Комплексний алгоритм профілювання кіберзлочинців методами графової та блокчейн-аналітики

Abstract

The presented abstracts explore and substantiate an innovative algorithmic model for automated collection, multilevel normalisation, and deep cross-platform analysis of digital traces, aimed at comprehensive de-anonymisation of modern cyber threats. A hybrid analytical pipeline has been developed that integrates natural language processing methods (in particular, named entity recognition and vector representation of words) for semantic parsing of unstructured text from shadow forums and closed messengers. Particular attention is paid to the application of weighted multigraph theory for the mathematical calculation of transitive links between isolated communication profiles, as well as the implementation of heuristic algorithms for clustering UTXO transactions in blockchain networks for the deobfuscation of financial flows. The practical application of the proposed mathematical model provides a transition from fragmented indicators of compromise to the formation of legally significant comprehensive profiles of cybercriminals. This creates a scientifically sound basis for the further legal blocking of illegal virtual assets on centralised cryptocurrency exchanges and the initiation of criminal proceedings within the framework of international law enforcement cooperation.