of Fuzzy logic application for assessing desertification intensity.

Keywords: fuzzy logic; Bonissone method; artificial intelligence; vulnerability; zoning

Abstract: Paper deals with the design of the model of hybrid threats and cyber deception platform and solution for cyber threat detection. National networks face a broad range of cyber threats. It includes advanced and persistent peril that can evade commercially available detection tools and defeat generic security measures. Cyber attacks are becoming more intense and complex as they reflect an increasing level of sophistication, e. g. by advanced persistent threat (APT) activity. This environment of menace is of a global nature when transcending geographic boundaries and characterized by the emerging development of offensive cyber capabilities that are an inherent part of conflicts. Deception methods and techniques are being successfully employed by attackers to breach networks and remain undetected in the physical and in the virtual worlds. However, in the world of cyber security, deception as a tactic and element of a more robust defensive strategy has been still largely underexploited. The broad concepts of deception within cyber security were introduced decades ago. Still, these were technological solutions focused on providing technical capabilities to distract, mislead or misdirect the attacker. Only recently has the focus shifted on to how to shape the attackers’ sense-making of what is happening as they illegitimately explore networks. In this way, Cyber Deception nowadays provides an opportunity to scare, deter, and retaliate against those that violate organizations’ systems. In connection with the foregoing authors created and presented the novel model of hybrid threats in hybrid warfare as a combination of multiple conventional and unconventional tools of warfare. Authors investigate the cyber deception platform and industrial model and solution for threat detection using deceptionbased methods.

Keywords: cyber attack; cyber deception; cyber threats; hybrid threats; hybrid warfare

Abstract: Data streams represent a continuous stream of data, in many forms, coming from different sources. Streams are often dynamic and its underlying structure usually changes over time. When solving predictive tasks on the streaming data, traditional models, trained on historical data, may become invalid, when such change occurs. Therefore, adaptive models, equipped with mechanisms to reflect the changes in the data, are suitable to solve these tasks. Adaptive ensemble models represent a popular group of such methods used in classification tasks on data streams. In this paper, we designed and implemented the modifications of the adaptive bagging methods, which utilizes internal class-weighting schemes for the model adaptation. Implemented models were evaluated on two simulated real-world data streams and compared with base classifiers and other adaptive methods. In addition to the performance evaluation, we also analyzed other models' characteristics, such as the duration of model update and memory requirements.

Keywords: concept drift; classification; data streams; ensemble learning

Abstract: High-quality entrepreneurial education has a positive influence on SMEs’ development, the establishment of start-ups, business innovation, and many macroeconomic indicators. The aim of this study is to examine and evaluate the quality of the educational system from the standpoint of SME owners and managers. The analysis is aimed at detecting connections between respective respondent categories based on their region, type of business, industry, length of business operation, attained employees´ education, and gender. The comparative analysis aims at examining the differences between Czech and Slovak SME entrepreneurs’ perceptions of the educational system quality. The research results prove that selected differential viewpoints enable us to see the differences in how the quality of high school and university education is perceived, partially due to managers’ diverse demands as to the workforce in their respective enterprises, its integration into the work process, and the job position. This micro-view is complemented by a macro-view - demands and requirements regarding the workforce are differentiated also by the size of the enterprise, its type, length of operation, etc. in both countries. The results provide valuable information for the authors of economic and educational policies, as well as for regional strategic development planning. They also emphasize the importance of a systematic approach in solving the issue of entrepreneurial education and the need to ensure support for enterprises on all levels of education, including lifelong learning.

Keywords: SME; entrepreneurial education; key competences; quality of education; college education; educational policies; competitiveness of educational systems

Abstract: Secure remote user authentication is an authentication technique in which the remote server authorizes the identity of the user through an insecure communication network. Since then diverse remote user authentication schemes have been proposed, but each category has its advantages and disadvantages. Besides its strength and weakness, remote user authentication systems have a great impact on real-time applications such as E-health applications, telemedicine applications, Internet of Things (IoT), Cloud, and Multi-server applications. The implementation of the Tele Medicine Information System (TMIS) over public networks continues to disclose confidential information to unauthorized entities. Similarly, remote user authentication techniques have become essential in accelerating IoT as well. Security is a major concern in IoT because it allows secure access to remote services. Cloud computing services and a Multi-server environment share data among different end-users through the internet which also needs security as its paramount concern. Although intensive efforts were made in designing remote user authentication scheme for health care, IoT, Multi-server and cloud applications, the majority of these applications suffers either from security attacks or lagging of critical features. This paper presents an analytical and comprehensive survey of various remote user authentication techniques and categorizes them based on different applications. Furthermore, the state of art recent remote user authentication techniques have been compared, their advantages, key features, computational cost, storage cost, and communication cost are highlighted.

Keywords: remote user authentication; e-health; telemedicine; internet of things; multi-server; security

Abstract: The geometric product specification (GPS), has an increasing importance in machine design, manufacturing and measuring. However, standards describe the interpretation of the different kinds of form and position tolerances; there are several methods in measuring, how these deviations can be evaluated. In this article the minimum zone method (MZ) is presented in case of flatness error by coordinate measuring device. Different search algorithms can be applied during the implementation, in order to solve the geometric problem of flatness evaluation. In this article, the genetic algorithm is investigated and compared with hill climbing algorithms. The optimization of the parameters of the genetic algorithm is also presented.

Keywords: geometric tolerances; coordinate measuring method; flatness; minimum zone method; genetic algorithm

Abstract: There is a premise that the activity diagrams can communicate their knowledge to the source code. This article analyzes the opportunity of the activity diagrams to improve the comprehensibility, orientation, reading, and modularization of the source code. It proposes an Activity Diagram Driven Approach (ADDA) and verifies application suitability of the approach in comparison to Use Case and Package-based approach. It highlights the strengths and weaknesses of such behavior description and discusses the identified limits and benefits of the proposed approach. It proposes an extension of the source code modularization at metamodel level based on source code parts associated with certain elements of the activity diagram. The proposed solution is evaluated over several test cases from different aspects using implemented plug-in and the results show appropriate use of the proposed approach.

