Authorship attribution (i.e., determining who is the author of a piece of source code) is an established research topic. State-of-theart results for the authorship attribution problem look promising for the software engineering field, where they could be applied to detect plagiarized code and prevent legal issues. With this article, we first introduce a new language-agnostic approach to authorship attribution of source code. Then, we discuss limitations of existing synthetic datasets for authorship attribution, and propose a data collection approach that delivers datasets that better reflect aspects important for potential practical use in software engineering. Finally, we demonstrate that high accuracy of authorship attribution models on existing datasets drastically drops when they are evaluated on more realistic data. We outline next steps for the design and evaluation of authorship attribution models that could bring the research efforts closer to practical use for software engineering. CCS CONCEPTS• Software and its engineering → Software maintenance tools; Software verification and validation; • Security and privacy → Malware and its mitigation.

No abstract

Authorship attribution of source code has been an established research topic for several decades. State-of-the-art results for the authorship attribution problem look promising for the software engineering field, where they could be applied to detect plagiarized code and prevent legal issues. With this study, we first introduce a language-agnostic approach to authorship attribution of source code. Two machine learning models based on our approach match or improve over state-of-the-art results, originally achieved by language-specific approaches, on existing datasets for code in C++, Python, and Java. After that, we discuss limitations of existing synthetic datasets for authorship attribution, and propose a data collection approach that delivers datasets that better reflect aspects important for potential practical use in software engineering. In particular, we discuss the concept of work context and its importance for authorship attribution. Finally, we demonstrate that high accuracy of authorship attribution models on existing datasets drastically drops when they are evaluated on more realistic data. We conclude the paper by outlining next steps in design and evaluation of authorship attribution models that could bring the research efforts closer to practical use.