The application of clustering algorithms for resolving point clouds to reduce data size while retaining key geometric information is discussed. The algorithms such as DBSCAN, OPTICS, and k-means are analyzed. The focus is on evaluating the performance of algorithms for compressing point clouds containing millions of elements in terms of retaining object structure and optimizing computational resources. The results show that, despite the high accuracy of some algorithms such as OPTICS and DBSCAN, their performance is limited in terms of processing time and memory consumption, making them less suitable for large datasets. On the contrary, the G-Means algorithm has demonstrated high efficiency in resolving point clouds while retaining meaningful geometric features and providing acceptable processing speed. The proposed approach can significantly reduce the amount of data and maintain sufficient accuracy for subsequent modeling and analysis.The study was conducted using two examples: a power line support and an abandoned building. The G-Means algorithm reduced the point cloud file size by reducing the number of elements for both objects. The power line support point cloud originally had a more uniform structure, which allowed the rarefaction algorithm to remove more points with Object Geometry Retentions.The relevance of the problem of processing three dimensional unstructured data to create digital twins is due to the increased demand for efficient methods of designing and modeling of objects in various industries, such as construction, engineering, medical diagnostics, etc.
Automatic communication components visual recognition plays an important role in a number of modern applications, such as human-computer interaction, accessibility for individuals with hearing impairments, and advanced communication analysis systems. This technology enables detailed evaluation of verbal, non-verbal and paraverbal communication components. In particular, determining the number of phonemes based on articulation can provide systems with subtle paraverbal elements, thus improving both human and machine interpretation of communication. Previous developments in visual determining of minimal linguistic units in speech encountered significant challenges, particularly due to insufficient and imbalanced training datasets. For example, the disparity in data representation among classes was substantial, with the number of examples per class differing by factors of ten or more. Such an imbalance can cause bias in machine learning models, leading to skewed predictions and reduced overall accuracy, particularly for underrepresented classes. As a result, data balancing is essential for achieving more accurate and reliable results. In this context, the present study focuses on evaluating various data balancing techniques to address the task of visual speech recognition and counting the number of phonemes in English words. The considered model is based on ResNet-18 with 3D convolutions. The study examines the effects of both undersampling and oversampling approaches. Undersampling is performed by cutting both video fragments randomly in each class and video fragments of specific words entirely. Oversampling is carried out using classical video augmentation methods, specifically via VidAug library, and by cross-dataset augmentation. Experimental results show that these data balancing methods positively influence the training process, enhancing the model's ability to generalize across classes and improving its accuracy to a certain extent. However, they are not fully decisive in determining the model's overall effectiveness. The study highlights the need for further exploration of advanced balancing techniques and alternative model architectures to address the challenges posed by class imbalance comprehensively.
There are many tasks from various fields of science and technology that must process huge amounts of data in real time in current time. Reconfigurable computing systems based on programmable logic integrated circuits (FPGA) can become the technical basis for such computers. FPGAs are characterized by an increase in real performance due to an annual increase in the number of logic gates in new architectural solutions. At the same time, the increase in the number of contacts on the FPGA occurs disproportionately slowly compared to the growth of the volume of computing equipment on the chip. This leads to the fact that with the growth of the potential performance of the device, its switching and transmission capabilities remain at the same level. This paper examines an approach to increasing the throughput of data transmission channels in FPGAs by using lossless data compression algorithms multiple of the average compression ratio of the selected algorithm. It is proposed that an increase in the channel throughput leads to an increase in the specific performance of the reconfigurable computing system. A modernized Huffman algorithm is presented, which allows compressing dense data streams with an optimal compression ratio at the rate of arrival at the RCS. The use of new principles for constructing a Huffman tree and forming the lengths of new codes made it possible to reduce latency, make it equal to the duty cycle of data receipt and fit into the task cyclogram. A simulation was conducted, which showed that the developed algorithm makes it possible to compress a data stream at a speed of 128 Gbit/s on one computing core (pipeline). Comparison with other modifications of the Huffman algorithms on RCS showed that, with the same flow processing speed, the developed algorithm occupies the smallest hardware resource. This ensures a 25 % increase in specific performance. Also, the performance of this implementation exceeds all known third-party implementations on hybrid computing systems by up to 5 times. The use of such compression systems will increase the speed of information processing when solving such promising problems on RCS as training convolutional neural networks or pairing high-speed computers with external memory units.
