Ellana Molchanova

Dr. Ellana Molchanova


Research Focus & Metrics

Academic Career Length

1Year (2026-Present)

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

Representative Publications

View All Publications
1 October 2026 in Information
Verified
Journal Article
Open access
PDF file
YKYuliia Kostiuk
Yuliia Kostiuk
EMEllana Molchanova
Ellana Molchanova
Pavlo M. SkladannyiPavlo M. Skladannyi
Pavlo M. Skladannyi
VSVolodymyr Sokolov
Volodymyr Sokolov
KKKaryna Khorolska
Karyna Khorolska
This paper proposes a semantically oriented knowledge-representation method for transforming primary cyber-event data into structured semantic knowledge to support explainable reasoning in cybersecurity decision-making. The core contribution is a transformation mechanism from primary semantic information (PSI) to secondary semantic information (SSI), based on a frame model of network events and attack patterns and followed by rule-based logical inference. This inference process maps semantic predicates to MITRE Adversarial Tactics, Techniques, and Common Knowledge (MITRE ATT&CK) techniques and reconstructs causal attack chains. The method is implemented within an Information-Semantic System (ISS) and integrated with explainable artificial intelligence (XAI), thereby ensuring that each decision can be traced to the rules and semantic relationships that produced it. Evaluation on the CICIDS2017 dataset, using a binary classification task (Normal vs. Attack), a stratified 60/20/20 train–validation–test split, and five experimental runs, yielded a precision of 0.94, a recall of 0.91, and an F1-score of 0.92. These results are comparable to those achieved by strong neural-network-based baselines, while the proposed method provides substantially greater interpretability and decision traceability. In two anonymised operational SOC/SIEM environments, the method reduced incident triage time by 22–38% and the false-positive rate by 18–34% compared with a baseline machine learning-only pipeline. The paper further examines transferability across additional datasets, presents ablation analyses, incorporates short-term-memory (STM)-based temporal context, and discusses explainability mechanisms and alignment with relevant cybersecurity standards.

17 June 2026 in Preprints.org
Verified
Preprint
Open access
PDF file
YKYuliia Kostiuk
Yuliia Kostiuk
EMEllana Molchanova
Ellana Molchanova
Pavlo M. SkladannyiPavlo M. Skladannyi
Pavlo M. Skladannyi
VSVolodymyr Sokolov
Volodymyr Sokolov
KKKaryna Khorolska
Karyna Khorolska
The article proposes a semantically oriented method for representing and processing knowledge in artificial intelligence systems for cybersecurity tasks. The method is based on a frame model that provides a structured rep-resentation of network events, threats, and attack patterns, and supports context-dependent decision-making by forming explainable conclusions. A logical inference mechanism based on rules and semantic generalisation opera-tions has been implemented, supplemented by a short-term memory (STM) module that accounts for the temporal characteristics of events within a sliding observation window. This allows explicit matching of semantic predicates (SSI) with MITRE ATT&CK techniques and accounts for the temporal characteristics of events with the formation of reproducible cause-and-effect attack chains. To improve interpretability, the results of logical inference are inte-grated with explainable artificial intelligence (XAI) methods, in particular SHAP and LIME, which ensure trans-parency of decisions at the event and attack-scenario levels. The scientific novelty lies in combining semantic knowledge structuring with the temporal context of STM and explainable rule-based reasoning in a single incident analysis pipeline. Experimental validation was performed on the CICIDS-2017 benchmark datasets with additional portability testing on KDD Cup 99 and NSL-KDD, confirming competitive accuracy and high explainability. The practical applicability of the method was demonstrated in anonymised production SOC environments with real telemetry flows (XDR/EDR, network sensors, IAM logs), resulting in reduced triage time and fewer false positives. The proposed approach complies with the requirements of international standards ISO/IEC 27001 and NIST SP 800-53, supports integration with the MITRE ATT&CK knowledge base and the Zero Trust architecture, and pro-motes the development of intelligent agents for reliable, explainable decision-making in cybersecurity.