De-Anonymization of the Author of the Source Code Using Machine Learning Algorithms

The authors of the article consider the relevant problem of determining the author of the anonymous source code of program. The importance of these solutions for the field of computer forensics is due to the rapid increase in the number of cybercrimes and necessity to disclose authors of various viruses and cyberattacks. In addition, methods of determining the author of the source code can be used as a test for plagiarism of student programming works, which makes the study valuable for educational organizations. Finally, such decisions may be useful in resolving intellectual property and copyright disputes in a commercial environment. The most popular among researchers approaches to the identification the author of the source code for the last 5 years are described in the article. The authors propose approaches based as on classical machine learning algorithms as on modern architectures of deep neural networks for this purpose. Experiments were conducted out on the corpus includes the most popular and rapidly developing programming languages (just above 200 thousand samples). The effectiveness of these methods in comparison with their analogues is confirmed by the accuracy of more than 95% and independence from language in which the author's programs were written.