Static Backward Program Slicing for Safety-Critical Systems

Abstract

Static program slicing is a technique to detect the program parts (i.e. the “slice”) of the given program possibly affecting a given property. The technique is of interest for analysing safety-critical software, since it can identify the program parts that may affect various safety properties. Verification efforts can then be directed towards those parts, leading to a more efficient verification process.

We have developed a novel method for static backward program slicing. The method works for well-structured programs, as commonly demanded by coding standards for safety-critical software. It utilises the program structure to obtain a highly efficient slicing process, where control dependencies are inferred from the program structure, and the slicing is done on-the-fly concurrently with the data dependence analysis.

We have evaluated our method experimentally. For applications that require few slices to be taken, like checking for a set of safety properties, we obtain large speedups as compared with the standard method for static backward program slicing. We have also investigated how the speedup varies with various parameters such as code size, size of the slice relative to the full program, and relative frequency of conditions in the code.