МЕТОД ПІДВИЩЕННЯ РІВНЯ ЖИВУЧОСТІ МЕРЕЖІ В УМОВАХ ГІБРИДНИХ КІБЕРАТАК ПІД ЧАС ВІДКЛЮЧЕНЬ ЕЛЕКТРОЕНЕРГІЇ

<jats:p>The subject of the study is the development process of SOHO (Small Office / Home Office) electronic communications networks based on a compact GPON (Gigabit-capable Passive Optical Network) mini OLT (Optical Line Terminal) with one PON (Passive Optical Network) port for up to 32 ONU/ONT (Optical Network Unit/Optical Network Terminal) subscriber terminals. This study aims to develop and validate an enhanced network survivability approach under hybrid cyberattacks during blackouts. The research tasks are as follows: a formal model of SOHO access survivability should be developed that considers the UPS energy budget and the structure of active nodes; a hybrid risk model in which energy failures are combined with cyber compromise should be formalized; the monitoring methodology IDS+SIEM (Intrusion Detection System + Security Information and Event Management) with machine learning should be developed; a comparative analysis of Ethernet and GPON architectures should be performed; practical recommendations for the implementation, diagnostics and streamlining of management should be formulated in accordance with the principles of Zero Trust and the requirements of information security management, taking into account the human factor and the ease of use of controls in SOHO. The following results were obtained. The survivability model was developed that combines the power backup circuit (UPS autonomy assessment), the availability of the active circuit, and a probabilistic model of the risk of compromising control components, including firmware attacks related to SNMP (Simple Network Management Protocol) and SSL/TLS (Secure Sockets Layer/Transport Layer Security) vulnerabilities in embedded software. A monitoring methodology that integrates IDS and SIEM with deep-learning-based analytical model classes was developed and represented by AE+LSTM-type (Autoencoder + Long Short-Term Memory) telemetry analysis and CNN+LSTM-type (Convolutional Neural Network + LSTM) firmware-oriented analysis. Comparative modeling of Ethernet Everywhere and GPON architectures with passive branching was performed for 5–20–32 connection scaling scenarios. Numerical examples of calculations, comparative tables of metrics, and the practical algorithm for diagnosing typical installation errors (APC/UPC, IN/OUT splitter, incompatibility of standards) were provided. The results were interpreted using Zero Trust principles for small networks and the prospects of multimodal AI for correlating SIEM events with firmware artifacts and physical optical channel telemetry were identified. Practical recommendations regarding segmentation, update policies, and privilege minimization for access control in the SOHO segment were offered. Conclusions. This study determines that enhancing network survivability is an important and urgent task. The scientific novelty of the results obtained is as follows: the integrated monitoring methodology with machine learning that ensure the transition from component composition and operational assumptions to numerical metrics of survivability, availability, and detection was developed; the formal survivability model was developed considering the UPS energy budget and the structure of active nodes was developed; the risk of hybrid events in which energy failures are combined with cyber compromise was formalized. Areas for further research: failure localization methodology; methodologies related to Zero Trust; multimodal AI analysis; decision support under conditions of limited SOHO resources.</jats:p>