Exploring developer patterns and code similarities: an in-depth study using code stylometry, developer profiling, and clone detection in software projects

Abstract

In recent years, code theft has become a significant threat to the software industry. Incidents of code being stolen and used anonymously are increasingly common, alongside other security challenges such as plagiarism, copyright disputes, software piracy, and unauthorized, anonymous contributions on platforms like GitHub. This paper presents a clear pipeline that uses code writing style, developer profiles, and a hybrid clone detector called CodeCloneX to find who wrote code and where code was reused. The dataset comprises source code samples from 10,000 prominent GitHub contributors, obtained through GitHub mining. We extracted simple filelevel style features, then made per-developer profiles by averaging those features. We compared three clustering methods: KMeans, Agglomerative, and BIRCH and our clone detector can detect clones of about 90% of lines of code and 57% of files. These results show that combining style profiles with clone analysis helps detect likely authors and reused code in large codebases. By providing a framework for recognizing coding styles and optimizing team collaboration, the findings contribute to improved code readability, maintainability, and software project management.