| Abstract | We propose a novel game-theoretic framework foranalyzing the robustness of trust-inference protocols
in the presence of adversarial (but rational) users.
To the best of our knowledge, this is the first such
framework which simultaneously (1) admits a rigor-
ous and precise definition, thereby enabling formal
proofs of security (in various adversarial settings)
for specific trust-inference protocols; (2) is flexible
enough to accommodate a full range of (realistic) ad-
versarial behavior and network models; and (3) is ap-
propriate for decentralized networks, and in particu-
lar does not posit a trusted, centralized party with
complete knowledge of the system history. We also
show some preliminary results regarding the design
of trust-inference protocols which can be rigorously
proven secure within our model.
In addition to establishing a solid foundation for
future work, our framework also enables a rigorous
and objective comparison among existing trust inference protocols.
|
