Model Driven Mutation Applied to Adaptative Systems Testing

TL;DR

This paper introduces a mutation testing approach for adaptation logic in dynamically adaptive systems, using a model-driven engineering framework to classify faults and generate mutants, with preliminary results indicating testing challenges.

Contribution

It proposes a language-independent fault model for adaptation logic, utilizing model-driven engineering to facilitate mutation testing in adaptive systems.

Findings

Mutants are difficult to kill in adaptive systems.

Testing adaptive systems is more complex than traditional systems.

Preliminary results highlight challenges in mutation testing for adaptation logic.

Abstract

Dynamically Adaptive Systems modify their behav- ior and structure in response to changes in their surrounding environment and according to an adaptation logic. Critical sys- tems increasingly incorporate dynamic adaptation capabilities; examples include disaster relief and space exploration systems. In this paper, we focus on mutation testing of the adaptation logic. We propose a fault model for adaptation logics that classifies faults into environmental completeness and adaptation correct- ness. Since there are several adaptation logic languages relying on the same underlying concepts, the fault model is expressed independently from specific adaptation languages. Taking benefit from model-driven engineering technology, we express these common concepts in a metamodel and define the operational semantics of mutation operators at this level. Mutation is applied on model elements and model transformations are used to propagate these changes to a given adaptation policy in the chosen formalism. Preliminary results on an adaptive web server highlight the difficulty of killing mutants for adaptive systems, and thus the difficulty of generating efficient tests.