Адаптивний метод забезпечення безпеки індивідуальних конфіденційних переговорів
O. A. Laptiev, Тетяна Лаптєва, Ян Рабчан et al. · The Scientific Issues of Ternopil Volodymyr Hnatiuk National Pedagogical University Series pedagogy · 2026
Subject of Research. An adaptive mechanism for ensuring confidentiality of acoustic speech in real time, which dynamically balances the level of secrecy and audio signal quality based on the analysis of semantic load, background noise, and external threats. Research Methodology. The problem is formalized as a multi-criteria optimization task with an objective function combining secrecy (Sec) and quality (Q) metrics through a priority-compromise coefficient α. Synchronized chaotic systems are employed for masking, generating noise within the 300–3400 Hz frequency band. The energy distribution of the masking noise is optimized using psychoacoustic models (MPEG, ITU-R). Computer simulations were implemented in Python, with performance evaluated using PESQ, MOS, and Sec metrics. The input signal is analyzed in real time to detect keywords, energy levels, pauses, and sensitive information. Results. The proposed method dynamically adjusts the parameters of chaotic noise depending on α, achieving high confidentiality (Sec = 0.87 at α = 0.60) with acceptable distortion (Q = 0.30), or preserving audio quality (Q = 0.47 at α = 0.30) under moderate protection. The system reactively intensifies masking during the transmission of semantically sensitive information (e.g., names, numerical data, financial details) and reduces it during non-critical segments. The chaotic nature of the masking noise ensures robustness against adaptive filtering: signal recovery is feasible only with the synchronization key. Psychoacoustic models enable energy-efficient concentration of masking power in the most perceptually sensitive frequency bands while maintaining protective efficacy. Consequently, the approach transforms passive protection into an intelligent, context-aware mechanism capable of real-time self-organization. Future integration of AI-driven automation of α (via NLP and neural networks) paves the way toward autonomous systems resilient to modern threats, particularly MEMS-based eavesdropping.
Cited by 0 works.