This paper presents a practical application of machine learning techniques for the detection of Distributed Denial of Service (DDoS) attacks in network traffic. The goal of the research is to evaluate the effectiveness of selected classification models, which are Logistic Regression, Decision Tree, Random Forest, and K-Nearest Neighbors, on the publicly available APA-DDoS dataset. The study involves a complete pipeline, including data preprocessing, model training, parameter optimization, and performance evaluation using metrics such as accuracy, precision, recall, F1-score, and cross-validation. Among the tested models, Random Forest achieved the highest classification performance and stability, demonstrating its suitability for real world deployment. The results indicate that machine learning provides a viable and adaptive approach to enhancing cybersecurity against evolving DDoS threats.

This study examines the use of large language models for grading assignments in computer science education. A dataset of authentic student submissions, including source code, written documents, and image-based content, was evaluated using a controlled prompt strategy designed to enforce uniform numeric scoring. The models were assessed based on their ability to produce grades aligned with human evaluations while adhering to strict output constraints. The analysis focuses on grading accuracy, consistency, and sensitivity to task structure and prompt formulation. The results indicate that language models can support grading in structured assignments with clearly defined expectations, while their reliability decreases in open-ended or loosely specified tasks. Prompt formulation influenced output stability, particularly for incomplete or ambiguous submissions. Overall, the findings suggest that large language models can assist instructors in scaling assessment, provided they are deployed with clear procedures and human oversight to maintain fairness and alignment with educational objectives.

This paper presents the development and implementation of an Extract-Transform-Load (ETL) model aimed at processing pandemic data. The model gathers data from various official sources, transforms it for easy analysis, and provides insights through graphical representations. Implemented using Python, the system automates data extraction and visualization, offering an accessible tool for understanding pandemic trends.

This work deals with procedural content generation, specializing in content generation in the fields of computer graphics and computer games. Even more precisely, it deals with the procedural generation of three-dimensional polygonal surface models of rivers. In the introductory part, the paper provides an overview of related procedural content generation principles using noise generating functions and more complex simulation techniques. Then, it presents a within this work designed algorithm for the procedural generation of a river model by creating its bed using the designed concept of control tiles, which allow controlling the direction and shape of the river. The generator's goal is to provide a tool for creating river models that can be integrated into the environment of 3D computer games.

This article highlights the impact of game creation on an introductory programming course (CS0, CS1) combined with problem-based learning. It introduces and defines the Game-Creative Learning (GCL) approach, which gradually transforms traditional 'learning from scratch' methods into an innovative framework where students develop their own computer games while learning a new programming language. Over 15 years, this approach has engaged over 10,000 students - novice programmers, addressing educational challenges and leading to higher course quality, increased motivation, and better student engagement. Beginning with simple tasks like playing with Karel the Robot, students' progress to creating their robots and ultimately developing 2D games. Throughout the course, scores and levels of engagement are assessed and tracked in a competitive environment. This approach aligns with Bloom's taxonomy by guiding students progressively through its cognitive levels.

This work deals with the issue of procedural content generation in the field of computer games, focusing on the generation of polygonal surface models of trees. As part of the work, an algorithm for a procedural content generator was created, using a recursive approach, to generate polygonal surface models of trees. The input of the algorithm consists of a set of parameters, which allow generating trees of various types, shapes and sizes. The algorithm was implemented in JavaScript and integrated into a web application, enabling users to generate, export, and interactively visualize tree models in real-time as they are generated, as well as apply transformations.

This paper explores filtration techniques for processing ECG signals, focusing on the evaluation of effective denoising methods. We highlight the effectiveness of Stationary Wavelet Transform as the most suitable approach for denoising ECG signals while preserving critical cardiac features. Stationary Wavelet Transform's superior performance was validated through rigorous testing, making it a very good choice for ECG signal filtration. We have also investigated other filtration techniques, including high-pass, Chebyshev type II, Kalman, notch, Savitzky-Golay Smoothing, Moving Average, Empirical Mode Decomposition, Empirical Wavelet Transform, and Stationary Wavelet Transform filtering (with tested wavelets 'db4', 'db5', 'db6', 'sym4', 'sym5', 'coif3', 'coif4', 'coif5', 'bior3.5', and 'rbio3.9'), to assess their effectiveness in ECG signal processing. The paper highlights the advantages of the Stationary Wavelet Transform filtration technique in maintaining signal integrity for subsequent analysis. The optimal configuration, determined to be the wavelet type 'rbio3.9' at level 5 with a threshold scale of 0.5, balances effective noise reduction with the preservation of crucial ECG signal details, thereby significantly contributing to clinical ECG analysis and advancing diagnostic precision.

