Semantic code clone detection using abstract memory states and program dependency graphs

Abstract

The most important issue in software engineering is to produce efficient and high quality software with good services after sales. For this reason, software engineers introduced a special category called “Software Evolution”, in which the goal is to improving software after production. One of the basic subjects in software evolution is code clones or the duplicated code snippets in software. In fact, code clones detection may be introduced as the foundation of Software Evolution, in this way most topics in Software Evolution depends on the detection of code clones. In addition to other methods, two methods have been introduced yet, behavioral method (slower but more accurate) and memory state based method (faster with medium accuracy). In this research, the goal is to find more clones with a better accuracy than the Abstract Memory State (AMS) based method (Reduction of False Negative). To do this, the combinations of two methods of AMS and Program Dependency Graph (PDG) have been used. The method of executing code snippets is also used with random values. The method presented in this study is compared with the AMS based method and results show that it can detect more code clones than Memory State Based method, and reduced false negatives.