scat: Learning from a single execution of a binary

Abstract

Retrieving information from a binary code is required in several application domains such as system integration or security analysis. Providing tools to help engineers in this task is therefore an important need. We present in this paper scat, an open-source toolbox, relying on lightweight runtime instrumentation to infer source-level and behavioral information from a binary code, like function prototypes or data-flow relations. We explain the functioning principle of this toolbox, and we give some results obtained on real examples to show its effectiveness.