Quantitative workflow resiliency

Abstract

A workflow is resilient when the unavailability of some users does not force to choose between a violation of the security policy or an early termination of the workflow. Although checking for the resiliency of a workflow is a well-studied problem, solutions usually only provide a binary answer to the problem, leaving a workflow designer with little help when the workflow is not resilient. We propose in this paper to provide instead a measure of quantitative resiliency, indicating how much a workflow is likely to terminate for a given security policy and a given user availability model. We define this notion by encoding the resiliency problem as a decision problem, reducing the finding of an optimal user-task assignment to that of solving a Markov Decision Process. We illustrate the flexibility of our encoding by considering different measures of resiliency, and we empirically analyse them, showing the existence of a trade-off between multiple aspects such as success rate, expected termination step and computation time, thus providing a toolbox that could help a workflow designer to improve or fix a workflow. © 2014 Springer International Publishing Switzerland.