Centralized intrusion detection in distributed corporate infrastructures (branches, remote offices, and enterprise/industrial IoT) creates privacy and compliance constraints because it requires centralizing sensitive telemetry. We study a privacy-preserving IDS design based on federated learning (FL) with differentially private local training (DP-SGD). Clients train locally on flow-derived features and structured event telemetry, and only model updates are shared with a coordinator for aggregation. We report the privacy budget (ε, δ) using an RDP accountant and evaluate detection quality using Accuracy and Macro-F1. Experiments on CICIDS 2018 and the CERT Insider Threat v6.2 dataset show the expected privacy-utility tradeoff: DP training reduces utility compared to non-private centralized learning, and FL with DP typically incurs an additional decrease under heterogeneous (non-IID) client partitions, while remaining practical at explicit privacy budgets (main comparison: ε ≈ 1.8 -2.0). We also report SOC-relevant operational indicators: training time, peak memory usage, inference latency, and model size-and compare against Random Forest and XGBoost baselines.

 The article discusses the role of graph neural networks (GNNs) as a new
computational tool in the development of the theory of semantic-associative analysis of text data. GNNs allow us to formalize associative connections between lexical and conceptual units in the form of dynamic semantic graphs, which brings machine models of language processing
closer to the cognitive mechanisms of human thinking. A conceptual architecture based on Graph Attention Networks is proposed, integrating linguistic dependencies and external semantic knowledge, which ensures contextual adaptation and interpretability of the analysis.
The theoretical contribution of GNN to overcoming the limitations of traditional statistical and vector models is substantiated: the transition from isolated representations of words to modeling the spread of semantic activation through a network of associations. The results of preliminary
experiments (based on synthetic data and the RuSentiment corpus) demonstrate an increase in the accuracy of semantic classification and interpretability of conclusions. The article contributes to the interdisciplinary synthesis of computational linguistics, cognitive science, and machine learning, paving the way for the construction of computable theories of meaning generation in texts.

The article investigates the resilience of the adaptive authentication model to adversarial attacks via hybrid digital fingerprint imitation. An adversarial extension of the adaptive authentication model based on a digital fingerprint combining technical and behavioral attributes is proposed. The attack is formalized as a conditional optimization problem with separate values of permissible perturbations of technical (ε_d) and behavioral (ε_b) attributes under conditions of limited feedback. To evaluate model resilence, a probabilistic resilience measure R is introduced, consistent with the classical FAR/FRR metrics. Experimental validation was performed on a dataset of 1280 sessions (32 users, 30 attributes) using the CMA-ES algorithm. The experimental results showed that models with adaptive weighting provide a twofold increase in resilience compared to the model with equal weights in the realistic attack regime (ε_d ≤0.2): R=0.66-0.79 versus 0.39–0.53, and reduce the probability of a successful attack to 0.36 under a strategy for optimizing technical and behavioral features. It has been experimentally established that the automatic reduction of weights with an increase in the variance of features complicates adversarial attacks by attackers. The identified security vulnerability at ε_d ≥0.5 (R≤0.07) defines the limits of the model's applicability and justifies the need for multifactor authentication with a high proportion of perturbations.

This paper presents a method for peak factor (PAPR) reduction in hybrid satellite communication networks using orthogonal access methods and shaping filters. The proposed approach combines flexible protective intervals and formative filtering methods to optimize signal transmission in low-orbit satellite networks. Several types of filters have been investigated, including the raised cosine (RC), the root raised cosine (RRC), the «better than» raised cosine (BTRC) and its optimized version (OBTRC). A new variant of the shaping filter based on OBTRC one is proposed. The method is evaluated by emulation low-orbit satellite communication networks. OFDM and DFT-s-OFDM technologies with zero-tail and unique word modifications. The simulation results show a decrease in PAPR by 0.5-2 dB compared to traditional approaches.

The goal of the article is to describe the development and progress of testing, on the basis of Novosibirsk State University, of a Smart Campus concept based on the approaches and best practices of the Smart City concept, taking into account adaptation to the specifics of the university environment.

The research is based on the translation of Smart City approaches into the university environment, taking into account differences in scale, actors, and the specifics of the university setting. The main methodological approaches include the analysis of strategic and regulatory documents, process modeling, as well as pilot implementation at the Novosibirsk State University site.

The key result is a comprehensive Smart Campus concept for Novosibirsk State University: key domains are described (a digital twin of the university, an infrastructure and resource management system, security, financial planning, a unified digital environment, information and client services), a smart campus governance model is described, and a roadmap and a system of KPIs for the stage-by-stage evaluation of the system's effectiveness have been developed.

The practical value of the results lies in the possibility of transferring the best practices and systems, tested in real-world conditions, to other universities.

The target audience of the article includes researchers in the field of education management and digital transformation, university and campus management teams, representatives of federal and regional government bodies, and technology partners.

The article examines the key principles for building an information security subsystem for MLOps (Machine Learning Operations), taking into account the increasing relevance and vulnerability of machine learning systems in modern conditions. The work considers the main regulatory requirements for information security of systems using machine learning, including an analysis of the requirements of the Russian Federal Service for Technical and Export Control (FSTEC). It argues for the need to adapt modern MLSecOps solutions to the conditions of Russian legislation. The analysis covers the key cybersecurity risks characteristic of each stage of the MLOps lifecycle, as well as existing methods and approaches for mitigating them. The work presents a mathematical formalization for assessing the security of the system, taking into account technical, regulatory, and import substitution requirements. The conceptual structure of the information security subsystem is presented. The results of the research are aimed at creating an effective and reliable protection system for MLOps pipelines and reducing the risks associated with the use of machine learning technologies.

The paper presents adaptation of a deep learning model based on residual neural network ResNet with CBAM attention modules for predicting nitrogen, phosphorus and potassium content in soils of central Russia using visible and near-infrared spectroscopy. The model pre-trained on the global LUCAS library was fine-tuned on a local sample of 100 soil samples from Voronezh, Kursk, Rostov regions and Krasnodar Krai. A partial layer freezing strategy was applied. After fine-tuning, the coefficient of determination of 0.91 and the ratio of performance to deviation of 3.01 were achieved for nitrogen, and 0.56 and 2.40 for phosphorus, respectively. The possibility of using portable spectrometers in precision farming systems is shown.

This article proposes a method for assessing the dynamic availability of telecommunications network equipment, taking into account seasonal variations in grounding device parameters. A mathematical model has been developed that accounts for the influence of soil temperature and humidity, thunderstorm activity, freeze-thaw processes, and corrosion on soil resistivity and grounding resistance. It is shown that in the spring, when the ground has not yet completely thawed and thunderstorm activity is already increasing, equipment availability decreases due to deterioration in grounding parameters. The need for continuous monitoring of grounding device parameters to promptly identify deviations and maintain the required level of reliability of telecommunications equipment is substantiated.