NSF Project III 1815247: Views of Data-Driven Business Processes: Foundations and Applications

PI: Victor Vianu

Students: Yuliang Li


In an organization governed by a complex business process, it is critical to provide stakeholders with views of the workflow, customized according to their role in the organization. Such views present the same workflow at various levels of abstraction that expose only the information relevant (or authorized) for a class of users. For example, an executive may require a high-level summary of a business process while a logistics engineer might need fine-grained control over actions involved in supply chain management. Workflow views are also needed to specify interactions with external business partners, to integrate different workflows, and to validate customizations of a workflow.

The objective of this project is to formally investigate specification mechanisms and semantics for views of data-driven workflows, to study techniques for view analysis and synthesis, and to develop view-based tools for the management of data-driven workflows.

The results of the project will benefit a wide variety of applications ranging from e-commerce to digital government to health- care and infrastructure for scientific applications. The project will contribute to the development of human resources by training graduate and undergraduate students in cutting-edge specification and static analysis techniques which are crucial to business processes. The project presents significant technical challenges. Analogously to database views, workflow views generally consist of hiding and restructuring some of the data and actions of the workflow. However, while database views are straightforwardly defined by queries, views of data-driven workflows must deal simultaneously with data and process. This requires operations whose expressivity goes beyond relational algebra's operators and results in a much richer landscape of specification mechanisms and semantics, which has not previously been explored. Moreover, since the workflow specifications include as first-class citizen the manipulation of data from an infinite domain, analysis tasks involve infinite-state systems. Static analysis of infinite-state systems is a notoriously challenging problem, to which classical techniques from general-purpose software verification cannot be easily applied, since these rely on finite-state abstractions. The project will bring into play techniques from logic, automata theory, complexity theory, algorithms, and automatic verification. It will develop formal foundations for views of data-driven workflows, including syntax and semantics for view specifications and consistency criteria connecting the view to the original workflow. It will produce tools for checking consistency and synthesizing guaranteed-consistent view specifications. It will yield a well-principled approach to integrating workflows from different organizational units, by extending the well-known local-as-view approach from data integration to workflow integration. Finally, the project will explore the use of views as abstraction mechanisms in automatic verification of business processes.