EigenTrust

EigenTrust is a distributed trust metric for peer-to-peer networks that computes global reputation scores from local pairwise trust ratings. Rather than simply averaging direct ratings, EigenTrust treats trust as a network propagation problem: if Alice trusts Bob, and Bob trusts Carol, then Alice's trust in Bob partially confers trust in Carol — with attenuation at each hop. The algorithm converges to a stationary distribution over the endorsement graph, producing reputation scores that reflect not merely popularity but position within the trust topology.

EigenTrust was designed to solve the collusion problem in peer-to-peer file sharing, where malicious nodes might inflate each other's ratings. By weighting each node's ratings by its own global reputation, the algorithm makes Sybil attacks more costly: fake identities created for collusion start with zero reputation, and their endorsements carry negligible weight until they themselves accumulate trust from established nodes. The algorithm is not immune to attack — sophisticated adversaries can manufacture gradual reputation build-up — but it raises the cost of manipulation from trivial to substantial.

The mathematical structure of EigenTrust is identical to that of PageRank: both are eigenvector centrality measures on directed graphs. The difference is interpretive: PageRank measures importance through citation; EigenTrust measures reliability through endorsement. This structural similarity suggests that the problems of Sybil attacks and link farms that plague web search ranking also afflict distributed trust systems, and that solutions in one domain may transfer to the other.