ANALYTICAL WEB SERVICE FOR IDENTIFYING SUSPICIOUS HIGH-RISK DIGITAL ASSET TRANSACTIONS

Abstract

The rapid growth of digital asset transactions significantly complicates their analysis and monitoring. Although blockchain provides transparency of operations, the high level of pseudonymity among participants creates substantial challenges in identifying the nature of interactions and detecting potentially undesirable activity. This increases the demand for modern tools capable of automatically processing large volumes of data, grouping addresses into clusters, and evaluating their behavior based on aggregated transactional characteristics. The development of a web service for digital asset transaction analytics that automatically forms address clusters and classifies them using machine learning models is proposed. The system provides an informative and interpretable data presentation, enabling users to assess the nature of activity associated with a given address. The methodology is based on heuristics for clustering blockchain addresses in networks utilizing the UTXO model, as well as on the calculation of structural and behavioral characteristics of the formed clusters. Random Forest, Extra Trees, and XGBoost models were used for classification, which were trained on a labeled subset of the scientific dataset. The technical implementation of the web service is based on the Symfony framework for the server side, the React library for the client side, MySQL and PostgreSQL DBMS for data storage, the Python programming language for machine learning, as well as Docker and Nginx tools for deployment and stability. The results of the study demonstrate that machine learning models can effectively classify clusters of digital asset addresses according to their behavioral characteristics, and the integration of algorithms into the web service provides automatic generation of analytical reports. The scientific novelty lies in combining heuristics for address clustering with machine learning models, which allows evaluating the behavior of clusters and identifying potential risk. The practical significance of the developed web service is defined by the possibility of its application for rapid preliminary checks of cryptocurrency addresses, detecting relationships between them, and assessing potential interaction risks, making it valuable for organizations and individual users alike.