Evaluating Workflow Trust Using Hidden Markov Modeling and Provenance Data

Abstract

In service-oriented environments, services with different functionalities are combined in a specific order to provide higher-level functionality. Keeping track of the composition process along with the data transformations and services provides a rich amount of information for later reasoning. This information, which is referred to as provenance, is of great importance and has found its way into areas of computer science such as bioinformatics, database, social, sensor networks, etc. Current exploitation and application of provenance data is limited as provenance systems have been developed mainly for specific applications. Therefore, there is a need for a multi-functional architecture, which is application-independent and can be deployed in any area. In this chapter we describe the multi-functional architecture as well as one component, which we call workflow evaluation. Assessing the trust value of a workflow helps to determine its rate of reliability. Therefore, the trustworthiness of the results of a workflow will be inferred to decide whether the workflow’s trust rate should be improved. The improvement can be done by replacing services with low trust levels with services with higher trust levels. We provide a new approach for evaluating workflow trust based on the Hidden Markov Model (HMM). We first present how the workflow trust evaluation can be modeled as a HMM and provide information on how the model and its associated probabilities can be assessed. Then, we investigate the behavior of our model by relaxing the stationary assumption of HMM and present another model based on non-stationary hidden Markov models. We compare the results of the two models and present our conclusions.