Inferring compact models of communication protocol entities

Abstract. Our overall goal is to support model-based approaches to verification and validation of communication protocols by techniques that automatically generate models of communication protocol entities from observations of their external behavior, using techniques based on regular inference (aka automata learning). In this paper, we address the problem that existing regular inference techniques produce "flat" state machines, whereas practically useful protocol models structure the internal state in terms of control locations and state variables, and describes dynamic behavior in a suitable (abstract) programming notation. We present a technique for introducing structure of an unstructured finite-state machine by introducing state variables and program-like descriptions of dynamic behavior, given a certain amount of user guidance. Our technique groups states with "similar control behavior" into control locations, and obtain program-like descriptions by means of decision tree generation. We have applied parts of our approach to an executable state machine specification of the Mobile Arts Advanced Mobile Location Center (A-MLC) protocol and evaluated the results by comparing them to the original specification