Information Technology and Society

AUTOMATED VERIFICATION OF STATEMENTS USING THE RAG MECHANISM AND SYMBOL CLASSIFICATION

Vitalii Dadyverin, Oleg Bisikalo — Mon, 01 Jun 2026 00:00:00 +0300

The object of the study is the problem of automatic fact verification in a digital environment saturated with disinformation. The paper analyzes modern approaches to fake news detection, including transformer architectures, neurosemantic and graph models. Additionally, the limitations of existing methods are identified, in particular, the popularity of the use of static features and poor generalization ability in a constant dynamic flow of information. The author proposes his own architecture of a multimodal model that combines style classification, AI text detection and a fact-checking module, supported by the search for relevant evidence through the RAG mechanism. The results of experiments on a test set of 1660 examples showed that the model achieves a high Recall indicator (84.6 %), while maintaining an acceptable balance of accuracy (Accuracy – 78.6 %, Precision – 74.4 %, F1 – 80.8 %). The obtained results indicate sufficient effectiveness of multi-task learning in truth-checking systems. This allows for effective detection of fake news from various sources, albeit with a certain number of false positives, but the balance between high Recall and lower Precision is justified, since the system is focused on reducing the possibility of missing fake news. The proposed model is suitable for use in real-world monitoring of the information space, in particular in the context of countering information threats. The effectiveness of the model is explained by the combination of several independent features (style, origin, factuality) and a flexible signal integration system. In addition, the use of the RAG mechanism provides an additional level of interpretability of the results obtained with reference to external sources. It can be used in online platforms with a large number of unstructured messages. The approach can be expanded with multimedia analysis and adapted for another specific language environment.

DYNAMIC ADAPTATION OF THE USER PROFILE IN A RECOMMENDATION SYSTEM BASED ON ANALYSIS OF INFORMATION ABOUT HIS BEHAVIOR

Mon, 01 Jun 2026 00:00:00 +0300

The rapid growth of digital content volumes and the increasing complexity of user behavioral patterns have intensified the requirements for the accuracy and adaptability of modern recommender systems. Under these conditions, methods for dynamic user profile updating become particularly important, as they enable the maintenance of relevant personalization in the presence of changing interests, contextual influences, and the irregular nature of user interaction with information services. The problem lies in the need to ensure the adaptation of such systems to changes in user interests in a dynamic environment, where short-term preferences change rapidly, and long-term ones retain inertia, which makes it difficult to maintain high personalization accuracy. The aim of the research is to develop a mathematical model and architectural solution for dynamic user profile adaptation, which allow to effectively combine the analysis of long-term and short-term interests taking into account contextual factors to increase the accuracy of personalization in recommender systems. The work is aimed at overcoming such shortcomings of existing systems as high sensitivity to random «noise» in user actions, complexity of interpreting updates and inability to respond in a timely manner to smooth changes in behavioral patterns. The work introduces a mechanism for dynamically managing the balance between short-term (STI) and long-term (LTI) interests using an adaptive coefficient, which allows the system to automatically switch to the current needs of the user in the event of a sharp change in behavior or rely on stable habits during stable interaction. A method for integrating contextual factors into the user profile update process has been developed, which provides the ability to enhance or weaken the significance of events depending on external conditions (time, device, location), transforming the profile into a context-sensitive structure. The technological approach is based on the architecture of pipeline data processing for streaming analysis of events in real time. Experimental validation of the proposed solutions was carried out by testing on the MovieLens dataset with a comparative analysis of the accuracy of recommendations of dynamic models relative to static ones. A characteristic difference of the proposed approach is the integration of contextual parameters into the profile update process, which turns it into a context-sensitive structure capable of adapting to interaction conditions. Technical tests have proven the ability of the recommendation system to work stably in conditions of intensive data flow, ensuring rapid updating of the system’s knowledge about a person without loss of overall performance. Practical application is possible in recommendation systems that operate in real time and work with intensive data flows, in particular in media services, e-commerce and information platforms, provided that there are means of collecting user events and a stream processing infrastructure.