This paper addresses the issue of how to represent the geometry of voxelized three dimensional scenes using domain-specific hierarchical data structures. It provides a comprehensive overview of these data structures, delving into the specifics of sparse voxel octrees in greater depth. Then, it discusses the proposed use of dimensionality upscaling, i.e. the transformation of a three-dimensional voxelized scene into a higher number of dimensions, to enable the representation of the geometry of this scene via an n-dimensional sparse voxel tree hierarchical data structure. This approach allows a smaller size of binary representation of the 3D scene geometry-compared to its representation using a sparse voxel octree. This is documented in the next section of the paper, where we summarize the results of tests encoding the geometries of the respective scenes (based on voxelization of polygonal surface models), into sparse voxel trees of different levels of dimensionality. We achieved the highest compactness of the binary representation of the resulting data structure, compared to the traditional sparse voxel octree (increase in compactness ranged from 2.02 to 4.44 times), using upscaling to 6D and 7D, respectively.

This paper addresses the issue of geometry representation of voxelized threedimensional scenes, using domain-specific hierarchical data structures, utilizing lossless compression. It discusses the details of the sparse voxel directed acyclic graph hierarchical data structure with implemented child node pointers, including a detailed analysis of its construction. Then, based on this analysis, we discuss the possibilities of adjusting its parameters, in order to optimize it in terms of its binary representation size. In particular, the possibility of using various numbers of child node pointer types and the binary representation lengths of these is considered; this is also related to the different binary representation lengths of the child node mask and the different binary representation lengths and meanings of the constituent header tags. The various settings of the data structure parameters are then tested on three models whose original representation-surface polygonal models-was transformed: each model was voxelized to six different resolutions. The results obtained were also compared with an uncompressed representation of the geometry of the voxelized three-dimensional scene, encoded using a regular threedimensional grid of one-bit scalar values (1b/vox). Based on these tests, using a 16b child node mask, three different child node pointer lengths (8b, 16b and 32b) and an 8b address word length came out as the best parameter combination of the investigated data structure in terms of data representation compactness. In the tests, a 1.88-3.06-fold data compression rate was achieved with this parameter configuration for the used test models and scene resolutions, compared to the traditional SVDAG structure using 32b pointers. The maximum data compression ratio (1098.3)-compared to the 3D grid of 1b/vox scalar values-was achieved with this parameter setting for a scene resolution of 40963 voxels.

The main objective of this research was to develop a classification model for the design of a cloud cognitive assistant. The classification model should be able to classify the textual description of the cloud architecture into desired multiple classes. For the purpose of implementation of such assistant, the analysis of current state of cognitive assistants and cloud computing was researched. As for the implementation of classifier, the overview of the possible ways how to create such classifier was also added to the research. Based on the analysis, a solution was proposed for the implementation of a model for text classification, and also a proposal for the implementation of an assistant that would use the proposed model. The Keras library was used to create a sequential text classifier. The IBM Watson cloud services were used to deploy the created model into live environment, and the services from the same group were used to develop a proposed cognitive assistant that was in the end connected to the classifier model.

How to distinguish truth from lies in the environment in which we are exposed to vast amounts of information from different sources and through different platforms - is a burning challenge today. Although lies and manipulations in public information space are not a novelty, the quantity and speed of spreading misinformation, especially through social networks and mobile communication applications present an unprecedented challenge. The level of pollution of the information environment and the complexity of the actors, the technique and the motivation behind its contributors have never been so present, within hybrid threats today. Novelty is, therefore, the speed of spreading misinformation through network channels, platforms and applications. In this paper, a new non-hierarchical conceptual model of hybrid threats aiming to improve awareness and systems' resilience is developed, which includes disinformation campaigns and provides an industrial example of a system for combating fake news.

The electrocardiogram (ECG) captures a person’s heart activity and is, therefore, one of the best ways to diagnose heart arrhythmia. Artificial intelligence can quickly learn and process a huge amount of data, the properties of which could simplify the process of diagnosing any disease. In this work, we created 4 convolutional neural network models for determining 6 types of heart arrhythmias classes based on two kinds of heartbeat images of ECG signals. The goal was to investigate the effects of different types of input data and model structure on the success of the prediction. We worked with a publicly available MIT-BIH database of arrhythmias. The division of ECG records into heartbeats was realized using our R peak detection method, which was able to detect up to 97.39 % of all R peaks. Each heartbeat was imaged as a representation in the time domain and then in the time-frequency domain using a short-time Fourier transform. The best model, trained over time-domain representations, consisted of 5 convolutional and 3 pooling layers and achieved a classification accuracy of 98.5 %.

