Vadym Shtanko

Vadym Shtanko

Auto-generated profile, based on publicly available information

Vadym Shtanko

Національний університет біоресурсів і природокористування України, Україна

Research Focus & Metrics

Academic Career Length

3Years (2024-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
25 December 2024 in Eastern-European Journal of Enterprise Technologies
Verified
Journal Article
Open access
PDF file
NKNurbek Konyrbaev
Nurbek Konyrbaev
YNYevheniy Nikitenko
Yevheniy Nikitenko
VSVadym Shtanko
Vadym Shtanko
VLValerii Lakhno
Valerii Lakhno
Zharasbek BaishemirovZharasbek Baishemirov
Zharasbek Baishemirov
EMErkebula Myrzabek
Erkebula Myrzabek
This study takes a look into the application of the Naive Bayes machine learning algorithm to enhance the accuracy of Intrusion Detection Systems (IDS). The primary focus is to assess the algorithm's performance in detecting various types of network attacks, particularly Denial of Service (DoS) attacks. This research proposes using Naive Bayes to improve intrusion detection systems that struggle to keep pace with evolving cyber threats. This study evaluated the efficiency scores of the Naive Bayes classifying model for two different dependency scenarios and identified strong and weak properties of this model. The Naive Bayes classifier demonstrated satisfactory results in detecting network intrusions, especially in binary classification scenarios where the goal is to distinguish normative and malicious traffic due to its simplicity and efficiency. However, its performance declined in multi-class classification tasks, where multiple types of attacks need to be differentiated. The study also highlighted the importance of data quality and quantity in training machine learning models because of the impact of those parameters on the model efficiency. The USB-IDS-1 dataset, while useful, has limitations in terms of the variety of attacks. Using datasets with a wider range of attack types could significantly improve the accuracy of IDS. The findings of this research can be applied to such domains as network security, cybersecurity, and data science. The Naive Bayes classifier can be integrated into IDS systems to enhance their ability to detect and respond to cyber threats. However, it is essential to consider the limitations of the algorithm and the specific conditions of its environment. To maximize the effectiveness of the Naive Bayes classifier, it could be promising to optimize and normalize the data to improve the accuracy of the model and combine Naive Bayes with the other machine learning algorithms to address its limitations

14 January 2026 in TECHNICAL SCIENCES AND TECHNOLOGIES
Verified
Journal Article
Open access
PDF file
VSVadym Shtanko
Vadym Shtanko
YNYevheniy Nikitenko
Yevheniy Nikitenko
ЄНЄвгеній Нікітенко
Євгеній Нікітенко
The growing complexity of cyber threats, especially multi-vector DDoS attacks, together with the rapid increase and heterogeneity of network traffic, creates a clear need for intrusion detection systems that can remain effective under highly dynamic network conditions. This work is motivated by the observation that conventional IDS approaches often fail to provide timely and reliable recognition of sophisticated attacks when they are hidden within large volumes of background traffic. For the field of information technology, it is therefore important to develop models that not only deliver high classification quality but can also be realistically applied in scenarios that are close to real time. At the same time, direct single-stage multiclass classification of network attacks remains a difficult task: traffic behaviour is highly variable, signatures of different threats can overlap with normal flows, and changes in the network environment directly affect the stability and consistency of the results. Addressing this challenge is essential for improving IDS reliability and for ensuring that attack types are correctly distinguished under changing traffic conditions. The aim of this study is to design and experimentally evaluate a two-level classification model that combines initial separation of suspicious flows with subsequent refinement of attack types, with the goal of achieving more stable and informative diagnostics. To reach this aim, machine-learning methods were employed, and dedicated tools for collecting, aggregating, and labeling traffic were developed and used. The proposed model was tested both on the public USB-IDS-1 dataset and on a custom traffic dataset generated in a controlled virtual environment. The main results of the study show improved consistency of classification, a reduction in misclassification events, and increased robustness of the model to changes in the operating environment. Overall, the two-level approach demonstrated practical effectiveness and potential for deployment in modern network-security monitoring systems, confirming that the stated research objectives have been achieved.