HYPpOTesT: Hypothesis Testing Toolkit for Uncertain Service-based Web Applications

EasyChair Preprint 2255

Abstract

This paper introduces a model-based testing framework and associated toolkit, so called HYPpOTesT, for uncertain service-based web applications specified as probabilistic systems with non-determinism. The framework connects input/output conformance theory with hypothesis testing in order to assess if the behavior of the application under test corresponds to its probabilistic formal specification. The core component is a (on-the-fly) model-based testing algorithm able to automatically generate, execute and evaluate test cases from a Markov Decision Process specification. The testing activity feeds a Bayesian inference process that quantifies and mitigates the system uncertainty by calibrating probability values in the initial specification. This paper illustrates the structure, features, and usage of HYPpOTesT using the U-Store exemplar, i.e., a web-based e-commerce application that exhibits uncertain behavior.

Keyphrases: Bayesian inference, model-based testing, probabilistic systems, service-based systems, uncertainty quantification

@booklet{EasyChair:2255,
  author    = {Matteo Camilli and Angelo Gargantini and Rosario Madaudo and Patrizia Scandurra},
  title     = {HYPpOTesT: Hypothesis Testing Toolkit for Uncertain Service-based Web Applications},
  howpublished = {EasyChair Preprint 2255},
  year      = {EasyChair, 2019}}