Modern IT infrastructures are facing an increasing number of incidents, that require a fast and accurate response. In this regard, there is a growing interest in using large language models (LLM) to predict solutions automatically to eliminate IT incidents. A comparative analysis of four LLMs – ChatGPT, DeepSeek, Claude Sonnet, LLaMA – in the context of generating solutions based on textual descriptions of incidentsv is presented. A multi-criteria analysis with the participation of an expert group was used to assess the quality of the proposed models. The solutions were evaluated according to four criteria: accuracy, detail, practical applicability and conciseness. Based on 50 real-life incidents, responses were received from each model and then evaluated on a 5-point scale. The results showed that DeepSeek demonstrated the highest scores in accuracy and detail, making it the most suitable for complex technical incidents. Claude Sonnet and LLaMA distinguished themselves by their conciseness, providing quick perception of solutions in operational situations. ChatGPT showed good results in terms of applicability, but was inferior in terms of detail. A comparative analysis confirmed, that the choice of the optimal LLM depends on the specifics of the task: DeepSeek is preferable for deep diagnostics, whereas Claude Sonnet and LLaMA are suitable for rapid response. The data obtained can be used to improve the efficiency of IT incident management systems by choosing a more accurate model depending on the operating conditions.
Quantification of the interaction of technical facilities with natural environments requires the development of algorithms and methods of digital support. The development of CAD in the direction of forecasting the aerodynamic flow of architectural and construction objects makes it possible to simplify the task of determining wind loads. Even with the use of modern software systems, the initial data for determining wind loads is prepared individually for each facility. When analyzing high-rise through structures consisting of a large number of elements, the preparation of initial data for determining the static components of wind loads is a laborious and very demanding task. In the case of even small errors in determining wind loads, the results of the design calculation may have a significant error. Methods that simplify the collection of wind load on the lattice structure lead to a decrease in the accuracy of the calculation results. An option for automating the collection of the static component of the wind load on the tower is given in the article. Automating the collection of static wind load on a lattice tower reduces the likelihood of errors, labor costs, and calculation time for the bearing capacity of cellular lattice poles and power lines. The developed algorithm will be used to automate the collection of wind static load on equipment, elements of stairs, platforms and cable routes of cellular communication poles and power lines. The automated wind load collection algorithm will be useful in developing software for calculating lattice structures.
The article presents a comparative analysis of algorithms for computing the unified Pareto front obtained by merging local fronts generated by genetic algorithms in multi-objective optimization problems. Various approaches are examined, including pairwise comparison methods, divide-and-conquer strategies, the sweep-line algorithm, and KD-tree–based search techniques. Special attention is paid to the theoretical analysis of each method’s operating principles and their asymptotic time complexities in worst-case and average-case scenarios. Subsequently, synthetic tests were performed on data sets varying in distribution geometry and the density of non-dominated solutions. In these experiments, the genotype dimensionality was fixed: each solution was represented by a three-parameter vector. To model distributional diversity, four a priori characteristics were employed: geometric balance, coefficient of variation, skewness, and normalized entropy. A total of 800 independent experiments were conducted with different combinations of population sizes, proportions of solutions in the resulting front, and ten levels for each a priori metric. Specifically, population sizes of 200, 400, 800, and 1600 were tested, with front proportions ranging from 10 % to 90 %. Each experiment was repeated several dozen times, and results were averaged to ensure statistical reliability. Based on the theoretical and empirical findings, recommendations are formulated for selecting the most appropriate algorithmic approaches that optimize the trade-off between computational performance and cost, depending on input-data characteristics and solution-quality requirements.
Digital support technique has been developed to analyze the interaction of surfaces of double curvature with aerodynamic flows, which is available for use in engineering practice. At the same time, digital technologies were used, which play a key role in modern aerodynamic research due to their modeling and analysis capabilities of complex physical processes. Computer modeling is actively used to develop various algorithms and techniques in scientific work, engineering, and other areas of life. To conduct
A detailed comparative analysis of two intelligent control methods, namely proportional-integral-differential (PID) control and fuzzy logic control (FLC), used in control systems for brushless motors (BLDC). Special attention is paid to the effectiveness of each approach in real-time conditions and in various industrial applications, where qualitatively, PID suits deterministic tasks (CNC machining), while FLC excels in dynamic environments (robotics). Hybrid fuzzy-PID approaches balance adaptability and precision. Implementation challenges include PID's recalibration needs and FLC's rule-design complexity. PID's computational efficiency contrasts with FLC's memory demands. The theoretical foundations of the PID controller and the FLC approach are considered. The properties of these methods are analyzed in the context of their application in electric drive control systems, where PID control, based on linear error minimization, offers simplicity and precision in stable systems but struggles with nonlinearities and disturbances. FLC, using heuristic rule-based inference, adapts to nonlinear dynamics without precise models but requires computational resources. Some indicators of the effectiveness of the management methods under consideration are investigated. Quantitative and qualitative comparisons are presented to objectively assess the feasibility of using each approach in different operational scenarios, and the advantages and limitations of both methods are described, where PID remains ideal for linear systems, while FLC addresses uncertainty. Selection hinges on prioritizing precision PID versus adaptability FLC. Future work explores machine learning-enhanced hybrids for industrial scalability.
