Cliche recognition in legacy software: a scalable, knowledge-based approach

Many software reverse engineering techniques that are sufficiently "light-weight" (i.e. computationally inexpensive) to be able to work on large systems tend to compute syntactic information that, while useful, does not capture the meaning of the program. At the same time, many "heavy-weight" (i.e. computationally expensive) techniques that compute information in terms of human strategies hidden in the software tend not to be efficient enough to work on large real-world systems. We are working on applying a heavy-weight technique of program cliche recognition to the real-world problem of software reverse engineering. This paper presents our approach to program cliche recognition and focuses on issues of scalability, robustness and human-system interaction. We demonstrate the approach by describing how it can be applied to the reverse engineering of a real-world software system.