Digitization and deployment of information and communication technologies is a global trend in all areas of human activities, including the area of capturing the surrounding landscape in the form of a digital twin. These processes are also reflected in the field of spatial planning, which can effectively deploy geospatial information technologies in its processes related to its entire life cycle, from the process of gathering and evaluating suggestions, spatial planning documentations and other relevant documents to the registration and publications of the full spatial planning documentation. That is why Slovakia proceeded to create the new legislation related to spatial planning, which came into force those days and whose philosophy is based on the gradual full digitization of this issue. Part of this process is also the creation of an information system that will become a platform usable by all stakeholders. The aim of this contribution is to present the initial phase of digitization of the spatial planning process, which includes the creation of a uniform data structure for spatial plan processors, which will enable the creation of an obligatory graphic part of the spatial plan in a standardized, fully digital form.

This paper explores the design and implementation of a cognitive assistant powered by cloud computing. We begin by reviewing existing cloud service providers and assistant solutions. Building on this foundation, we define a set of requirements for our cognitive assistant and utilize them to guide the development process. To validate the functionality of our creation, we compare its outputs with established reference architectures.

Accurate measurement of the flow rate of air entering the jet engine is very important from the point of view of the design of control algorithms and diagnostic systems (e.g. when classifying situational frameworks of situational control, or when predicting unmeasured parameters or the failure of other measured parameters within the diagnostic system, the design and implementation of which for micro (JetCat P-80, TJ-20) and small (iSTC-21v, TJ-100) jet engines is devoted to the scientific research team of the Laboratory of Intelligent Aircraft Engines). In order to accurately measure the air flow rate in the entire cross-section of the inlet device of the iSTC-21v small jet engine based on the value of the local velocity, temperature and pressure, it is necessary to know the distribution of the absolute velocity at the entrance to the engine (e.g. we may be interested in the homogeneity of the absolute velocity and turbulence, by sensing which using CTA equipment or pressure probes, another scientific and research team of the Aerodynamic Tunnel Laboratory). The speed distribution is also influenced by the shape of the bell mouth - the leading edge of the newly designed iSTC-21v engine inlet assembly (as opposed to the classic engine inlet assembly, where the flow rate of air is not measured). Using the numerical CFD simulation in Ansys and the program for calculating the superellipse coordinates in Matlab, in the presented article we are devoted to the pilot design of the inlet device with the sensing of the air flow amount of the small jet engine iSTC-21v.

Online voting is a topic that companies have been dealing with for many years. Several countries tried and are still trying to implement such voting system instead of a classical paper election system we know the most nowadays, as it brings many advantages. Blockchain technology became known to people mainly thanks to the Bitcoin cryptocurrency which was created in 2008. Blockchain presented a new solution that was almost ideal for voting purposes. With the use of blockchain, the security and reliability of online voting systems has also increased over time. Voting became transparent. The work focuses on the development of an online voting management system built on such blockchain technology. The introductory part is focused on the blockchain technology and online voting. The work focuses on the design and implementation of a system based on the blockchain technology which allows the administrator to create new votes, start them and stop them.

This article presents a solution to obtain adequate ECG signals using low-cost devices. We believe that a pilot ECG examination in the home environment can detect certain forms of arrhythmias. If arrhythmias are diagnosed and treated in time, a person can completely avoid serious health problems. Performing an ECG examination at home is hindered by the high cost of professional equipment. That’s why some low-cost devices capable of recording an ECG signal have drawn our attention. The results of our research showed that an affordable ECG can be recorded at home using a 1-lead ECG machine.

This research paper examines the use of technology in managing mass casualty incidents. This work assesses the current state of technical advancements, such software inventions, hardware technology, robotics, telemedicine, and artificial intelligence. It also identifies the challenges and limitations of using technology, such as technical issues and financial constraints. The paper provides insights and recommendations for effective integration of technology in mass casualty incident management to improve response times, reduce casualties, and enhance situational awareness. As a result, we have proposed a software solution for patient triage and transportation. All of which increases chances of successfully managing mass casualty incidents.

This paper focuses on training machine learning models using the XGBoost and extremely randomized trees algorithms on two datasets obtained using static and dynamic analysis of real malicious and benign samples. We then compare their success rates—both mutually and with other algorithms, such as the random forest, the decision tree, the support vector machine, and the naïve Bayes algorithms, which we compared in our previous work on the same datasets. The best performing classification models, using the XGBoost algorithm, achieved 91.9% detection accuracy and 98.2% sensitivity, 0.853 AUC, and 0.949 F1 score on the static analysis dataset, and 96.4% accuracy and 98.5% sensitivity, 0.940 AUC, and 0.977 F1 score on the dynamic analysis dataset. Then, we exported the best performing machine learning models and used them in our proposed MLMD program, automating the process of static and dynamic analysis and allowing the trained models to be used for classification on new samples.