Lightweight Parsing and Slicing for Bug Identification in C

Abstract

Program slicing has been used to semi- or fully-automatically help developers find errors and vulnerabilities in their programs. For example, Dashevskyi et al. (IEEE TSE 2018) introduced a lightweight slicer for Java that can be used for vulnerability analysis. However, a similar lightweight slicer for C/C++ is still missing. In this work we propose a comparison method for parsers, evaluate it on two commonly-used parsers, and develop a lightweight slicer for C/C++ using the “better” parser from our comparison. From our evaluation, the Joern parsing method (island grammar) could parse non-standard C/C++ code but its resulting structure may contain semantic errors that can affect subsequent analysis. ANTLR4 is faster in returning a result, and when manually cleared of non-standard C/C++ codes, it is more accurate than Joern. We then built our C/C++ thin slicer extension using ANTLR4, and we observed that it is promising from both precision and performance perspectives. As a future work, we plan to improve the logic behind processing pointers. In particular, we consider doing deeper pointer analysis.