Data Mining Application: Attack Resistant Trust Metrics

Exploring the Applications of Attack Resistant Trust Metrics

Abstract

This Dissertation characterizes the space of trust metrics, under both the scalar assumption where each assertion is evaluated independently, and the group assumption where a group of assertions are evaluated in tandem. We present a quantitative framework for evaluating the attack resistance of trust metrics,and give examples of trust metrics that are within a small factor of optimum compared to theoretical upper bounds. We discuss experiences with a real world deployment of a group trust metric, the Advogato website. Finally, we explore possible applications of attack resistant trust metrics, including using it as to build a distributed name server, verifying metadata in peer-to-peer networks such as music sharing systems, and a proposal for highly spam resistant e-mail delivery.