Analysis of Source Code Authorship Attribution Problem
A. D. Bogdanova, Mirko Farina, Zamira Kholmatova, Artem V. Kruglov, Vitaly Romanov, Giancarlo Succi · 2022
Source Code Authorship Attribution (SCAA) has become very important for the functioning of our societies. For example, it is central in copyright and plagiarism issues, for detecting authors of malware, and even in recruitment and selection processes. The goal of this review is to analyze existing approaches to SCAA, compare them, and identify the most common feature types and architectures of neural networks underlying them. We identified the most common taxonomy of the feature types in SCAA. These are: a. lexical, b. layout, and c. structural. We also found that the combination of the lexical (most language agnostic) and structural (most language dependent) features usually provides the most accurate results.