The research work is devoted to the theory of fuzzy sets and fuzzy logic, its realization in Fuzzy Logic Toolbox of MATLAB environment. The theoretical part of the study introduces the theory of fuzzy sets and fuzzy logic, oriented to the construction of mathematical models, taking into account the incompleteness and imprecision of data. The main purpose of the study is to identify the shortcomings of the theory implementation by means of MATLAB computer environment, limiting the correctness of fuzzy inference, the possibility of visualization of systems with three or more variables, as well as the visibility of in the construction of fuzzy systems. As ways of solution, methods based on mathematical modeling, revision of 3D model construction and creation of additional possibilities of visualization of interrelationships of different parameters were proposed. In addition, the paper presents methodological solutions that extend the capabilities of Fuzzy Logic Toolbox, increasing its correctness, accuracy and applicability in a wide range of engineering and scientific disciplines. The main solutions are related to the elimination of the following limitations: unstable operation of specific editors, impossibility of 3D visualization of three or more input variables, distorted visual representation of a certain number of variables. Other disadvantages of the Fuzzy Logic Toolbox package are also considered, which refer to additional costs in the process of using the software and possible risks associated with installing an unlicensed version of MATLAB. Additionally, a narrow application of fuzzy set theory is noted, which does not take into account additional criteria, weighting coefficients, specific indices that could be used for a comprehensive evaluation of the tudied systems of different spectrum. It is concluded that further research will be aimed at eliminating the identified limitations and realizing the proposed functions as an independent application.
The article presents the results of introducing objects – device processors designated by words – into the categorical and network сNnet-scheme of the Markov algorithm and the synthesis of their morphisms. The applied (technical, technological) goal of defining the processors of devices in the composition of this algorithm is the synthesis of such a scheme that would be most suitable for use in Markov networks with abnormal inference, implemented as separate hardware devices (calculators). These networks are distinguished by the presence of an output that cannot be considered normal according to Markov, such an output is called “abnormal”. The implementation of such an output in hardware devices, in particular, processors, is of scientific and applied interest, since it allows implementing a new method of parallel and distributed computing by managing a list of data sources, in fact, a new technology. The “classical”, intuitionistic creativity according to Brouwer in Markov networks of indirect propagation is considered, some principles of its implementation are defined. To study the points of application of creativity to the Markov algorithm, in addition to non-material (mathematical, artificial, virtual) objects, objects of the world around us are taken into account, in particular, processors and some other devices that ensure the execution of the algorithm's instructions. The scientific goal of joint consideration of such objects is to find a way to synthesize Markov networks with abnormal inference capable of processing “all” morphisms, in other words, to synthesize a complete (in the mathematical sense) list of morphisms. The conducted research of creativity in such networks is also aimed at finding new, previously unaccounted for morphisms, generalization and classification of which is a constructive clarification of the methods of interaction of objects. The ultimate goal of synthesizing a complete list of morphisms is to use Markov networks with abnormal inference together with artificial intelligence. Therefore, the study regularly introduces various aspects of their interaction, otherwise – the morphism from the network to artificial intelligence and the inverse morphism are specified. The first part of the study will show the results of determining the following processors in the сNnet-scheme of the Markov algorithm: control channel, input-output, small and large correlators. Conclusions are made that the obtained сNnet-scheme of the Markov algorithm with small correlators can be used in telecommunications, in particular, for processing data from digital communication channels in conditions of interference.