MATHEMATICAL MODEL OF ADAPTIVE CONTROL OF UAV MOTION USING PHASE COORDINATES

Svitlana Kashkevych, Dmytro Mykolyuk — Mon, 01 Jun 2026 00:00:00 +0300

This article examines the problem of adaptive motion control for unmanned aerial vehicles under conditions of dynamic changes in the external environment and limited computational resources of on-board systems. An analysis of modern approaches to the control of unmanned aerial vehicles has been carried out, in particular swarm intelligence methods, neural network algorithms and classical optimisation approaches, which has made it possible to identify their main advantages and limitations in terms of accuracy and computational complexity. An approach is proposed for constructing a mathematical model of unmanned aerial vehicle motion using phase coordinates, which reduces the dimensionality of the control problem without losing the informativeness of the flight dynamics description. A phase state vector has been formulated, which includes velocity, angular and spatial motion parameters, allowing the synthesis of control actions based on the current state of the system. A mathematical model of adaptive control has been developed, based on the formation of indirect and direct control vectors, taking into account the design characteristics of unmanned aerial vehicles and external disturbances. Simulation modelling of the motion of unmanned aerial vehicles in phase coordinates has been carried out, the results of which confirm the adequacy of the model, the stability of the control system and the correct execution of the specified program trajectories. The proposed approach was implemented in the Mission Planner environment, confirming that the developed model can be integrated into existing automatic control systems without significantly increasing the computational load. The results obtained demonstrate the feasibility of using phase coordinates to improve the efficiency of adaptive control of unmanned aerial vehicles.

A METHOD OF RISK-BASED QUALITY MANAGEMENT OF AN IT PRODUCT BASED ON DATA ANALYSIS AND MACHINE LEARNING WITHIN SDLC

Yurii Kish, Ihor Liakh — Mon, 01 Jun 2026 00:00:00 +0300

The article is devoted to the development of a risk-oriented quality management method for IT products within the software development life cycle. The object of the study is the process of ensuring software quality under conditions of uncertainty and limited resources, while the problem addressed is the lack of an integrated mechanism for quantitative assessment, prediction, and prioritization of quality risks based on SDLC data. The paper proposes a formalized mathematical model in which risk is interpreted as a probabilistic characteristic of software component defectiveness, as well as a mechanism for integrating risk assessments into quality assurance processes in the form of adaptive quality gates and a decision support system prototype. The obtained results made it possible to solve the identified problem due to the combination of probabilistic modeling, machine learning methods, and risk-oriented test prioritization, which ensures the transition from static control to data-driven quality management. The results are explained by establishing relationships between software metrics and defect probability, enabling component ranking by risk level, formation of a risk register, and optimization of verification resource allocation. Experimental validation on the open NASA Metrics Data Program dataset demonstrated that the model achieves ROC-AUC = 0.669 and PR-AUC = 0.382, and in scenario analysis provides recall@top-k ≈ 0.38 compared to ≈ 0.37 for the LOC-based approach and ≈ 0.18 for random selection. Practical application of the results is advisable in software quality management systems, DevOps analytics, CI/CD environments, and test prioritization processes, provided that historical data on code metrics, testing outcomes, module characteristics, and defectiveness are available under resource constraints.

METHOD OF SEMANTIC PREFILTERING OF CAUSAL RELATIONSHIPS IN HIGH-DIMENSIONAL NETWORKS

Oleksii Kovenko, Nataliia Apenko — Mon, 01 Jun 2026 00:00:00 +0300

The object of research is the process of discovering causal relationships in high-dimensional networks. The addressed problem is the exponential computational complexity of classical structural learning algorithms and their inability to operate effectively without historical statistical observations. The obtained results involve the development and empirical validation of a novel semantic pre-filtering method for causal graphs. Relying on the formulated semantic sparsity hypothesis, the method narrows the search space based solely on node metadata. The algorithm comprises four stages: generating textual interpretations for each node using language models, transforming these descriptions into dense numerical vectors, calculating a cosine similarity matrix, and applying an adaptive pruning strategy. These results solved the problem by rejecting up to 88.3 percent of irrelevant node pairs for massive graphs while preserving over 90 percent of true causal edges. This efficiency is explained by the fact that in large systems, causal links occur predominantly between semantically related entities. Vectorization models quantify this semantic proximity and scale naturally with network size. The approach can be practically applied in high-tech domains as a pre-filtering tool before executing traditional causal discovery algorithms. Conditions for its effective use include high-density networks and scenarios where collecting massive historical datasets is technically impossible, provided qualitative metadata is available.

