Когнітивний кіберполігон на основі побудови двосторонньої семантичної мережі

Scenario analysis in a cyber range requires a structured representation of infrastructure and threat knowledge. This paper proposes a cognitive cyber range model based on two-sided semantic network construction and examines its graph component using a software prototype. The theoretical model combines forward and backward expansion, scenario synthesis, and multicriteria assessment. The prototype represents document-level concept co-occurrence as an undirected contextual semantic network, weights its links using normalized pointwise mutual information, and generates ranked paths for subsequent interpretation by a large language model. Three search configurations are evaluated on 20 test scenarios of cyber intrusions into infrastructure. Mean reference-step recall across up to five paths is 0.6083 for bidirectional search performed separately for each supplied initial concept and 0.2042 for one-sided search without supplied initial concepts. These values compare configurations that differ in both input information and traversal procedure; they do not isolate the effect of the search algorithm. Recall of the first-ranked path, preservation of step order, and node precision are assessed separately. Literal identifier grounding before registry validation supports the preparation of source documents. The prototype supports traceable exploration of scenario hypotheses. Causal consistency, action feasibility, numerical multicriteria assessment, and synchronization with live infrastructure remain outside the validated implementation and require further research.