Annotation: The current trend of increasing labor productivity and efficien-cy of business processes entails optimization of software development pro-cesses through the use of generative artificial intelligence models trained on various code bases and manual copying of code fragments. Taking into ac-count the growing number of reported vulnerabilities, methods for detecting clones of program code are needed. In this paper, we propose a method for evaluating the similarity of fragments of the program code of binary executable files, which is based on the representation of the code in the form of an FA-AAST tree and the apparatus of graph neural networks. The results obtained during testing on open and closed source software demonstrate the correctness of the proposed method and higher accuracy in comparison with considered solutions

Keywords: code clone, AST, FA-AAST, graph neural network, cyberse-curity