ENTERPRISE OSINT FOR RISK MANAGEMENT, MONITORING THE DIGITAL FOOTPRINT OF THE COMPANY AND EMPLOYEES

Valeriia Slatvinska, Viacheslav Bevza — Mon, 01 Jun 2026 00:00:00 +0300

The object of the study is the process of identifying, interpreting and using open-source data for enterprise cyber risk management under the conditions of an expanding digital perimeter. The problem is that traditional internal monitoring mechanisms do not provide early detection of leaks, compromised accounts, shadow digital assets and behavioral signals connected with the digital footprint of employees. The paper improves the approach to Enterprise OSINT as a continuous cycle of collection, normalization, verification and correlation of external indicators with internal security events. The results include a structural architecture model of Enterprise OSINT, a matrix of threat vectors and detection methods, and a procedure for integrating OSINT data into the ISO/IEC 27001, SIEM and CTI control loop. The proposed results solve the problem of fragmented external monitoring through the combination of technical, organizational and analytical components within a single risk management contour. Their distinctive feature is the focus not only on the company infrastructure, but also on employees digital footprints, partner mentions, leaks in the Surface, Deep and Dark Web, and subsequent false-positive verification. The results are explained by the fact that external data are treated not as background information, but as operational risk indicators suitable for automated validation and prioritization. Practical use is possible in corporate information security systems, SOCs, economic security services and compliance units provided that ethical monitoring policies, response procedures and repeated source verification are in place.

HYBRID VISUAL–INERTIAL SLAM ARCHITECTURE FOR CONTINUOUS NAVIGATION IN LOW-VISIBILITY CONDITIONS

Oleksandr Usov — Mon, 01 Jun 2026 00:00:00 +0300

The relevance of the study is determined by the need to ensure reliable and continuous navigation of autonomous mobile systems under low-visibility conditions, where conventional visual localization methods lose effectiveness, while inertial sensors are subject to error accumulation. Under such conditions, the integration of heterogeneous sensor data within hybrid architectures capable of adapting to variations in input data quality becomes critically important. The aim of the study is to theoretically substantiate and practically generalize approaches to the development of a hybrid visual–inertial SLAM architecture focused on ensuring continuous and robust navigation in low-visibility environments. The research methods are based on system analysis, generalization of modern scientific approaches, comparative analysis of algorithmic solutions, and structural-functional modeling of visual–inertial data integration processes in navigation systems. As a result of the study, the principles of sensor data integration and approaches to constructing hybrid SLAM architectures have been investigated, and methods for improving localization accuracy using inertial measurements have been generalized. It has been established that system efficiency is determined by adaptability to visual data degradation and the ability to compensate for inertial drift. Quantitative analysis has shown that the proposed hybrid architecture enables a reduction of inertial drift accumulation by 65–70 %, a decrease in computational latency by 60–70 %, and an improvement in localization accuracy by 6–8 times under critical conditions (reducing RMSE from approximately 1.7–1.8 m to 0.18–0.25 m). The scientific novelty lies in the systematic generalization of principles for constructing hybrid visual–inertial architectures, taking into account sensor data degradation, and in substantiating adaptive approaches to ensuring robust localization. The practical significance of the obtained results lies in their applicability to the development of navigation systems for unmanned platforms, mobile robotics, and rescue operations, where the ability to operate in real time under low or variable visibility conditions is critical.

A CASCADE METHOD FOR IMPROVING OBJECT DETECTION SYSTEMS

Lyudmila Yudina, Yuri Degtyar — Mon, 01 Jun 2026 00:00:00 +0300

The paper addresses the problem of instability of neural network-based object detection systems deployed on unmanned aerial vehicles (UAVs) under dynamically changing environmental conditions. It is shown that conventional universal models, including the YOLO family architectures, demonstrate a significant degradation of key performance metrics such as Precision and Recall when operating under varying weather conditions, illumination changes, and different object scales. The aim of the study is to improve the efficiency and robustness of real-time object detection systems by developing an adaptive cascade-based processing method. The proposed Cascade YOLO approach is based on sequential utilization of specialized neural network models combined with an adaptive switching mechanism depending on the confidence level of the current detection. Unlike classical ensemble approaches, the proposed method avoids reprocessing the same frame and instead applies the next model to a new incoming frame, which prevents latency accumulation and preserves real-time performance. The research methodology includes the development of a mathematical model of the cascade algorithm, its software implementation, and experimental validation on heterogeneous video datasets with varying observation conditions. Comparative analysis with a baseline Single-YOLO model demonstrated an increase in Precision up to 0.867 and Recall up t 0.824, with only a minor decrease in processing speed (within 5–10 %). The obtained results confirm the effectiveness of the proposed approach in enhancing the reliability of UAV-based computer vision systems. The practical significance of the research lies in the possibility of integrating the method into onboard data processing systems to ensure stable performance under uncertainty and dynamic environmental conditions.