Certification of non-destructive testing specialists entails legal consequences, so the authenticity of the test result should be treated as an information-security task. Password protection confirms the subject's identity only at login and does not control who actually answers the questions during the session. Behavioral biometrics, covert in the way features are collected, closes this gap. This paper assesses the reliability of confirming an examinee's identity by the dynamics of single mouse clicks and identifies the most discriminative features. The data source is the log of an operational certification system accumulated since 2020, containing over 3.4 million click events from 250 users. The hold duration and the relative coordinates of the click point are recorded for each click. A set of sixteen session-level features and an open-set verification scheme using a random forest as the baseline model are proposed. Interval quality estimates are obtained by user-level bootstrap. On the test set, an equal error rate (EER) of 18.3 % was achieved for single-session scoring (95 % confidence interval [16.2; 19.5] %) with an area under the characteristic curve of 0.899. Fusing decisions over several sessions reduces the error to fractions of a percent (0.47 % for twelve sessions), which, given the statistical dependence of one user's sessions, should be understood as an optimistic bound. The spatial features of the click point, complemented by the button hold duration, contribute most to discriminability. The applicability of click dynamics as an additional, covertly collected factor for confirming subject authenticity in certification systems without extra hardware is demonstrated.

The aim is to study the possibilities of reducing the average latency in a multipath wireless network. Delays are minimized by adaptive traffic redistribution in nodes that act as switches between alternative data delivery channels to the target recipient. To conduct the experiments, a simulation model has been developed in the OMNeT++ environment using the INET framework. A simulation model of a multipath network with the possibility of route segmentation by redirecting data in switch nodes is presented, taking into account the influence of physical obstacles on average network delays. The dependence of the average end-to-end delay on the proportion of traffic redirected through each of the routes, as well as on the location of physical obstacles in the network, is determined. The existence of an optimal traffic distribution for different network configurations is shown, taking into account the variable location of signal propagation obstacles.

The rapid expansion of millimeter-wave frequencies in 5G cellular networks necessitates the use of deterministic methods for calculating radio coverage that take into account multipath signal propagation in complex urban environments. In this regard, ray tracing technology, particularly the SBR method, has become increasingly wide-spread. It allows for the identification of all significant signal propagation paths between a transmitter and receiver. The change in amplitude, phase, and polarization for each ray is estimated by calculating its angles of incidence and reflection at the points of interaction with environmental elements whose surfaces possess specific electro-physical properties. The need to analyze existing software implementations of the SBR method is driven by the need to verify, validate, and further modify the ray tracing model in the Matlab environment.

This paper presents an antenna design that enables the generation of vortex waves or waves with orbital angular momentum of the first two modes over a wide wavelength range. To implement the radiation, a design based on a radial waveguide and circularly polarized helical antenna elements is proposed. To generate vortex radiation, mechanical rotation of the antenna elements around their axis is used to create a phase shift. Based on this design, an analysis of the generation of vortex waves of the first two modes over a wide wavelength range is presented. The paper presents the characteristics of vortex waves, as well as their generation depending on the operating frequency ranges of the antenna array. The presented results demonstrate the possibility of generating vortex waves over a wide frequency range using a ring antenna array with an excitation system based on a radial waveguide.

The causal relationship between human capital and the level of current and prospective socio‑economic relations is demonstrated. The negative factors of the learning process are examined in the context of increasingly complex admission conditions. A comprehensive adaptive model and a method for Pareto‑optimal management in conducting a university’s admissions campaign, based on a fuzzy‑set framework, are proposed; they link the individual psychological characteristics of applicants (qualitative features) with performance indicators during the learning process and in the labour market. The process of conducting the acceptance campaign is presented as a black box, the mathematical description of which is provided through two parameter transformation circuits based on the Mamdani and Sugeno fuzzy algorithms. An integral efficiency criterion has been formed, ensuring the system’s transition to its target state. The implementation of the described tools has been justified through the formalization of requirements and constraints, which serves as a guideline for the development of harmony, well‑being, security, and economic development of society

This paper presents an experimental verification of the parallax-shift method for non-contact determination of height differences of fuel assembly heads from a series of television images obtained during arc-shaped movement of a single CCD camera. Consecutive video frames are considered as quasi-stereo pairs, from which the disparity of the centers of fuel assembly heads is calculated. An image processing algorithm has been implemented, including noise suppression, CLAHE contrast enhancement, adaptive binarization, morphological filtering, the Hough transform, and subpixel localization. The measurement results obtained with different quasi-stereo pair baselines are compared, and the influence of subpixel localization on the final accuracy is evaluated. It is shown that the most stable results are achieved with a baseline of 12 mm, at which the average error in determining height differences is approximately ±1.3 mm. The obtained results confirm the fundamental feasibility of the parallax-shift method for television-based inspection of objects with a regular spatial structure.

The rapid spread of Sharia- and law-related disinformation across Arabic digital platforms poses serious risks to public safety, social stability, inter-confessional relations, and national information security in Arab states. Existing detection approaches remain largely limited to binary “fake/real” classification and rarely incorporate national legislation or Sharia norms, nor do they adequately support operational monitoring and counter-disinformation systems. To address this gap, this paper proposes a hybrid deep learning system based on AraBERT, CNN, and BiLSTM with a cross-attention mechanism for multi-class detection, classification, and risk assessment of religiously and legally sensitive harmful content. The system is explicitly designed as a core technical component for national monitoring, early warning, and response platforms, where classification outcomes can support risk prioritization and operational decision-making. Experiments on the newly developed ArabFake-2025 dataset, which comprises 38,750 documents across five Sharia-oriented categories, demonstrate strong performance, with 96.9% accuracy and 96.3% Macro-F1. Ablation studies further confirm the complementary contribution of each architectural component. this work contributes methods and models for detecting and mitigating false and malicious information through formalized threat modeling, legal risk evaluation, and pathways for practical operational integration.

The article proposes a method for embedding hidden data into a message using a BCH code that corrects one error. It is based on a block steganographic algorithm in the spatial data area, based on the choice of minimal distortions. The set of container message blocks are represented as a vector space. The corrective BCH code is used to partition the vector space into syndrome equivalence classes. Several bits of the hidden message are embedded in each block by changing only one bit. The selection of the bit to be changed is based on a scheme for determining the changed bit using a BCH code that corrects one error. This approach increases the capacity of the hidden channel with a constant level of container distortion. The proposed embedding scheme is applied to bitmap images for embedding a hidden message in a zero-bit layer. The 8-bit block partitioning used. In a computational experiment, image distortion metrics are determined. It is shown that the throughput increases 6 times with an increase in distortion 2 times.