SKStanislava Kudrenko

Stanislava Kudrenko

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Stanislava Kudrenko has a total of 33 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Effectiveness of anomaly detection methods in network traffic under class imbalance using the detection reliability index

    by H.V. Pekurovskyi, S.O. Kudrenko, Y. V. Kotukh

    The paper addresses the problem of network traffic anomaly detection under class imbalance, which remains a critical challenge for modern intrusion detection systems due to the limited availability of labeled attack data and the inherent scarcity of anomalous traffic compared to normal traffic. Since supervised learning approaches require representative samples of malicious traffic for every attack type, which is practically infeasible given the constant emergence of new attacks, the study employs one-class learning methods trained exclusively on normal traffic. The authors perform a comparative analysis of three such methods — a multilayer-perceptron-based autoencoder, an isolation forest, and a one-class support vector machine (OCSVM) — on a subset of the widely used CICIDS2017 dataset, which combines records of normal traffic and traffic from a LOIC HTTP DDoS attack. The merged dataset contains 755,663 records with a class imbalance ratio of IR = 4.90, and the feature space consists of 76 numerical network-flow characteristics extracted with the CICFlowMeter tool. The key scientific contribution of the work is a new evaluation metric, the Detection Reliability Index (DRI), which, unlike conventional F1-score and ROC-AUC metrics, explicitly accounts for the operational burden of false positive alerts (FPR) by introducing a multiplicative penalty in the denominator of a weighted F-like score. The metric parameters were the recall weight beta = 0.8 and the penalty coefficient alpha = 0.5, which are justified by the asymmetry between the cost of a missed attack and the cost of a false alarm for security administrators. Experimental results show that the autoencoder achieves the highest precision (0.7245), ROC-AUC (0.9154), and proposed DRI value (0.6342) owing to its minimal false positive rate (FPR = 0.049). The isolation forest attains the highest recall (0.637) but generates substantially more false alarms (FPR = 0.169), which considerably reduces its practical applicability and results in the lowest DRI among the compared methods (0.5496). The one-class SVM, trained on a limited subsample due to the computational cost of the method on large datasets, occupies an intermediate position (DRI = 0.5835). The findings confirm that maximizing recall without accounting for the false alarm rate is a suboptimal strategy for operational intrusion detection systems, and that the proposed DRI metric reveals practical advantages of methods that remain hidden when using traditional classification quality measures. The work is of practical relevance for cybersecurity practitioners selecting anomaly detection methods for deployment in real network environments with limited administrator capacity for processing security alerts, and it calls for further validation on other attack types.

  • Deepfake Detection Through Multimodal Feature Fusion and Adaptive Security Frameworks

    by Al Humam Adnan Yousif, Hanaa Hameed Merzah, Stanislava Kudrenko

  • Analysis and Detection of False-Negative Injection Attacks in IIoT Environments

    by Olena Chebanyuk, Stanislava Kudrenko, Vitalii Alkema, Anna Stoliar

  • Open Access

    Problems and prospects of building authentication methods in quantum key distribution channels

    by Ye.V. Kotukh, M.V. Korobchynskyi, V.V. Kozlovskyi, H.Z. Khalimov, S.O. Kudrenko, D.V. Kozlovska

    The paper addresses entity authentication in quantum key distribution (QKD) systems as a decisive condition of their practical security. It is shown that the information-theoretic security of quantum key agreement does not eliminate the need to authenticate the communicating parties: an unauthenticated classical channel leaves the system exposed to the man-in-the-middle attack, since the eavesdropper can run independent QKD instances with each party and reconcile two keys under full control. Existing authentication methods are analysed and classified by the underlying cryptographic primitive: symmetric schemes based on Wegman–Carter universal hashing, pre-shared and fixed keys, public-key infrastructure, two-way authentication, quantum entity/identity authentication, and zero-knowledge proofs. For each class the operating principle, advantages and limitations are determined, with emphasis on key management, scalability and trust distribution. It is established that symmetric and quantum-layer methods rely on pre-shared secrets with a quadratic growth of key material, public-key infrastructure introduces a single trust bottleneck and quantum-vulnerable primitives, while existing zero-knowledge authentication schemes are quantum and bound to the physical layer or solve network properties other than identity. A comparative analysis reveals an unresolved scientific gap: the absence of a scalable entity-authentication method that simultaneously provides non-disclosure of the secret, quantum resistance, sub-quadratic scalability and minimisation of trust assumptions. On this basis, a prospective research direction is substantiated – the construction of entity-authentication methods based on post-quantum zero-knowledge proofs operating over the classical control plane of scalable QKD networks. The requirements for such a method are formulated, and its compatibility with formal QKD security proofs is discussed.

