Due diligence through risk analysis and belief propagation over provenance.

Abstract

In many domains, the concept of due diligence is defined as taking reasonable care to protect something from unforeseen problems. The concept of due diligence is expected to be demonstrated to avoid undesired events or to anticipate (un)expected events. Moreover, demonstrating due diligence can mean complying with regulations and guidance. Thus, organisations can avoid a penalty in case something bad happens on the basis that they have taken reasonable precautions according to the applicable regulations and guidance. The importance of due diligence and its implications in various domains motivates this research to develop a general scientific and systematic approach to demonstrate due diligence. Our claim is that scientific approaches should be utilised to develop an approach as they are usually proven to be valid and based on strong evidence. Set against this, since due diligence is contextual, it is challenging to develop a general approach that applies across multiple domains. In developing our general approach to demonstrate due diligence, we systematically combine several scientific approaches into a framework called prFrame. These approaches are Provenance, Risk, and the Probabilistic Graphical Model (PGM), and it is these that comprise the three central pillars in constructing prFrame. We situate our research into the process of demonstrating due diligence in the general business product supply chain, where multiple product operators and authorities are expected to demonstrate due diligence. Indeed, our discussion with them suggests that they are continuously looking for a better approach to demonstrate due diligence. The first pillar in prFrame is provenance. We consider provenance to be the most important aspect in prFrame since it underlies the other two pillars. Provenance is understood as a piece of historical information to explain how something is derived, who responsible for any changes to it, and what service is used to make those changes. The second pillar is risk, which can be described as a chance of undesired consequences. Finally, PGM constitutes the last pillar in prFrame on account of its role as the principle means of risk propagation in the provenance-based product supply chain. To evaluate prFrame, we perform a set of exercises, as follows: 1) develop an ontology to conceptualise the domain of interest, 2) model the supply chain with the ontology 3) assess and infer the risk along the product supply chain, 4) establish experiments to evaluate the result of inference. A specific ontology is developed as an extension of a general ontology to model provenance, PROV-O (Ontology), so as to capture the risk along the product supply chain. The product supply chain itself is constructed based on the provenance of the product. The notion of risk that is captured comprises set risk models and risk factors, which are the main properties to assess risk quantitatively. The inference of risk is done by performing Belief Propagation as an inference technique based on the provenance of a product. Finally, a set of experiments is conducted to v evaluate the intuitive of the propagation result and its accuracy in the linear and nonlinear product supply chain. Our contributions in this research are therefore, 1) Development of an ontology to model the provenance of a product and map other legacy ontologies, 2) Create a risk model to overlie Provenance Graphs (PGs), 3) Establish a provenance-based description as a basis for Monte-Carlo simulations, 4) Develop a systematic provenance-based factorisation to allow a Belief Propagation technique, 5) Conduct a systematic evaluation for accuracy of states by Belief Propagation technique.

More information

Identifiers

ORCID for Belfrit Victor Batlajery: orcid.org/0000-0001-7637-9481

ORCID for Mark Weal: orcid.org/0000-0001-6251-8786

Catalogue record

Export record