Annotation: The paper proposes a method for detecting malicious executable files by analyzing disassembled code. This method is based on static analysis of assembler instructions of executable files using a special neural network model, the architecture of which is also presented in this paper. In addition, through several different metrics, the effectiveness of the method has been demonstrated, showing a significant reduction of the second-order error compared to other state-of-the-art methods. The results obtained can be used as a basis for designing static malware analysis systems.

Keywords: detection of malicious software, static analysis, machine learning, deep neural networks, disassembled code analysis, transformer, BERT.