The article presents results of SO-CAL lexicon-based method adaptation for sentiment analysis in Russian texts by modification lexicon using aggregated attention weights of ruRoBERTa-large neural language model. The proposed method includes four steps. At the first step of the algorithm, ruRoBERTa-large neural language model was fine-tuned on the first text corpus. The second step involves obtaining sentiment and attention weights of fine-tuned model for each text on the second text corpus. Aggregation of attention weights was carried out according to a special algorithm. At the third step, new lexicon are formed based on existing lexicon and aggregated attention weights of neural language model for SO-CAL lexicon-based method. The first three approaches involved incorporating words with high aggregated attention weights, differing in their handling of ambiguous sentiment cases. The fourth method removed terms with low aggregated attention weight. The fifth implemented a hybrid approach: replacement of low weight lexical items with their high attention counterparts. At the last step, the optimal lexicon was selected from the list of all received lexicons. Then optimal lexicon was evaluated and classification results for initial and optimal lexicons were compared. We used macro F1-score to compare lexicons. The results showed that combined method of lexicon formation was superior to the others: for validation data of SentiRuEval-2016 corpus macro F1-score was increased by 4,28 %, for test data of SentiRuEval-2015 corpus this metric was increased by 15,86 %.
Icing is the most dangerous among environmental impacts on aircraft. The freezing of ice on the surface of the wing leads to a change in its bearing properties, due to a change in the geometric shape of the wing surface, which leads to a change in the structure of the aerodynamic flow. Aircraft icing most often occurs when flying in clouds mainly due to the presence of water in a droplet-like state. The results of geometric and digital modeling of ice build-up processes on the wing surface in a wide temperature range 0…–14 °C are presented. The modeling was performed in the ANSYS software package based on the results of digital modeling. Digital modeling of ice build-up on the wing surface was performed using the FENSAP-ICE program. A comparison of the calculated and experimental data showed a satisfactory agreement of the results. The geometric characteristics of the shape of ice growths at different temperatures for the specified values of water content and angle of attack were determined. A mathematical model to describe the geometric characteristics of the wing surface, taking into account icing as a function of temperature was proposed. Based on the results of digital modeling, a mathematical model that expresses the dependence between the geometric characteristic of an ice outgrowth – its length along the chord of the wing – and temperature has been developed. The characteristic type of dependence of the wing mass increment on temperature was determined. The nature of the influence of ice on the aerodynamic characteristics – drag and aerodynamic quality of the wing – was shown. The results of the work are recommended for use in digital modeling of aerodynamics and dynamics of aircraft movement in icing conditions.
A classification of creative sequences by L. E. Ya. Brauer is proposed by: control, purpose, classification, source data. The paper shows a method for formalizing Brouwer’s creative sequences in the theory of Markov algorithms using the cNnet-scheme of the Markov algorithm, developed to replace the well-known γ-scheme of the normal algorithm, and differing from it in that it cannot be considered normal. The article considers the “Trulstra result” about the creative sequence calculator mentioned in the work of N. N. Nepeyvoda, from which the scientific problem of synthesizing the mathematical formula of such a calculator is extracted and a method for implementing its algorithm is shown. The results of achieving the research objective – classification and selection of creative sequences – are presented. A part of the algorithm (logical) executed by the controller is synthesized, thereby achieving the practical goal of the study. The “main” creative sequence is identified and disclosed from the theoretical-algorithmic positions. A hypothesis is put forward that the “main” creative sequences arise in generative algorithms. The selection of such algorithms allowed us to show an additional hypothesis about the division of all algorithms into: generative, processing and modernizing. The Brauer creativity associator is introduced, which is included in the cNnet-scheme of the Markov algorithm and, as a consequence, in network with abnormal inference. It is noted that the method of calculating creativity in the cNnet-scheme of the Markov algorithm is limited by the assumption about its result. A classification of some basic types of lawless sequences that can be defined in the theory of Markov algorithms is given. A Markov network of indirect propagation with a demonstration of the “Troelstra result” is synthesized. It is noted that the assumption of this mathematician “about three inputs” cannot be understood literally, since it is more accurate to talk about three types of inputs, where there can be several inputs of each type. Conclusions were made that allowed us to put forward a hypothesis that category theory and n-categories are quite suitable for generating artificial intelligence, at the current stage of its comprehension. The functioning algorithm can be represented by a composition of algorithms already known in science. The modernizing algorithm should be considered in further research.
