Student Project Proposal:

Agent-oriented Framework for Decentralized E-Commerce using Reputation Mechanisms


Supervisors

Radu Jurca
Office: INR 240
Tel: +41-21-6936711
Email: radu.jurca@epfl.ch

Project Type

Diploma project for one or two students.

Description

Background

The availability of ubiquitous communication through the Internet is driving the migration of business transactions from direct contact between people to electronically mediated interactions. People interact electronically either through human-computer interfaces or through programs representing humans, so-called agents. In either case, no physical interactions among entities occur and the systems are much more susceptible to fraud and deception.

Traditional methods to avoid cheating, involving strong cryptography and Trusted Third Parties (TTP's) that overlook every transaction, are very costly and sometimes even impossible to apply due to the complexity and heterogeneity of the environment. The maintenance of the TTP's can incur substantial costs, and network communities often have a strong desire to be independent of any authorities, as illustrated by the success of P2P systems.

Reputation mechanisms offer a novel and efficient way of ensuring the necessary level of trust which is essential to the functioning of any market. They are based on the observation that agent strategies change when we consider that interactions are repeated: the other party will remember past cheating, and changes its terms of business accordingly in the future. In this case, the expected future gains due to future transactions in which the agent has a higher reputation can offset the loss incurred by not cheating in the present transaction. This effect can be amplified considerably if such reputation information is shared among a large population and thus multiplies the expected future gains made accessible by honest behavior. Reputation mechanisms define a new notion of "soft" security according to which misbehavior is not impossible, but economically irrational.

Existing reputation mechanisms enjoy huge success. Systems such as eBay or Amazon implement successful reputation mechanisms which are partly credited for the businesses' success. Studies show that buyers seriously take into account the reputation of the seller when placing their bids in online auctions and that despite the incentive to free ride, feedback is provided in more than half of the transactions on eBay.

Goal

The goal of this project is to design an agent-based framework for decentralized e-commerce applications in which reputation mechanisms are employed to generate trust. The basic scenario is following: Agents possess different resources which can be shared with, or sold to other agents. Agents constantly need to exchange or acquire different resources. For that, they need to be able to find the possessor of the resource they are looking for, they need to be able to negotiate a price for the desired resource, and they need to be able to exchange the resource. Tools for comparing the quality of different resources, for recording different transactions and for drawing market statistics also need to be considered.

A Reputation Mechanism will be integrated in this framework in order to replace/complement traditional security mechanisms. Agents are able to check the reputation (i.e. information about past behavior) of a partner agent, and submitt feedback after every interaction. Clear guidelines should also be provided in order to help agents use the reputation information for making decisions.

Four particular markets will be implemented:

  1. A P2P File Sharing Service. Agents possess different files which can be shared with other agents. Agents should be able to find the possessor of the desired file and request the download. Agents need to trust the quality and the availability of the shared resource, media files in this case. The risk lies in not being satisfied with the downloaded file, loose more time and energy to find another provider, spend some more on downloading costs. Reputation information is based on the quality of service and availability. No exchange of money occurs.
  2. A Decentralized Electronic Auction Marketplace. Agents possess goods that can be sold to other agents through an auction (of varying type). Goods have different characteristics. Agents should be able to participate in auctions of different types. Payment procedures need to be implemented. The buyer agent needs to trust that the seller will correctly describe the product offered for sale and that he will cooperate and actually ship the product. Reputation information is based on honesty (correctly reporting the attributes of the product), predictability (degree to which the agent will behave as predicted), and integrity (agent makes deals in good faith). On the other hand, the reputation mechanism itself must be trustworthy in order to ensure the proper functioning of the market.
  3. A marketplace for web services. Agents can deliver services for which they are paid. Requesting agents should be able to find the provider of the service they need, and should be able to negotiate the conditions of the service. A payment system is also needed. Consumers need to trust that service providers will offer the promised service at the promised quality level. Reputation information is context-dependent, i.e. depends on the type of service requested.
  4. A marketplace for financial information. Has the same characteristics as the file sharing market, just that agents have to pay for the information. Decision makers need to trust that the information on which they base their decisions is correct. A particularity of this application is that conflicts of interests among organizations may strategically influence the quality of provided information

Remarks

This project will be part of a bigger project that studies reputation-based security for open distributed markets. The results of this project will be used for empirically evaluating different reputation mechanisms developed at LIA. However, the goal of the project is much more general. This framework could constitute a basis for a new generation of decentralized e-commerce applications.

Skills

Java, English

Environment

OS independent

Radu Jurca
Last modified: Thu May 13 2004