Learnlib Framework for Extrapolating Behavioral Models

Organisation:	Dortmund Technical University, Potsdam University (GERMANY) Uppsala University (SWEDEN)
Method:	DFA (Deterministic Finite Automata) Mealy machines
Tools used:	CADP (Construction and Analysis of Distributed Processes)
Domain:	Communication Protocols.
Period:	2009
Size:	n/a
Description:	Telecommunications protocol specifications are often defined in natural language text documents, and there is no formal link between the specification and the implementation, making any formal validation or model checking difficult. Automata learning techniques have been proposed to overcome this situation, by allowing one to construct and later update behavioral models automatically. LearnLib is a framework of tools for automata learning, which is explicitly designed for the systematic experimental analysis of the profile of available learning algorithms and corresponding optimizations. This case study examines the use of automata learning by means of LearnLib, to construct behavioural models tailored to specific contexts and learning scenarios. In particular, the study uses three protocols from the CADP demos, applying LearnLib to "learn" each protocol behaviour, and experimenting with various options for optimization and tailoring.
Conclusions:	The experimental results show the value of using tailoring in automata learning algorithms, including in one example an increase in performance of one order of magnitude. Though this case study involved telecommunications protocols, it demonstrates an approach that could be applied to the more general setting of reactive systems that are component-based, addressing the problem of under-specification.
Publications:	[Raffelt-Steffen-Berg-Margaria-09] Harald Raffelt, Bernhard Steffen, Therese Berg, and Tiziana Margaria "LearnLib: a framework for extrapolating behavioral models". International Journal on Software Tools for Technology Transfer (STTT) Volume 11, Number 5, 393-407, Springer, 2009. Available on-line at: http://www.springerlink.com/content/7581g372n82206v4/ or from the CADP Web site in PDF or PostScript
Contact:	Bernard Steffen Chair of Programming Systems, TU Dortmund, Otto-Hahn-Str. 14 44227 Dortmund, Germany Email: steffen@uni-dortmund.de
Further remarks:	This case study, amongst others, is described on the CADP Web site: http://cadp.inria.fr/case-studies

Back to the CADP case studies page