Information extraction by type analysis

This thesis investigates an alternative use of type reconstruction, as a tool for extracting knowledge from programs written in weakly typed language. We explore this avenue along two different, but related directions. In the first part we present a static analyzer that exploits typing techniques to extract information from the COBOL source code: reconstructing informative types is an effective way for automatically generating a basic tier of documentation for legacy software, and is also a reliable starting point for performing further, higher-level program understanding processing. In the second part of we apply similar principles to an apparently distant context: validating inter-component communication of Android applications by reconstructing the types of data within Intents - the building blocks of message passing in Android. Both for COBOL and Android, we present a distinct implementation of the static analysis system proposed.