Introduction. Today, mobile phones and applications working with them penetrate into all spheres of life of modern digital society, including the field of education, in which a new direction is being developed – mobile training (training using portable mobile devices – smartphones, tablets, smart watches, etc.). Considering the fact that each mobile device has its own technical characteristics and each student has its own individual psychophysiological features, there is a problem of developing an application interface that can adapt to them to increase the usability of the software product and ensure comfortable conditions for studying the material of the discipline. Method. The paper presents an author's method of forming adaptive interfaces of mobile applications by integrating two artificial neural networks (ANN). The first convolutional ANN – determines the age, gender and emotional state of the user. The results of its work, along with the values of the formation of other features of the trainees (professional, physical, mental), the characteristics of the smartphone, are used in the work of a deep ANN, presented in the form of a multilayer perceptron. Based on the work of the deep ANN to determine the values of design ergonomic indicators, an interface is generated in the application, adapted for each student. Main results. For the first time, a method has been introduced that integrates the operation of two ANNs: convolutional and deep. The convolutional ANN is implemented through the DeepFace Python library and is used to assess gender and age characteristics, as well as the emotional state of the user, based on his face recognition. Deep ANN is used to select the values of design ergonomic indicators. A mobile application of the training system (MATS) was developed, the interface part of which is adapted based on the use of the work of two ANN. Its testing showed an increase in the effectiveness of student education, by ensuring a comfortable interaction between the student and the smartphone, as well as reducing its physiological and psycho-emotional load. Discussion. The results obtained during the study can be used to create adaptive interfaces for mobile applications, both in the field of education and in other areas in which the user works with the software interface.
A simulation model of a clustered computer system with container virtualization is proposed, enabling the assessment of the impact of dynamic partitioning of shared limited server resources on queueing delays and the performance of active containers (those processing requests). The service rate in active containers depends on the total number of containers deployed on the server and the variable number of active containers, which is dependent on the current queue length. A simulation program, implemented using the Python programming language and the SimPy library, is presented. The simulation investigates a two-level cluster system that includes a node acting as a load balancer and several servers, each composed of groups of virtual containers. A key feature of the proposed simulation program is its consideration of the dynamic dependence of the request processing rate of containers on the number of active and inactive containers that dynamically share the server's limited resources. The relationship between container service rate and the total number of containers and the active portion of those containers has been experimentally determined. The experiment found that the decrease in service rate in active containers depends non-linearly on their number. In the simulation, the load balancer node is represented as a single-channel queuing system (QS) with a simplest input stream. Servers, equipped with virtual containers, are represented as multi-channel QSs. A distinctive feature of the investigated QSs, which differentiates them from classical QSs with a common queue, is that the service channels (containers) are considered dependent due to the dynamic partitioning of shared limited server resources between active and inactive containers. The simulation considers server QS options with unlimited and limited queue lengths. The results of the simulation are compared with analytical calculations. The simulation results show the existence of an optimal number of containers deployed on servers, at which a minimum in request processing delays is achieved. The number of containers deployed on the servers should vary with changing traffic. The proposed simulation models can be applied to the optimization of clusters with container virtualization, including justifying the choice of the number of containers to deploy in cluster nodes.
The results of inclusion of the algorithm and Markov network with abnormal inference of a large correlator in the сNnet-scheme are presented. A new Markov network scheme is synthesized, in which the large correlator is assigned the role of a device detecting coincidences in words and compositions of interacting words – categories obtained as a result of processing input data samples from many sources of such. This new network scheme is developed for processing input data samples from organizational systems to classify the interactions of words denoting objects of these systems with some useful purpose. The main scientific goal of the study has been achieved: a method for introducing known classifiers, which are not formally considered to be constructed using the Markov algorithm, into the Markov network with abnormal inference has been substantiated and synthesized. The problem of research on the synthesis of morphisms of a large correlator in the сNnet-scheme of the Markov algorithm is solved. An additional contribution is made to solving a larger scientific problem: synthesizing a complete list of morphisms for Markov algorithms and networks. Practical examples of implementing Brouwer creativity in the control channel and by manipulating input data are considered. An additional source type associator typeн is introduced into the сNnet-scheme of the Markov algorithm, which specifies a specific method for processing input data samples for individual schemes that make up the network. In the Markov network with abnormal inference, the commands of the large correlator Dcorb for its processor Пcorb are defined. Associators of sources of a special type for controlling the large correlator are introduced into consideration and investigated. A formula of the Markov network with abnormal inference with an integrated large correlator is synthesized, it is shown that it must be developed to determine the result of executing the commands of the control channel. It was concluded that morphisms of a large correlator were synthesized and included in the сNnet-scheme of the algorithm and the Markov network of indirect propagation, which made it possible to determine the method for introducing known classifiers into it.
