Detecting Similar Classes and Methods Using Program Execution Traces

Abstract

Recently, plagiarism of source code has increased. It can be categorized into whole and partial plagiarisms. For the identification of partial plagiarism, code clone detection techniques can be used because they can detect duplicated parts from two software systems. However, obfuscation techniques allow obscuring plagiarism by rewriting software to make it difficult to understand. In this paper, we propose an approach to detecting pairs of similar classes and methods based on the similarity of execution traces that are resistant to obfuscation. This approach divides an execution traces into phases, and then match them based on the similarity of method calls for the detection of pairs of similar classes and methods. In case study, the approach identified the correspondences of original and obfuscated components in actual applications.