Usage patterns: Extracting system functionality from observed profiles

A technique for reverse engineering software systems has been developed that identifies the functionality of a system through patterns of recurring software modules. This technique is different from its predecessors in that it uses run-time information and seeks to identify the principal functionalities of the software. Each time the system is run the set of transitions that executed is collected in an observed profile. Each observed profile is examined and patterns of transitions are extracted. The patterns that are identified are called usage patterns. These usage patterns can be used to identify the principal functionalities of a software system as well as to determine the probability of execution associated with each of these functionalities. This is possible due to properties of the software calling structure and the relationship between software modules and functionalities. The technique has been applied to three different software systems and the results of the usage pattern extraction process are presented. Usage patterns have application in many aspects of software development, including design verification, reliability modeling, security and efficiency enhancement activities.