An interval type-2 fuzzy logic based framework for reputation management in Peer-to-Peer e-commerce

During the last two decades, the Internet has changed people's habits and improved their daily life activities and services. In particular, the emergence of e-commerce provided manufactures and vendors with more business opportunities. This allowed customers to benefit from a global, quicker and cheaper shopping environment. However, e-commerce is evolving from a centralised approach, where consumers directly purchase products and services from businesses, to a Peer-to-Peer (P2P) perspective, in which customers buy and sell goods amongst themselves. In P2P scenarios, it is crucial to protect both buyers and sellers (the peers) from being victimised by possible fraud arising from the uncertainties, vagueness and ambiguities that characterise the interactions amongst unknown business entities. For this reason, the so-called reputation models are becoming a key architectural component of any e-commerce portal. These systems are intended to evaluate the basic features of each entity (buyer, seller, goods, etc.) involved in a given trading transaction in order to assess the trust level of the given transaction and minimise fraud. However, in spite of their wide deployment, the reputation models need to be enhanced to handle the various sources of uncertainties in order to produce more accurate outputs which will allow to increase the trust and decrease the fraud levels within e-commerce systems. In this paper, we present an interval type-2 fuzzy logic based framework for reputation management in (P2P) e-commerce which is capable of better handling the faced uncertainties. We have carried out various experiments based on eBay®-like transaction datasets which have shown that the proposed type-2 fuzzy logic based system can provide better performance (in terms of malicious peer detection and exchanged message overhead) when compared to the other well-known and heavily used approaches like the eBay® approach, EigenTrust, PeerTrust as well as the type-1 fuzzy based counterpart approach. © 2015 Elsevier Inc.