Keywords: UML; activity diagram; source code; orientation; comprehensibility; modularization

Abstract: Using brain waves, for user authentication, is an important emerging technology. The University of Defence, in the Belgrade and the Serbian Ministry of Defence (MoD), have recognized the importance of biometric applications in identity management. As a result, an intensive research, under the project named “Access control management of protected resources in Ministry of Defence and Serbian Armed Forces computer networks based on multimodal user identification”, has been conducted over the last few years. The main contribution of this paper, is a software platform, for learning about brain wave acquisition and analysis. This platform was developed as a part of the project, with the main goal to improve and increase participant’s knowledge in biometrics. A research study was conducted, with the aim of comparing traditional learning methods and learning methods based on the developed platform.

Keywords: biometrics; brain waves; learning tool; software platform; user authentication

Abstract: Pneumatic pressure vessels, or more commonly used names, air tanks, are used for storing compressed air to operate different pneumatic tools, for example, wrenches, grinders, sanders and others. The objective of design optimization of air tanks is cost reduction, by reducing the total mass of the vessel, the quantity of the welding seams with adequate strength and stiffness. This study describes the steps of this design optimization, but also deals with high cycle fatigue, to determine the lifetime of the equipment and the effect of the fire load. The optimization was made with the Generalized Reduced Gradient (GRG) method. Nozzles with the same geometry will have a different impact on the utilization, in the various diameters of shells. Therefore, the allowed number of cycles will change. The fire load was modelled as an ISO fire and corresponding temperature/pressure increase rates were investigated.

Keywords: pressure vessel; fire load; lumped model; optimization

Abstract: In this study, a terminal sliding mode controller is introduced, for trajectory tracking control, of a flexible link robot manipulator. Two important parameters are considered; angle of the link and tip deflection. To demonstrate the effectiveness of the developed controller, control gain parameters in the sliding mode controller are determined by using Fuzzy Logic Control law. Another important feature of the developed controller is that it is robust against external disturbances. Stability analysis of the system is guaranteed by the Lyapunov theory. Trajectory tracking errors, angles and tip deflection of the links are investigated for two different trajectories. When the results are examined, it is seen that the developed controller, with fuzzy logic, is effective, even if there are external disturbances and parametric uncertainties in the system.

Keywords: flexible link manipulator; fuzzy controller; terminal sliding mode controller

Abstract: The landing task is fundamental to Micro air vehicles (MAVs) when attempting to land in an unpredictable environment (e.g., presence of static obstacles or moving obstacles). The MAV should immediately detect the environment through its sensors and decide its actions for landing. This paper addresses the problem of the autonomous landing approach of a commercial AR. Drone 2.0 in presence of uncertain obstacles in an indoor environment. A localization methodology to estimate the drone’s pose based on the sensor fusion techniques which fuses IMU and Poxyz signals is proposed. In addition, a vision-based approach to detect and estimate the velocity, position of the moving obstacle in the drone’s working environment is presented. To control the drone landing accurately, a cascade control based on an Accelerated Particle Swarm Optimization algorithm (APSO) is designed. The simulation and experimental results demonstrate that the obtained model is appropriate for the measured data.

Keywords: Micro air vehicles (MAVs); autonomous landing; obstacle avoidance; dynamics obstacle’ velocity estimation; sensor fusion; optimal control

Abstract: The issue of capital structure is among the most commonly discussed fields within Corporate Finance Theory. By optimizing the capital structure, it is possible to achieve an increase in the company value and increase the company flexibility and competitiveness. Economists have, for more than half a century, seen tax benefits (tax shields) as a capital structure determinant. Nonetheless, leverage is also influenced by earnings management, which can significantly reduce information asymmetry, between stakeholders when used correctly. This paper examines the influence of the tax shield and earnings management on a corporate capital structure, in V4 countries. In order to determine the influence of the tax shield and earnings management, panel data model was used. Net sample consisted of 10627 companies from the V4 countries from 2014-2017. The results of the model indicate that corporate behavior in the area of capital structure follows Pecking order; short-term trade credit is the most commonly used liability. The interest tax shield is of little importance for deciding between debt and equity, while the non-debt tax shield is negatively correlated with debt. Furthermore, an inverse correlation between debt and earnings management, as measured by the modified Jones model, indicates that debt monitoring, reduces agent costs and reduces the application of earnings management techniques.

Keywords: tax advantage; earnings management; emerging economies; capital structure determinants; Visegrad countries

Abstract: Two unimodal deep networks and one multimodal deep network are created to test for possible mechanisms of sensory integration that may shed more light on how sensory integration is carried out in biological organisms. One unimodal network is provided with pictures and the other with mel-spectrograms created from sounds. Adapted pre-trained VGG16 network was used for unimodal networks. After training consisting of 30 epochs and repeated for 100 runs the unimodal networks achieved an average accuracy of 0.57 and 0.73 respectively. The multimodal network received processed features from both unimodal networks and after training consisting of 30 epochs and repeated for 100 runs outperformed both unimodal networks with the average accuracy of 0.79. Next, noise was applied to the test data to see how unimodal and multimodal networks compare in noisy environments. Unimodal networks achieved an average accuracy of 0.63 and 0.69 respectively. Again, the multimodal network outperformed both unimodal networks with an average accuracy of 0.73. Pre-trained networks were used and limited training data were provided to the networks to simulate conditions similar to animal brains.

Keywords: sensory integration; deep learning; neural network