Modern times dictate the active use of numerical methods in solving various practical problems. In this study, a mathematical model of information dissemination in a small professional network was formulated using the example of the problem of scientific citations. In the course of the article, the main parameters of a professional network were identified: agents, connections, information messages and information, decision-making, and many external entities. The team of authors presented the professional network as a graph, the vertices of which are randomly connected to each other. As in any community, there are main agents in a professional network, there are standard agents, and in the course of the article, a formula for the importance coefficient of any network representative was determined. The mathematical model, presented by a differential equation, allows us to determine the time of information dissemination. Thanks to the approximation by the least squares method, the approximation coefficients and the approximating polynomial were determined, and a trend was constructed based on the values obtained. The determination coefficient is 0.99, which indicates a high reliability of the result. This mathematical model is suitable for small professional networks, because in the future, calculations will be difficult to describe with a polynomial dependence. In the example, the authors considered a network with a size of 16 participants. In the context of such a group size, the model is adequate, it can be used to predict the time of information dissemination in the context of any given task.
A methodology and program for calculating the geometric and aerodynamic characteristics of a horizontal-axial wind turbine are proposed by statistically evaluating its main operational parameters: average annual electricity generation and wind energy utilization coefficient. Confidence intervals of changes in the studied parameters are obtained. Confidence intervals using stochastic modelling are determined. The calculations varied the values of the aerodynamic torque coefficient of the blade profile, the wind energy utilization coefficient and the estimated wind speed. The values of the coefficients and velocities varied within the limits of their permissible deviations distributed according to the normal law. A number of aerodynamic profiles with maximum aerodynamic efficiency have been studied. As a result, the mathematical expectation, variance and standard deviation (RMS) of the studied parameters were determined. The range of parameter changes in the confidence interval widely used in engineering calculations was estimated equal to ±3RMS. The technique makes it possible to select blade profiles according to their aerodynamic properties, optimize control laws and geometric parameters. The methodology is applicable to the calculation of horizontal-axial wind turbines of any power, allows to set a statistical description of the tolerances for the main source data. The results of the calculation of the main operational and geometric parameters of a wind turbine of medium power class are obtained.
This scientific article examines in detail the main issues related to the development of information support for the expert electronic document management system in the area of information security. The analysis of the system's information flows is performed. A flow chart of information flows in the automated document management system in the field of information security is given. Based on the system analysis of the area of document support for information security, a model architecture of the automated electronic document management system is proposed. Its subsystems are detailed, and its components – structural modules – are identified. The interaction of these modules and subsystems is considered. An information model of the database of electronic documents in the field of information security is proposed, presented in IDEF1x notation. The features of the formation of this database are described, taking into account the specifics of the subject area and the dynamics of legislation in the field of information security. The subject area of documentation support in the field of information security is characterized by dynamic development and the lack of a unified systematic approach to the formalization of electronic document management processes. In the context of constant changes in legislation and tightening of requirements for information security, the issue of developing information support for electronic document management systems is becoming relevant. A systematic approach to the analysis of processes and information flows allows identifying key aspects that require attention, as well as developing effective models and architectures of automated control systems. The purpose of this scientific article is to develop information support for electronic document management processes in the field of information security based on an expert information and analytical system (EIAS).
The growing adoption of Kubernetes for container orchestration has intensified the need for robust security practices. This paper examines key antipatterns and vulnerabilities in securing Kubernetes clusters and proposes strategies for mitigating these risks. The focus is on evaluating traditional and modern security approaches, with an emphasis on automating security checks without compromising the performance of containerized environments. The study explores the challenges of balancing security and resource utilization in Kubernetes clusters. It discusses the limitations of deploying traditional security tools, such as antivirus software, on Kubernetes nodes, which can negatively impact system performance. The research highlights alternative security measures optimized for containerized infrastructures, such as using distroless images and immutable operating systems to reduce the attack surface. Additionally, the paper advocates for shifting security checks earlier in the development lifecycle through tools that scan container images for vulnerabilities before deployment. Automated enforcement of security policies in Kubernetes, using Validating Admission Webhooks, is explored as a method to prevent compromised images from entering production environments. The study concludes that securing Kubernetes clusters requires a multilayered approach that integrates automated tools and minimizes the performance impact. It emphasizes the need to balance security measures with operational efficiency, providing practical recommendations for achieving this in production environments.