  • Open Access

    Neural Network Approach to Ensuring the Resilience of Industrial IoT Systems under Conditions of Peripheral Node Integrity Violation

    by Stanislava Kudrenko, Valiryi Kozlovskyi, Vitalii Alkema

    The paper addresses the problem of ensuring the functional resilience of industrial IoT systems under conditions of partial integrity violation of peripheral nodes. The relevance of the study is driven by the growing cybersecurity risks in multi-layer edge-fog-cloud architectures, where peripheral devices perform primary processing and transmission of telemetry data. Degradation or compromise of individual nodes in such systems leads to the accumulation of aggregation errors, distortion of analytical results, and the potential adoption of incorrect management decisions. Traditional approaches based on the complete disconnection of suspicious nodes may negatively affect the continuity of technological processes and reduce the overall fault tolerance of the system. A neural network-based approach to maintaining functional resilience is proposed, combining real-time node state assessment with adaptive adjustment of their weighted contribution to processing results without complete exclusion from the computational loop. A formal model describing the impact of disturbances on the aggregated signal is developed. Within this model, a node state variable is introduced as an integral indicator of information reliability and the potential level of compromise. The assessment procedure is implemented using a gated recurrent unit (GRU) neural network, which accounts for the temporal dynamics of node behavior and enables the detection of anomalous deviations in telemetry streams. Simulation modeling demonstrated a reduction in system-level aggregation error and improved stabilization of the resulting signal compared to conventional threshold-based response methods. The obtained results confirm the effectiveness of the proposed approach for industrial IoT environments and its ability to ensure a balance between cybersecurity requirements, data reliability, and the continuity of technological processes.

  • Open Access

    А robust client update aggregation algorithm for counteracting model poisoning attacks in federated learning

    by С.о. Кудренко, О.в. Німич, І.г. Макєєв, В.в. Алькема

    The article considers the problem of counteracting model poisoning attacks in federated learning systems. It is shown that the client update aggregation stage is security-critical, since compromised clients may transmit intentionally distorted local updates and influence the quality of the global model. A robust client update aggregation algorithm is proposed. The algorithm combines the standard FedAvg weight, which reflects the size of the local client dataset, with an additional reliability coefficient of the corresponding update. To estimate reliability, the normalized Euclidean distance between a client update and the reference update is used. The reference update is defined as the component-wise median of the received client updates. This approach makes it possible to reduce the influence of suspicious or potentially poisoned updates without rigidly excluding them from the aggregation process. The paper presents a formal description of the algorithm, its workflow, an analysis of expected behavior for typical, heterogeneous, and suspicious client updates, and an illustrative numerical example of robust weight calculation. The proposed approach can be used to improve the resilience of machine learning and deep learning systems under cyber threats, especially in tasks related to data protection, information security, and the protection of information infrastructure in critical infrastructure environments.

  • Open Access

    METHOD FOR DETECTING ANOMALOUS LOCAL UPDATES AND ISOLATING MALICIOUS PARTICIPANTS IN FEDERATED LEARNING SYSTEMS

    by Stanislava Kudrenko, Oleksii Nimych, Ihor Makieiev

    This paper proposes a method for detecting anomalous local updates and isolating malicious participants in federated learning systems. The method is aimed at improving the resilience of distributed machine learning models to model poisoning attacks and distorted local updates. The proposed approach combines two protection levels: anomaly assessment of local updates based on their deviation from the collective update pattern and isolation of participants that repeatedly demonstrate suspicious behavior. An anomaly score is used to evaluate local updates, while an accumulated anomaly counter is applied to control long-term participant behavior. The paper presents the formalization of the proposed method, describes the adaptive anomaly threshold mechanism, and defines the procedure for forming a trusted participant set for further global model aggregation. An illustrative numerical example is provided to demonstrate the operation of the proposed approach and to show the possibility of detecting anomalous local updates while reducing their influence on the global model. A comparative analysis of the proposed approach with existing aggregation and malicious participant detection methods in federated learning systems is also presented. It is shown that, unlike conventional aggregation schemes, the proposed method provides not only suspicious update filtering but also control of repeated anomalous participant behavior. The proposed approach can be applied in federated learning systems for edge and fog environments, as well as in information systems of critical infrastructure facilities.

