https://repository.hneu.edu.ua/handle/123456789/144 2026-07-08T12:22:11Z https://repository.hneu.edu.ua/handle/123456789/41335 Назва: Perplexity-based AI-generated text classification in Ukrainian using small language models Автори: Gorokhovatskyi O. Короткий огляд (реферат): The aim of the research. The rapid advancement of generative artificial intelligence language models has introduced new complexities in discerning the authorship and quality of textual content. In this paper, we explored the feasibility of using perplexity – a measure of token predictability – as the only discriminative feature for classifying AI-generated versus human-written texts in Ukrainian within the IT domain. Our approach employed small language models to calculate perplexity and detect content generated by state-of-the-art models, evaluating the potential for lightweight solutions. Research results. Initial experiments using a single perplexity threshold across Gemma 3 / Llama 3.2 1B models yielded classification accuracies around 0.70. The full token-level probability sequences were proposed as feature vectors, enabling us to achieve an accuracy of 0.68 via simple KNN classification. Finally, the convolutional neural network architectures trained on these features allowed us to obtain 0.82–0.87 accuracy. Conclusions. The comparative analysis with a traditional NLP-based discriminative neural network model revealed that direct text piece classification outperforms perplexity-based methods, although the latter still demonstrate practical utility. 2026-01-01T00:00:00Z https://repository.hneu.edu.ua/handle/123456789/41287 Назва: Intelligent Guidance Algorithms for Autonomous Unmanned Interception Systems: Smart Information Processing and Decision-Making Автори: Milevskyi S.; Brynza N.; Serhiienko O.; Mashchenko M.; Chernova N.; Dydiak R. Короткий огляд (реферат): The paper addresses the problem of autonomous guidance for unmanned interceptor systems operating against highly maneuverable small aerial targets. A comparative analytical and numerical study of fundamental guidance laws Proportional Navigation (PN), Augmented Proportional Navigation (APN), Pursuit Navigation, and Linear Quadratic (LQ) optimal control - is conducted, with emphasis on their applicability under real physical constraints of small UAV platforms, including actuator inertia, limited available overload, and measurement noise. To overcome the limitations of classical methods, a combined adaptive guidance algorithm is proposed, integrating APN-based target acceleration compensation, Zero-Effort Miss (ZEM) trajectory prediction, and a nonlinear correction term justified through Lyapunov stability theory. The Lyapunov function approach, employing the squared line-of-sight angular rate, guarantees partial stability with respect to the guidance error variable across a wide range of initial conditions. A Kalman filter is incorporated into the guidance loop to provide reliable real-time estimates of target acceleration and time-to-go under high noise conditions. Three-degree-of-freedom numerical simulations confirm that the proposed algorithm achieves a miss distance of 0.1 m - a reduction of approximately 97 % compared to classical $\text{PN}(16.5$ m) and 97 % compared to APN (3.42 m) - while reducing average interception time by 12 % with only a moderate increase in computational cost. The results validate the effectiveness of combining nonlinear adaptive corrections with stochastic filtering for autonomous terminal guidance applications. 2026-01-01T00:00:00Z https://repository.hneu.edu.ua/handle/123456789/41286 Назва: Hardware Security Evaluation of IoT Microcontrollers: Threats and Countermeasures Автори: Pohasii S.; Tkach M.; Holdobin S.; Brynza N.; Stetsenko V.; Tiutiunyk O. Короткий огляд (реферат): This study analyzes the susceptibility of IoT microcontrollers to reverse engineering, defined as unauthorized data access for replication, and proposes countermeasures. STM32F0 demonstrates superior resistance due to dense integration, rapid protection activation, and robust encryption, effectively countering attacks like Cold Boot Stepping, Glitch, Electromagnetic Analysis, and invasive methods. GD32 offers moderate security but is vulnerable to faster attacks. CH32 and MM32F0 show the least resilience due to weak protection. Architectural analysis highlights STM32F0's secure interfaces, while GD32's multi-chip design and CH32/MM32F0's simpler cores increase risks. Security metrics confirm STM32F0's lead, with GD32, CH32, and MM32F0 needing enhancements. Recommendations include hardware shielding, signal filters, and dynamic encryption, implementable in 2-14 days at $0.1-1.0 per device. STM32F0 suits high-security IoT, GD32 mid-level needs, and CH32/MM32F0 low-risk applications. Comprehensive hardware and software modifications are critical to mitigate reverse engineering threats across all microcontrollers. 2025-01-01T00:00:00Z https://repository.hneu.edu.ua/handle/123456789/41285 Назва: Analysis of software vulnerability detection methods Автори: Gryzun L.; Havrylova A.; Tkachov A.; Hapon A.; Brynza N. Короткий огляд (реферат): The author proposes a direction of improvement of existing software protection systems, focusing efforts on increasing their ability to detect new types of malware. The most promising direction for the development of technologies for detecting vulnerabilities and vulnerabilities in software is an approach that combines different methods of analysis. This allows achieving higher accuracy of the results, as well as increasing the productivity of the tools used to check the code. A method of combining static analysis of programs and dynamic symbolic execution is proposed to improve the accuracy of vulnerability detection while maintaining high performance of analysis tools. This approach will significantly reduce the risk of errors that can be missed when using one of the analysis methods separately, and also improves the efficiency of the overall software security process. 2025-01-01T00:00:00Z