Using software birthmarks and clustering to identify similar classes and major functionalities

Abstract

Software birthmarks are a class of software metrics designed to identify copies of software. An article published in 2006 examined additional applications of software birthmarks. The article described an experiment using software birthmarks to identify similar classes and major functionalities in software applications. This study replicates and extends that experiment, using a modern software birthmark tool and larger dataset, while improving the precision of the research questions and methodologies used in the original article. We found that one of the conclusions of the original article could be replicated while the the other conclusion could not. While software birthmarks provide an effective method for identifying similar class files, they do not offer a reliable, objective, and generalizable method for finding major functionalities in a software release.