  • Open Access

    PROACTIVE GNSS INTERFERENCE MITIGATION AGAINST JAMMING AND SPOOFING USING COMPACT NEURAL MODELS

    by Stanislava Кudrenko

    This paper proposes a proactive approach to GNSS interference mitigation based on compact neural models designed for real-time prediction of interference conditions. Unlike conventional reactive methods, the proposed approach enables forecasting of signal degradation by estimating the probability of interference occurrence, its type (jamming or spoofing), and intensity. The problem is formulated as a time-series modeling task using a sliding observation window, allowing the system to capture temporal dynamics of GNSS signal behavior. The proposed architecture integrates a feature extraction module, a lightweight neural prediction model, and an adaptive control mechanism for proactive adjustment of receiver parameters. The activation of mitigation is governed by a threshold condition applied to the predicted interference probability. A key feature of the approach is its compliance with real-time constraints and suitability for resource-limited GNSS receivers due to the use of efficient neural architectures. The proposed method improves system robustness against radio-frequency interference and reduces positioning degradation under adverse conditions.

  • Open Access

    Trust-driven adaptive weighting for robust GNSS state Estimation under interference condition

    by Stanislava Kudrenko, Valerii Kozlovskiy, Iryna Zarubinska, Yuriy Zozulya

    Reliable state estimation in Global Navigation Satellite Systems (GNSS) becomes challenging when measurement statistics are distorted by signal interference. Jamming, spoofing, multipath propagation, and environmental disturbances may significantly increase residual variance and introduce biases in pseudorange observations, which leads to degradation of navigation accuracy and loss of estimation reliability. Conventional robustness mechanisms typically rely on threshold-based fault detection, residual monitoring, or classical robust estimators. Although such approaches mitigate large outliers, they often employ binary measurement exclusion or purely statistical reweighting and therefore do not explicitly account for the reliability of individual measurement channels. This paper proposes a trust-driven adaptive weighting framework for GNSS state estimation under interference conditions. The method introduces a dynamic trust coefficient assigned to each measurement channel based on statistical quality assessment and risk-aware evaluation of its potential impact on the navigation solution. Residual variance is used to quantify measurement stability, while a channel risk factor incorporates geometric sensitivity and anomaly likelihood. The resulting trust coefficients regulate measurement influence through adaptive modification of the measurement covariance matrix within the Extended Kalman Filter.

  • Open Access

    МЕТОД ПРОГНОЗУВАННЯ КОМПРОМЕТАЦІЇ ВУЗЛІВ У СЕРЕДОВИЩІ EDGE ТА FOG ДЛЯ ОБ’ЄКТІВ КРИТИЧНОЇ ІНФРАСТРУКТУРИ

    by Станіслава Кудренко, Олексій Німич, Ігор Макєєв

    У статті запропоновано метод прогнозування компрометації вузлів у середовищі edge та fog для системкритичної інфраструктури. Актуальність дослідження зумовлена зростанням кількості кібератак нарозподілені IoT-мережі, у яких порушення роботи окремих компонентів може призводити до втрати цілісностіданих, зниження надійності вузлів та порушення функціонування інформаційної інфраструктури.Запропонований метод базується на аналізі поведінкових і мережевих характеристик компонентів edge/fog-середовища, формуванні вектора ознак, розрахунку показника аномальності, інтегрального ризикукомпрометації, рівня довіри та узагальненого показника небезпеки. Особливістю підходу є поєднання детекціїаномалій із прогнозуванням станів, що дозволяє не лише виявляти поточні відхилення у поведінці вузлів, а йоцінювати ймовірність їх переходу до потенційно небезпечного стану. Використання елементів машинногонавчання та нейронних мереж створює передумови для адаптації методу до динамічних умов функціонуваннярозподілених систем. Запропонований підхід може бути використаний для підвищення рівня захисту даних,своєчасного виявлення аномальної поведінки та підтримки проактивних механізмів безпеки мереж IoT усистемах критичної інфраструктури.