TRUTHFUL MECHANISMS FOR BUILDING TRUST IN E-COMMERCE
A fundamental issue for a real uptake of commercial transactions over the web regards trust among the transacting entities, frequently unknown to each other. One solution to increase confidence in transactions is to use a network of TSPs (Trust Service Providers), called a trust web, which are third parties known and trusted by both entities, and an algorithm that establishes a trust path before carrying-out any e-commerce transaction.
In this paper we study the problem of building trust paths linking an entity initiating a transaction to a set of final merchants in a trust web from a “mechanism design” point of view. Namely, we consider TSPs as strategic agents which respond to incentives and may deviate from the protocol for a tangible gain. A truthful mechanism should define both the protocol and a suitable payment rule such that each agent maximizes her own utility when not deviating from the protocol, regardless of what other agents do.
We first address the problem from a “protocol design” perspective and, assuming that TSPs are honest/obedient, we propose a distributed search algorithm based on a probabilistic trust degree model [Mau96, DIM02] which generalizes that based only on boolean trust relationships proposed in [Ati02], and reduces the search space complexity by pruning the alternatives that do not satisfy (besides cost constraints) trust degree constraints (e.g., the “transitive” degree of trust accumulated along the path has to be greater than a given threshold value).
Then, when considering TSPs as strategic agents, we use this algorithm as a substrate to define truthful mechanisms for building suitable trust paths. Indeed, the main scope of this paper is to provide an answer to the following fundamental problem: does a payment function exist for the described problem such that the the resulting mechanism is truthful? By applying recent results appeared in [MPPW+04] we provide both positive and negative answers, depending on which constraint we add/drop and on which parameters are considered as a private information of agents.