TUSUR: Program Code Analysis System

2022: Announcement of the AI system

In December 2022, it became known about the creation at Tomsk State University of control systems and radio electronics (TUSUR) of an artificial intelligence (AI) system, which is able to determine the author of the program code by writing style. This technology will track the authors of viruses and help identify precedents for code theft.

Russian development, according to its creators, allows you to determine the individual "handwriting" of the programmer and distinguish between code that is written by a person or generated by a neural network. You can also install the author or neural network that created the code. Accuracy depends on input and programming languages, ranging from 76% to 87% (by December 2022). The system determines the code generated by the machine with an accuracy of 81.5%. It is also stated that the average accuracy of identifying source codes generated from commits is at 84%.

An AI system for analyzing program code was created in Russia

The technology developed by domestic scientists is a kind of analogue of handwriting - a section of forensic science dedicated to the handwriting of a person. Handwriting allows you to establish, for example, whether the signature on the document is fake, as well as identify the person who left this signature or wrote this or that text.

TUSUR scientists have published information about their development in the journal Future Internet. In their article, they state that their creation allows them to quickly resolve copyright disputes in the field of software, including commercial.

TUSUR says that Russian AI has already been included in the strategic development plan of Tomsk University until 2030. The researchers intend to analyze the compiled code of programs to increase the recognition of viruses and their authors.^[1]

Notes

Шаблон:Remarks